# Start a 1-Person Business with Claude (4 HOUR COURSE 2026)

https://www.youtube.com/watch?v=mKyaNr3jK-E
Translation: zh-TW

[00:00] If I were me right now and graduating

[00:02] college, I would feel like the luckiest

[00:03] kid in all of history.

[00:04] >> Why?

[00:08] >> Wake up. AI is here. And if you assume

[00:11] any rate of improvement over any

[00:13] reasonable time period, learning how to

[00:15] use AI should become your number one

[00:18] priority.

[00:18] >> I think it is possible now to start a

[00:21] company that is a oneperson company that

[00:22] will go on to be worth more than a

[00:24] billion dollars. The New York Times just

[00:26] ran a story about a guy who built a

[00:27] billion dollar company with AI. We all

[00:29] knew it would happen. One person

[00:31] creating a $1 billion company.

[00:33] >> You have access to tools that can let

[00:34] you do what used to take teams of

[00:36] hundreds. And you just have to learn how

[00:38] to use these tools and come up with a

[00:39] great idea.

[00:41] >> You probably heard it about a thousand

[00:44] times. Oh, AI is going to make you rich.

[00:47] Oh, I just made Claude code wipe my or I

[00:50] just replaced my entire family with

[00:52] Claude Code. There's a bunch of those

[00:54] videos out there and they have taken the

[00:55] internet by storm. And I know what

[00:57] you're thinking. No, another one of

[00:59] these drop shipping NFT get-richqu

[01:01] schemes. But that is not what this

[01:03] course is about. And if you're looking

[01:05] for the next getrichqu thing, then you

[01:08] might as well just click off this video

[01:10] cuz this video is not for you. And I

[01:11] know that because I actually thought the

[01:13] exact same when I got into this space a

[01:15] little over 3 years ago. If you don't

[01:17] know me, my name is Albert and I am by

[01:20] no means the smartest, but I still

[01:22] somehow managed to build two AI

[01:24] companies that combined have done over a

[01:27] million dollars, which is weird cuz I

[01:30] don't have a university degree in

[01:32] computer science. I actually never went

[01:33] to university. So, I'm kind of on my

[01:35] what, like seventh gap year or

[01:37] something. The only reason that this was

[01:39] possible was because of AI. And I give

[01:41] full credit to Chad GBT and Claude like

[01:44] 100%. I couldn't have done that without

[01:46] these tools. But let me show you how

[01:48] this was even possible. You might have

[01:50] seen this graph going around. Each dot

[01:53] on this graph represents 3.2 million

[01:56] people. So in total, we have 2,500 dots,

[02:00] which amounts to 8.1 billion humans. And

[02:02] what this shows is how many people in

[02:05] 2026 that have even used AI. And what

[02:08] you will find is that 84% of people, 84%

[02:13] which is 6.8 8 billion people have never

[02:16] used AI in their life. They have never

[02:18] even given chatb a regular prompt, which

[02:21] is hard to believe because if you're

[02:23] like me, you see AI everywhere. Every

[02:25] time you open your phone, you hear AI

[02:27] news, but that is because your algorithm

[02:28] knows that you're interested in that

[02:30] stuff. 16% of people have used free

[02:33] chatbots. So that is free chat GBT or

[02:36] free claude. That is 1.3 billion people.

[02:39] But if you are just a little into AI,

[02:41] you know that the free models, you

[02:42] almost can't do anything with those. And

[02:44] then if we zoom in, we have the people

[02:46] that pays $20 a month for AI. That is 25

[02:51] million people represented by these

[02:53] small yellow squares right here. 0.3% of

[02:57] the population. And my guess is that's

[02:59] probably where you are right now. Or

[03:01] maybe you are part of the 0.04%

[03:04] which is the max users. So the people

[03:06] using the most powerful models like

[03:09] claude code and codeex that is only 3.6

[03:12] million people or 0.04%.

[03:15] The reason that I'm showing you this is

[03:17] that I want to show you how far behind

[03:19] the world still is when it comes to AI.

[03:22] Many say that AI is going to be the

[03:24] greatest productivity boost that the

[03:26] world has ever seen. And even then after

[03:28] AI has existed for a couple of years

[03:30] now, 84% of people has still never even

[03:33] used it. I hope you see how this creates

[03:35] a massive opportunity because this

[03:38] allows us, the 0.04%,

[03:41] to provide services to the rest of the 8

[03:45] billion people that don't know how to

[03:47] use AI effectively yet. To these people,

[03:50] your AI services is going to seem like

[03:53] magic. And the reason that I'm dropping

[03:54] this full course right now is that the

[03:57] world is finally starting to understand

[03:59] AI. When I started providing AI services

[04:01] three years ago, the only real markets

[04:04] that we could go after was the US and

[04:07] Canada, the UK and Australia, every

[04:10] other market simply wasn't ready yet.

[04:12] But because of the mass adoption that's

[04:14] happening right now in AI, all of a

[04:16] sudden most other countries are starting

[04:18] to pick up as well. And this opens up

[04:20] markets like South America. Europe is

[04:22] one of the biggest upand cominging

[04:23] markets in AI right now. Countries in

[04:26] Asia like India is also starting to pick

[04:28] up on AI which means that companies are

[04:31] ready and they know that they need AI

[04:33] implemented into their business and it's

[04:35] not going to take long before the entire

[04:37] world is at the same wavelength. You

[04:39] might be thinking well AI companies

[04:41] already implementing AI I might be too

[04:43] late but I showed this to you to make

[04:44] you understand that it's not too late

[04:46] yet and by the end of this course when

[04:48] you stick around you will 100% be in the

[04:50] 0.04% 04%

[04:53] of AI users that will actually be able

[04:55] to capitalize on AI. But don't get too

[04:58] excited yet because if you've ever tried

[05:00] to start a business before, you might

[05:02] know or definitely have filled the

[05:05] excitement curve before. When you've

[05:06] just heard about a new business model or

[05:08] you want to start something or a new

[05:09] project or a new business, you're going

[05:11] to be extremely excited. Your excitement

[05:14] is going to be through the roof. You're

[05:16] going to be thinking about all of the

[05:17] good things. So, for example, with

[05:18] building an AI business, this would be,

[05:20] "Oh, I can automate my entire product

[05:22] delivery. Oh, I can get unlimited

[05:23] clients. Everyone's going to be

[05:24] interested in my services. I'm going to

[05:26] be rich." That's the thoughts that are

[05:28] going through your head when you're just

[05:29] starting out. And the reason that I told

[05:30] you right at the start that you

[05:32] shouldn't expect this to be a get-rich

[05:33] thing is because your excitement will

[05:36] then drop. This is 100%. All of a

[05:38] sudden, you realize all of the problems

[05:41] that are in the business, that it's

[05:43] actually not easy to get your first

[05:45] client, that you have to get really,

[05:46] really good at what you do. You need to

[05:48] watch videos like this. You need to

[05:49] learn before you can actually start

[05:51] selling these services, and that you

[05:52] probably won't land your first client

[05:54] the couple of weeks after starting. That

[05:56] is the reality of every single business

[05:58] model. There are always issues. There

[06:00] are always things. And things are never

[06:02] as easy as they seem. So, you reach a

[06:05] low of excitement. And this right here

[06:08] is where 95% of people give up. They

[06:13] don't want to go through the pain of

[06:15] figuring out, okay, how do I actually

[06:16] land clients? They don't want to spend

[06:17] all of the time learning the skills that

[06:19] it actually takes in order to provide

[06:21] good AI services. So, they do one of two

[06:23] things. Either they say, okay, this AI

[06:26] thing, scam, onto the next, they try

[06:28] dropshipping or NFTts or crypto or some

[06:31] other thing. or they simply just stop

[06:33] business altogether and they go back and

[06:34] just focus on the regular thing that

[06:36] they're doing the 9 to5 or whatever but

[06:38] a small percent of people 5% or so are

[06:43] going to work through the excitement

[06:46] drop and it's not a steep curve it takes

[06:49] time and all of a sudden they fix a

[06:50] couple of issues and then they land

[06:52] their first client so it might go up

[06:53] like this but then they realize okay

[06:55] it's actually not as easy as I thought

[06:56] providing the service so it stalls again

[06:59] and then they maybe figure out okay how

[07:00] do I actually provide at the service.

[07:01] So, it goes up a bit again and then

[07:03] their first client drops and all of a

[07:05] sudden they're back with no clients. But

[07:07] as they keep working, this curve starts

[07:09] going up and all of a sudden they crack

[07:11] how it works. And after long enough

[07:13] time, they're going to be very excited.

[07:15] They're going to know, okay, this is how

[07:16] I actually scale this business. This is

[07:18] how I do it. But getting to here takes

[07:20] an immense amount of time and work. I'm

[07:22] setting these expectations right now

[07:25] because if you're not ready to do this,

[07:26] then it doesn't make sense for you to

[07:28] spend hours watching this. you can just

[07:29] click off the video. But if you're

[07:31] actually committed to do this, if you

[07:33] are the top 5% that does to give up,

[07:36] then you are the person that I made this

[07:38] video for. The worst thing you can do is

[07:40] to get stuck in this loop of trying a

[07:42] new business model, getting very

[07:44] excited, getting the drop in excitement

[07:46] when you actually try it and say, "Okay,

[07:48] this doesn't work." And go on to the

[07:49] next one. because all of a sudden you

[07:51] get stuck in a loop of just trying stuff

[07:52] but never actually committing and going

[07:54] all in and actually staying with it,

[07:56] which is what it takes to get success in

[07:58] anything in life. I hope you're ready to

[08:00] put in the work. And if you are, then

[08:02] good. Listen up. These are the things

[08:04] that we're going to go through in this

[08:06] full course. Don't worry if you are a

[08:08] beginner in all of this claw stuff.

[08:10] We're going to start all the way from

[08:11] scratch designing and building our AI

[08:13] operating system from nothing. So, if

[08:16] you're just starting out, this is the

[08:17] perfect video for you. Most other guides

[08:19] on YouTube show you how to build the

[08:21] tech, but they actually never show you

[08:22] how to land your first client. So,

[08:24] that's actually the first thing that

[08:25] we're going to do in this course. We're

[08:27] going to focus on getting clients,

[08:28] getting clients, getting clients,

[08:30] because that is really what matters and

[08:31] really what moves the needle when you're

[08:33] just starting your AI business. The best

[08:35] way to land clients when you're just a

[08:36] beginner is to do something where you

[08:38] can reach out to a lot of people, but do

[08:41] it very cheaply. So, I'm going to show

[08:43] you how you can set up cold email

[08:44] campaigns, and I'm going to show you how

[08:46] you sign up to Upwork to get your first

[08:48] couple of freelance clients. Then, I'm

[08:49] going to show you how you take those

[08:51] interested leads that we got from cold

[08:52] email and Upwork, and how you book them

[08:55] in on meetings and how you close them.

[08:57] And I'm even going to show you a secret

[08:59] trick for how we can get Claw to listen

[09:01] in on our meetings so it gets the

[09:02] context for exactly what we need to

[09:04] build and can even start the building

[09:06] process before we even end the meeting.

[09:08] So, stay around for that cuz that's

[09:10] absolutely going to blow your mind. Then

[09:11] I'm going to show you how you actually

[09:13] do the service delivery. I'm going to

[09:15] show you how to build websites,

[09:16] automations, mini apps, and the process

[09:19] that I use for basically being able to

[09:21] build anything with no university degree

[09:24] where Claude Code does the heavy lifting

[09:26] for us. And then we're going to go over

[09:27] a very important step which a lot of AI

[09:29] business owners, they miss. We're going

[09:31] to take the cool things that we have

[09:32] built and we're going to show them to

[09:34] the world. We're going to post this on

[09:35] LinkedIn. We're going to post it to

[09:37] YouTube. And this will have a

[09:38] compounding effect where people all of a

[09:40] sudden see the proof of the stuff that

[09:42] you have built which is going to get a

[09:43] ball rolling to get you even more

[09:45] clients. And at the end I'm going to

[09:46] show you how not only to automate your

[09:48] clients businesses but also how you

[09:50] automate your own business so you truly

[09:52] achieve the oneperson AI business using

[09:54] clawed code. This is probably going to

[09:56] be the most valuable thing that I've

[09:58] ever released and I'm releasing it for

[09:59] free. So I'm very excited. Let's get

[10:01] into it. The beauty of Claude Code is

[10:03] that we can manage our entire business

[10:05] with it.

[10:08] So, let's set that up, shall we? If you

[10:10] have already set up plot code inside of

[10:12] Visual Studio Code, then you can just

[10:13] skip to the next chapter of the video.

[10:16] But let's set up Cloud Code the right

[10:17] way. The first thing you want to do is

[10:18] to open a browser. Then you want to

[10:21] search for Visual Studio Code, click the

[10:24] top link, and then download it for your

[10:27] computer, either Mac or Windows. Visual

[10:30] Studio Code is what's called a code

[10:32] editor. But don't worry, we're basically

[10:34] not going to create any code ourself. CL

[10:37] code is going to do absolutely

[10:38] everything for us. We then click on the

[10:40] installer. Then we drag it over to the

[10:42] applications. We let it install. And now

[10:45] we can open up Visual Studio Code.

[10:47] You're going to land on this page right

[10:50] here. Before we do anything inside of

[10:51] here, let's create a folder that we can

[10:54] work within. Damn.

[10:56] >> Excuse the mess on my desktop. I promise

[10:58] you we will get that cleaned up. I'm

[11:00] going to create a new folder and I'm

[11:01] going to call it shiny which is the name

[11:04] of our agency. This is going to be where

[11:06] our AI operating system lives inside of

[11:10] this folder. So call it something that

[11:12] you can remember. Then head back under

[11:14] Visual Studio Code and click this open

[11:16] button right here. Then go to wherever

[11:18] you saved the folder. I saved it on my

[11:20] desktop. So I'm going to open this shiny

[11:22] folder. Click open. And there we go.

[11:24] We'll then click yes. I trust the

[11:26] authors. and we can close down this

[11:29] welcome message. Now we are inside of

[11:31] our code editor and we are ready to

[11:32] start setting up clawed code. What you

[11:34] want to do is that you want to open the

[11:36] terminal. So I'm going to write click

[11:38] inside of this space and click new

[11:40] terminal. And this is going to open up

[11:42] this terminal view right here. Don't

[11:44] worry, this is going to look a bit

[11:46] technical, but you can just watch

[11:47] exactly what I do onetoone and it's

[11:49] really not going to be that hard. You

[11:50] don't need to be that technical for

[11:51] this. Then I'm going to open a new tab.

[11:54] I'm going to search for claude code

[11:56] install.

[11:58] Click the first one that's called quick

[12:00] start cla code docs. And then we have

[12:02] three commands right here for installing

[12:04] it on Mac or on Windows. So I'm going to

[12:06] choose the Mac install that fits with my

[12:08] computer. Then I'm going to paste in

[12:10] this command and that's going to install

[12:12] cloud code on our computer. And there we

[12:14] go. It now says setting up claude code.

[12:16] Installing cloud code native build

[12:18] latest. It then says claude code

[12:20] successfully installed. It shows the

[12:22] version. It shows where it was

[12:23] installed. So now we can write clear.

[12:26] That's going to clear our terminal. When

[12:27] we now write cla, you can see it's going

[12:30] to say welcome to clawed code. And we

[12:32] can choose what style we want. The way

[12:34] that we navigate is with the arrow keys

[12:36] up and down. And I'm definitely going to

[12:38] want dark mode. So I'm just going to hit

[12:40] enter to dark mode. And now it gives us

[12:42] three options. Either a claw account

[12:44] with a subscription or using an API key

[12:47] through the console. I'll definitely

[12:49] recommend you use claw with a

[12:51] subscription. either pro or max. I'm

[12:54] personally on the max plan right now

[12:56] because I need that bit more usage. But

[12:58] if you're just starting out, they don't

[12:59] recommend that you start on the pro

[13:00] plan. So just go to claude.ai, sign up

[13:03] to the pro plan, and you can see that

[13:05] gives you access to Claude code directly

[13:07] in your codebase. When you've signed up

[13:08] with an account and you have the pro

[13:10] plan, you now go to Claude account with

[13:13] subscription and you hit enter. And this

[13:14] is going to ask you to authorize with

[13:16] your Claude account. You can see it says

[13:18] Claude code would like to connect to

[13:19] your Claude chat account. So, we just

[13:21] click authorize. Now, it says build

[13:23] something. Great. How fitting. And you

[13:25] can see it says login successful. Press

[13:27] enter to continue. So, that's what we're

[13:29] going to do. We're going to click enter.

[13:31] And then we're going to hit enter again.

[13:32] And then claude is asking us if we want

[13:34] to use recommended settings. I'm going

[13:36] to hit enter. And there we go. Now,

[13:37] we're inside of cloud code. And then we

[13:39] can write just claude like this in our

[13:42] terminal after we've installed it.

[13:43] That's going to open up claude code on

[13:45] our computer. It says quick safety

[13:47] check. Is this a project you created or

[13:49] one you trust? So, I'm going to just hit

[13:51] enter. The way that you navigate inside

[13:52] of these options is that you can use

[13:54] your arrow keys up and down to go up and

[13:56] down. I'm going to click yes, I trust

[13:58] the author, and then hit enter. And

[14:00] there we go. Now, we are inside our claw

[14:02] code view. And we have successfully

[14:03] installed clawed code. The next thing

[14:05] we're going to need is some clawed

[14:08] skills inside of our completely free

[14:10] group. You can grab my clawed skills.

[14:12] I'll leave a link right below this

[14:13] video. And then you can find those under

[14:15] classroom and then go into AI learning

[14:18] hub. Scroll down until you find claude

[14:21] skills right here. And this gives us

[14:22] this drive folder with all of our claw

[14:25] skills. So I'm going to click claude

[14:27] skills right here. I'm going to click

[14:28] download. And now you can see it's

[14:30] downloading all of these skills in a

[14:32] folder. Claude skills are basically

[14:35] portable skills that you can give to

[14:36] claude code. So inside of these skills,

[14:39] we have some documents that explains how

[14:41] to use something. It's basically just a

[14:42] list of instructions. But by having the

[14:45] right skills and having the right

[14:46] instructions, you can make claude code

[14:48] even smarter than it already is. So if

[14:50] we go back into claude code and we then

[14:52] doubleclick on the zip file that we just

[14:54] created, this gives us our claw skills.

[14:56] Now going to write exactly this inside

[14:59] of my folder claude skills 2. This was

[15:04] the folder name, right? The folder name

[15:06] is claude skills 2 in my downloads

[15:09] because it's inside of our downloads

[15:11] folder on our computer. I want you to

[15:13] install these skills in this project

[15:20] folder. And then I'm going to call the

[15:22] folder shiny where we are right now.

[15:26] Then we're going to write exactly this.

[15:27] I want you to research the documentation

[15:31] for

[15:33] installing claude skills in a project.

[15:38] Please research that and install these

[15:41] skills in this current folder. Hit

[15:45] enter. And now what we're basically

[15:46] forcing Claw to do is that we're forcing

[15:48] it to go out and research first before

[15:51] we have it do anything. And that is also

[15:52] how you usually work with Claude.

[15:55] Sometimes it doesn't have all of the

[15:57] context that it needs. So by telling it

[15:59] to go out and find that context, we're

[16:01] going to get much better results. That

[16:03] took about well exactly 43 seconds. And

[16:07] if we go inside of thisclaw folder that

[16:09] was now created on the left hand side

[16:12] and inside of the skills folder, we can

[16:14] now see all of the skills that we have

[16:16] inside of this folder. What you want to

[16:19] do now is that you want to write clear

[16:21] inside of cloud code. Then we can hit

[16:24] Ctrl C which is basically just going to

[16:26] stop cloud code and then we're going to

[16:29] force it to restart. After installing

[16:31] skills, you need to restart the session

[16:33] for the skills to actually be enabled.

[16:36] If I now write slash and then front end

[16:40] design, you can now see we have a

[16:42] front-end design skill right here that

[16:44] creates distinctive production grade

[16:46] front-end interfaces with high design

[16:48] quality. Use this skill when the user

[16:50] were asked to build web components,

[16:52] pages or applications blah blah blah. So

[16:54] now we have a front-end design skill and

[16:56] we have all of these other skills as

[16:57] well like composio or customer support

[17:00] or research or scalability or security

[17:02] etc. All of these skills are now inside

[17:04] of this. claude folder inside of our

[17:07] workspace. The next thing you want to do

[17:08] is that you want to be able to save this

[17:11] to the cloud. Right now, all of this

[17:13] information, all of these skills are

[17:15] living on our computer, but we want to

[17:17] have a copy of this in the cloud as

[17:19] well. That way, if I smashed my computer

[17:21] and for some reason it was destroyed, we

[17:23] would still be able to recover our

[17:25] workspace, which if you have been

[17:27] working in a workspace for, let's say, a

[17:29] couple of weeks, you definitely don't

[17:31] want to lose that workspace and all of

[17:33] those skills. Let me show you how to do

[17:34] that. Go back into your browser. Then

[17:37] search for github.com and create a user

[17:40] if you don't have one. After you've

[17:42] created a user, you're going to land on

[17:43] this page right here. Then you want to

[17:45] click new. Now I'm just going to call

[17:47] this shiny workspace. I'm going to

[17:51] choose myself as the owner. And then I'm

[17:54] going to make it not public because we

[17:56] don't want everyone to be able to see

[17:57] our workspace. That is a high value

[17:59] asset. So I'm going to choose private

[18:02] and then click create repository. This

[18:04] will basically create a folder inside of

[18:07] GitHub. And with GitHub, we can

[18:09] basically publish our code that allows

[18:11] us to save a version of it in the cloud.

[18:13] The next thing you want to do is that

[18:14] you want to copy this URL. Paste this in

[18:17] and say, please push this current folder

[18:21] we are in to this repo. And now what

[18:24] claude is going to do is that it's going

[18:25] to take these claw skills, which is

[18:27] everything we have inside of the repo

[18:28] yet, and it's going to take all of that

[18:30] and push it into the cloud. So again, if

[18:32] something happened to our computer, if

[18:34] we lost access to it, we don't lose our

[18:37] workspace. Very important. My GitHub

[18:39] account is already connected to my

[18:41] computer. When you're doing this for the

[18:43] first time, it's going to pop up and ask

[18:45] you to authorize with your GitHub

[18:47] account. So just make sure to do that

[18:49] when you're doing this for yourself. You

[18:50] can now see it says the push was blocked

[18:52] by an auto mode classifier. That's

[18:54] because we have clawed in auto mode

[18:55] right now. What I'll do is that I'm

[18:57] going to set it to just like default

[18:58] mode. Write try again. And then we just

[19:02] have to accept that it actually takes

[19:03] this workspace and pushes it into the

[19:05] cloud. Do we want to proceed? I'm going

[19:07] to go down and say yes and don't ask

[19:08] again. Hit enter. And this will now be

[19:10] pushed into the cloud. Push successfully

[19:12] to Albert shiny shiny workspace main

[19:14] branch 52 files. If we go back into

[19:17] GitHub, you can see this is what a blank

[19:19] workspace looks like. If you now hit a

[19:20] refresh, we'll be able to see our claude

[19:22] skills right here. This is the exact

[19:24] same files that are inside of this

[19:26] folder right here. Now, they're just

[19:28] also in the cloud. Great.

[19:30] Congratulations on having all of this

[19:31] set up. Now comes probably the most

[19:34] valuable part of this entire course, and

[19:36] it's really what no one else is showing

[19:38] you on YouTube. Everyone is talking

[19:41] about the tech. But the hard part for

[19:43] almost everyone is, okay, now I know how

[19:46] to build stuff. How do I actually sell

[19:48] it? How do I go out and actually land

[19:50] clients? A wise man once said that you

[19:53] have to get good before you can get

[19:56] rich. What do I mean with that? What

[19:58] you've probably seen on Instagram and

[20:00] all the other socials as well is people

[20:02] saying, "Go out, learn how to build

[20:04] websites, and sell your first one for

[20:06] $10,000." And while it is true that you

[20:09] can sell websites for this much, is it

[20:12] realistic that you're going to go out

[20:13] and close a 10K deal as your first one?

[20:17] Well, probably not. And that is also

[20:18] what I mentioned earlier when we talked

[20:20] about expectations. It's probably not

[20:22] going to go that way. Why is that? Well,

[20:24] there's a couple of reasons. The first

[20:26] one is that you have absolutely no

[20:28] trust. And how do you build trust? Well,

[20:31] you need testimonials and you need

[20:32] previous experience. So, when you hop on

[20:35] a call with someone that you actually

[20:37] know what you're talking about, that

[20:38] you're actually certain in yourself

[20:41] that, okay, you can help this person get

[20:43] better results. In this course, I'll

[20:44] show you how to get both of those

[20:46] things. And what I'll also show you is

[20:48] the next part that you're lacking, which

[20:50] is skills. For some reason, these days,

[20:53] people think that you can just go out

[20:54] and learn something in 5 minutes and all

[20:56] of a sudden charge thousands of dollars

[20:58] for it. I'm sorry, but that is not

[20:59] reality. You've got to understand that

[21:01] what people are paying you for is the

[21:04] time and effort you've put in to gather

[21:06] skills. Skills that this person that are

[21:09] paying you doesn't have themselves. If

[21:11] it was easy, the thing that you doing,

[21:13] the thing that you had learned, then no

[21:15] one would pay you for it. So, you have

[21:16] to do exactly what it is that you're

[21:18] doing right now when watching this

[21:19] video, you have to go out and you have

[21:20] to learn. You have to build these skills

[21:23] in order to get paid. And that's exactly

[21:25] what I mean when I say you have to get

[21:26] good before you get rich. And of course,

[21:28] that's also the entire point of this

[21:29] video is to get you to build those

[21:31] skills. When I started my AI business,

[21:34] it didn't take a week. It didn't take a

[21:36] month. It didn't take two months. It

[21:38] took four months before I closed my

[21:42] first paying client for $400. And guess

[21:45] what I did before that? I did not charge

[21:47] $10,000 for a website. I actually did

[21:51] free work in order to actually build

[21:53] trust, to gain experience, and to gain

[21:55] testimonials. And by doing the free

[21:57] work, I also gained the skills that I

[21:59] could later monetize and actually charge

[22:01] for. So, if you're thinking right now,

[22:02] okay, I'm going to go out, I'm going to

[22:03] learn this, and then the next week I'm

[22:05] going to sell a 10K website, leave this

[22:06] video. This video is not for you.

[22:08] Understand that you have to build the

[22:10] skills before you can charge anything.

[22:11] And understand that this takes months.

[22:14] The way that I think about it is that

[22:16] there's kind of like three different

[22:17] levels to what you can be building when

[22:18] you're just starting out. The first

[22:20] level that I would recommend that you

[22:21] start with is AI generated websites.

[22:24] Specifically with Claude Code, it's

[22:26] extremely good at building websites. And

[22:27] let me show you something very, very

[22:29] interesting. If I head over to Claude

[22:30] and I ask how many small businesses

[22:34] still don't have a website, while I then

[22:37] drink my coffee, Claude is now going to

[22:39] research the web and find out the

[22:41] number. And now we can see that roughly

[22:44] 27 to 30% of businesses in 2026 have no

[22:49] website. So roughly one in three small

[22:51] businesses still don't have a website.

[22:53] So if you learn that skill of building

[22:55] websites, all of a sudden you can

[22:56] provide services to that 30%. and that

[22:58] 30% does not have a website. You can

[23:01] also build better websites and sell it

[23:03] to the rest of the 70% if you provide

[23:05] something that's better than what they

[23:06] currently have. So I can write okay so

[23:09] if it's 30% how many businesses in that

[23:11] in let's say the US just to put a number

[23:14] for how many potential businesses that

[23:15] you could actually be working with. I'll

[23:16] take another sip of the coffee while

[23:18] Claude is researching. So in the US

[23:21] there are 36 million small businesses

[23:24] and 30% of that is then 10 million

[23:27] businesses without a website and that is

[23:30] in my opinion the best place to start.

[23:32] Now you have millions of businesses that

[23:34] you could service with your skills. They

[23:36] don't know how to build websites with

[23:38] something like cloud code. They are not

[23:40] going to learn it. They are focused on

[23:42] their craft. So when you in this course

[23:44] is going to learn how to build websites,

[23:46] all of a sudden you can service all of

[23:47] those businesses and provide a valuable

[23:49] service. That is level one out of free.

[23:52] The next thing we're going to take a

[23:53] look at is the step up. Now you're not

[23:55] only building websites, you're also

[23:56] automating their business with

[23:58] automations and agents. This requires

[24:01] some more knowledge into integrations

[24:03] and APIs, but don't worry, you're going

[24:04] to be learning how to do that as well.

[24:06] And I would say the final level is

[24:07] building out full AI systems for

[24:10] businesses that not only integrate into

[24:12] what they already have, but also where

[24:14] you are collecting all of the

[24:15] automations that you have built from

[24:17] level two and then maybe creating a

[24:19] dashboard or somehow building it into a

[24:21] full AI system. But of course, we don't

[24:23] start at level three. We start at level

[24:25] one, websites. And stay around because

[24:27] later in the video, I'm going to show

[24:28] you exactly how you build those out in

[24:30] just a couple of minutes. I wanted to go

[24:32] over these three levels so you know what

[24:33] to actually sell because in the next

[24:35] part of the course, we're going to go

[24:36] into, okay, how do I actually land

[24:37] clients? One of the biggest mistakes I

[24:39] see beginners make is that they focus

[24:40] all of their time on building and not

[24:42] enough time actually outreaching and

[24:44] talking to clients. We're going to flip

[24:46] that on its head. So, before you even

[24:47] know how to build anything, we're going

[24:49] to start with the outreach because that

[24:50] means we can have the outreach running

[24:52] while we are then building stuff. So,

[24:54] let's get into probably the most

[24:56] valuable part of the course, how to

[24:57] actually get clients. Let's get into it.

[25:01] to do our outreach and actually build up

[25:03] the marketing system. We will of course

[25:05] also be using Claude Code. We're

[25:07] basically going to be using Claude Code

[25:08] throughout the entire video. And Claude

[25:11] Code will probably do 90 to 95% of the

[25:14] work in this business model that we are

[25:16] building up right now. I'll be showing a

[25:18] bunch of cool stuff that you can do

[25:20] inside of Claude Code. If you're looking

[25:22] for a full guide to Claude Code, then

[25:23] I'll definitely recommend that you check

[25:25] out the Claude Code masterass inside of

[25:27] our close community, the 1% in AI. This

[25:30] is also where you can get help directly

[25:32] from me. You have a full master class in

[25:34] absolutely everything inside of Claude

[25:36] Code that takes you from a complete

[25:37] beginner to a master in AI coding

[25:40] models. This community is also where you

[25:42] can get help from me. I answer every

[25:44] single post inside of this community. If

[25:46] I click on these icons right here and

[25:47] then go to new, then you can see that I

[25:49] answer absolutely every single post in

[25:51] here. This right here is my face and you

[25:53] can see I have commented and answered

[25:55] every single post inside of this

[25:57] community. I can actually see a post was

[25:59] just made 1 minute ago and just now. So,

[26:01] let me answer those right quick. That

[26:03] was one. Let me answer this as well.

[26:04] There we go. And the best thing about

[26:06] this community is that right now we are

[26:08] running the 30-day AI challenge. What

[26:10] this is is basically 30 days of videos.

[26:12] And when you finish those videos and

[26:14] stay consistent for 30 days, then you

[26:16] actually get your first month completely

[26:18] refunded. So, if you're consistent, you

[26:19] can start the 30-day AI challenge that

[26:21] takes you from a complete beginner to

[26:23] setting up an agency in 30 days. And

[26:25] then you can get to try all of this

[26:26] completely for free because when you

[26:28] finish, then you get all of your money

[26:29] back. If that sounds interesting, then

[26:31] watch this video right here. This

[26:33] explains everything. I'll leave a link

[26:34] right below this video. All right, let's

[26:36] continue with the course. Let me show

[26:37] you how you can actually get clients.

[26:38] All right, let's get into it. This is

[26:40] how I'd recommend that you get your

[26:42] first couple of clients and how you

[26:44] build trust and how you build

[26:45] testimonials. The first ones are always

[26:48] the hardest. It's like rolling a big

[26:50] stone. It's going to require a lot of

[26:52] force when you're just starting out. But

[26:53] as soon as you get the ball rolling,

[26:55] then you can use the momentum and it

[26:56] will start rolling faster. It's the

[26:58] exact same when it comes to landing

[26:59] clients and building the business. The

[27:01] first couple of months, the first couple

[27:02] of clients are an absolute pain. But let

[27:05] me show you how I recommend that you get

[27:07] your first couple of clients. To do

[27:08] this, we are going to use Upwork. If you

[27:11] don't know what Upwork is, Upwork is

[27:13] basically a freelancing platform that

[27:15] allows you to post yourself as a

[27:17] freelancer to say, "Okay, I can help

[27:18] with these things." And then businesses

[27:20] will also go to the platform and say I

[27:22] need help with these things. So this

[27:24] platform basically connects freelancers

[27:26] with actual businesses. The good thing

[27:28] about Upwork is that you can find leads

[27:29] directly that need your exact services.

[27:32] So if you're building websites then you

[27:34] can find people that need something

[27:35] improved on their website. If you're

[27:36] doing air automation system or go high

[27:39] level or inn you can find people that

[27:41] are specifically looking for that

[27:43] service. This doesn't mean that Upwork

[27:45] is just like free clients and that

[27:46] you're going to get a client as soon as

[27:47] you sign up to Upwork. There are really

[27:49] levels to the game when it comes to

[27:50] Upwork. How you build your profile, how

[27:52] you send your applications to post, but

[27:54] I'll cover all of that in this module.

[27:56] So stay around because there's a high

[27:57] chance that this I'm about to tell you

[27:59] right now will get you your first couple

[28:00] of clients. First thing you want to do,

[28:02] go to upwork.com, click the top right

[28:04] corner, and click sign up. And then sign

[28:06] up as a freelancer. I'm then going to

[28:08] use my Google account to sign up. There

[28:10] we go. I write in my first name. I write

[28:11] in my last name. And then I choose the

[28:13] country, which is going to be Denmark.

[28:15] Then I'm going to check this box right

[28:16] here and click create my account. And

[28:18] there we go. Congratulations. You have

[28:20] been signed up. Now we need to create

[28:22] our profile. So I'm going to click get

[28:24] started. I recommend that you write I

[28:25] have some experience right here. Click

[28:27] next. What's your biggest goal for

[28:29] freelancing to earn my main income?

[28:31] Click next. Then check both these on and

[28:34] check this right here and click next.

[28:36] Create a profile. Here you can do a

[28:38] couple of different things. If you have

[28:39] a LinkedIn that already now shows your

[28:41] experience, then you can connect your

[28:42] LinkedIn already. So that's what I'm

[28:44] going to do. I'm going to click import

[28:45] from LinkedIn. Then I will go to

[28:47] LinkedIn. Go to my profile. Click these

[28:49] three dots and click save to PDF. That

[28:51] basically creates a PDF of our LinkedIn

[28:53] profile. Then I can click upload to save

[28:55] LinkedIn PDF. Choose this. And this will

[28:58] now upload our PDF. So Upwork gets all

[29:00] of our information from our LinkedIn.

[29:02] Don't worry if you are not set up on

[29:03] LinkedIn yet. If not, just click fill

[29:05] out manually right here and go through

[29:07] the process. But for now, I'm going to

[29:08] click continue editing your profile. Now

[29:10] you're choosing what kind of work we're

[29:11] in. I'm going to go inside of IT and

[29:13] network and then choose database

[29:15] management and administration and CRM

[29:17] software right here. This is what best

[29:19] fits like a classic automation agency.

[29:22] Now we click next. Add your skills. Now

[29:24] we need to add a couple of skills. We

[29:26] could add CRM software already here. We

[29:28] could add our platform that we're

[29:30] working like in for example. We could

[29:32] add cloud. We could also add things like

[29:35] marketing. If you're doing Facebook ads

[29:37] for example, you could choose Facebook

[29:38] advertising. You could choose things

[29:40] like Google ads. You could choose things

[29:42] like building websites. You can always

[29:44] change this later, but for now it's good

[29:46] to show, okay, these are the things that

[29:48] you are usually working with. You could

[29:49] also choose AI agent development. And

[29:51] sometimes it's also good to choose some

[29:53] coding languages like TypeScript,

[29:56] Python, etc. That can sometimes help you

[29:58] get some jobs. You could also choose

[30:00] things like Sabia for example. This can

[30:03] help us as well. So basically like the

[30:04] automation platforms. Also choose

[30:06] make.com. And there we go. Now we have

[30:08] added a bunch of skills that shows what

[30:10] we can offer. We can then click next and

[30:13] add a profile title in our professional

[30:15] role. We can write something like this.

[30:16] AI agents and savior go high level make

[30:19] it in AI consulting and implementation.

[30:21] Then click next. Add your experience.

[30:23] You can see it has found that I'm

[30:25] co-founder of shiny AI solutions. You

[30:27] can click edit right here and then we

[30:28] can add more information like the

[30:30] location for example is in Denmark.

[30:34] Click I'm currently working in this

[30:35] role. Click save right here. Now we can

[30:38] add our education. Then you can add it

[30:40] right here. For now, we're not going to

[30:41] do that. Now we're going to click skip

[30:42] for now and then add languages. I'll

[30:44] definitely recommend that for languages

[30:45] that you choose native or bilingual

[30:48] right here. You then choose all of the

[30:50] languages that you know. I'm load Danish

[30:52] for example. I know a bit of German and

[30:54] I know a bit of Swedish as well. I can

[30:56] then write next. Write an overview.

[30:58] We're going to come back to this later.

[31:00] I actually already like the this AI

[31:02] version right here. As a co-founder with

[31:03] four years of experience in tech

[31:05] landscape. We can change this in AI

[31:08] automation AI agents. I specialize in

[31:10] developing solutions that drive

[31:12] efficiency and productivity and solve

[31:13] real world problems. And then we end it

[31:16] off with if you need someone who is an

[31:17] expert in automation, I am your guy.

[31:19] This one I probably want to leave. We

[31:21] don't want too big of a description. I

[31:22] like this one right here. It's also a

[31:23] good idea to write in the tools that you

[31:25] use. So I specialize in developing. I'll

[31:28] change that to building. I specialize in

[31:30] building solutions. specialized in

[31:32] building helpful solutions that also

[31:34] made it improve real world problems. I

[31:37] use tools such as go high level

[31:41] savior make inn like this if need

[31:46] someone who's an expert in operation I'm

[31:47] your guy we can even also write claw

[31:49] code I use tools such as go high level

[31:51] savior make inn and plot codec

[31:56] if you need someone who's an expert in

[31:58] automation I am your guy there we go now

[32:01] we go next set your rate I'd recommend

[32:03] not charging too much in the start what

[32:05] you'll see the expert on Upwork charge

[32:07] is something around like $50 an hour.

[32:09] For now, you just want to go out and

[32:10] land your first couple of jobs because

[32:12] those are going to show the success rate

[32:14] is going to show on your profile. So, I

[32:16] recommend that you set this to something

[32:17] like $15 an hour. The service fee, you

[32:20] cannot change that, which means you will

[32:21] get $135 more. Now, we click next. Add

[32:24] your photo and location. For the profile

[32:26] photo, you want to add a professional

[32:28] photo of yourself like this one right

[32:30] here. So, I'm going to click upload

[32:31] photo and then upload this photo of

[32:34] ourselves right here. Click attach

[32:35] photo. Then you want to write in your

[32:37] birthday, your country. Fill out all of

[32:39] this information and then review our

[32:41] profile. There we go. Agents and Savior

[32:44] go high level make and then consulting

[32:45] and implementation. Set our hourly rate,

[32:48] set our skills, set our work history,

[32:50] and then submit the profile. Now, Upwork

[32:52] is asking us to buy some connects. I'll

[32:54] definitely recommend that we do do that

[32:56] at one point. For now, I'm going to

[32:58] click browse without bidding. And we're

[33:00] going to go to this page right here. And

[33:01] you can see we are 60% complete. Now, we

[33:03] still want to add a bunch of stuff to

[33:05] our Upwork profile. It's important that

[33:07] you fill all of this information out as

[33:09] much as possible because this helps

[33:10] build trust with Upwork, which will help

[33:12] you land more jobs. So, inside of

[33:14] profile settings, I'm going to set

[33:16] myself as an expert in the space. We can

[33:19] link our accounts. Like, if you want to

[33:20] link our GitHub, we could do that. Going

[33:22] to click authorize. There we go. Then, I

[33:25] recommend that you complete this working

[33:26] style right here, which is basically

[33:27] just a small quiz. I click start

[33:29] assessment. Just fill out these

[33:30] questions with what you believe. stay

[33:32] organized. I like to communicate my

[33:34] schedule. When scheduling my week, I

[33:36] usually plan detailed steps, lock time.

[33:38] When I deliver work to a new client for

[33:40] the first time, I usually double check

[33:42] every part before sending it out. When

[33:44] timelines are tight, I prefer to focus

[33:47] on the key task that ensures quality.

[33:49] Before submitting work, I usually review

[33:51] the brief. I confirm it meets the

[33:52] client's intent. When collaborating with

[33:54] other freelancers, I usually cross check

[33:56] shared work for consistency. When

[33:57] project deadlines change suddenly, I

[33:59] talk with clients to confirm what's most

[34:01] important to finish first. I feel most

[34:03] productive. When I'm creating solutions,

[34:06] when I'm faced with something new, I

[34:07] usually start by brainstorming. When I

[34:10] need to make a decision, I typically

[34:11] make the call and take responsibility.

[34:13] Line the golden key so I can make like

[34:15] this. Before I start to work with a new

[34:17] client, I prefer to identify potential

[34:19] challenges early and plan the way to

[34:20] handle. There we go. Clear communicator

[34:22] is now visible on our profile. All of

[34:24] these things basically just helps you

[34:25] build trust with both Upwork and people

[34:28] looking at your profile and all of these

[34:29] are really important. The next thing you

[34:30] want to do is verify your identity. You

[34:33] do that by clicking this button right

[34:34] next to your name. Click on portfolio

[34:36] and then you can add a project title

[34:38] like a solar sales AI agent that I've

[34:41] done for example reactivation of old

[34:44] solar leads. This could literally be

[34:45] anything. Also, if you had a previous

[34:47] job, write in what you did and what you

[34:49] accomplished. Then I have this picture

[34:51] right here from our website. Let me take

[34:53] a screenshot that. Now we can add

[34:55] images. I also have a picture right here

[34:57] of a completely booked calendar that I

[34:58] worked on. I can write book Google

[35:00] calendar booked by an AI agent. I can

[35:02] add more images like this one. Free

[35:04] appointments booked in a month. Next

[35:06] preview. And there we have it. Solar

[35:07] sales agent developer. Reactivation of

[35:10] all solar leads. Next thumbnail. Let's

[35:12] choose this as the thumbnail because

[35:14] that's probably the most catching

[35:15] picture. And there we go. Our portfolio

[35:17] is published. Now you can see we are 75%

[35:21] done. We could also create a video

[35:22] introduction if you wanted to. We could

[35:24] add certifications. If you have any

[35:25] certifications that you've done that are

[35:27] slightly similar to what it is that

[35:28] you're offering now, make sure to show

[35:30] that on your Upwork profile as well,

[35:32] including an employment history is also

[35:34] smart. So, if you have a previous job,

[35:35] include that, too. Basically, fill out

[35:37] as much information on this Outboy

[35:39] profile that you can. One of the things

[35:41] that can make or break your Upboy

[35:42] profile as well is the location right

[35:45] here. So, you want to make sure to go

[35:47] inside of profile settings. Go inside of

[35:49] contact info and then change this to a

[35:51] location that people know. Right now,

[35:52] I've set it to a city called for exper

[35:54] Denmark. But I'm going to get better

[35:56] results if I set it to something that

[35:58] people know like Copenhagen. Then we can

[36:00] click update. And now you can see it

[36:02] says Copenhagen, Denmark. And people

[36:03] know what Copenhagen is, which one

[36:05] sparks new conversations, and two builds

[36:08] credibility, which is what all of this

[36:09] is about. But we can improve this upward

[36:11] profile a whole lot more. And of course,

[36:13] we're going to use Claude for that.

[36:14] Let's get into it.

[36:15] >> Before we continue, let's take a look at

[36:16] the people that are absolutely crushing

[36:18] it on Upwork. Like this guy, for

[36:20] example, Vipa D. Total earnings a

[36:24] million on Upwork, 2,000 hours worked.

[36:27] And if we take a look at his profile, we

[36:29] can see it says Istanbul, Turkey. You

[36:31] know, a capital big city that people

[36:33] know. 98% job success and top rated. And

[36:36] the great thing about these Upwork

[36:37] profiles is that they are 100% public.

[36:40] So, we can just go in and steal what we

[36:42] want. You can see he's offering a

[36:44] consultation first and then he's showing

[36:45] some work history with his reviews. You

[36:49] can see that he has a bunch of languages

[36:50] right here that he's showing. He's ID

[36:52] verified. I'll show you how to do that

[36:54] later as well. It's showing education

[36:56] and then showing a bunch of reviews.

[36:57] It's showing his portfolio of things

[36:59] that he's already worked on. It's

[37:00] showing certifications, employment

[37:02] history. You can see he has a very

[37:04] stacked Upwork profile. The most

[37:05] important thing for an Upwork profile is

[37:08] this job success right here. And also

[37:10] this text right here that says available

[37:12] now for the job success. We literally

[37:14] just need to get some jobs done. And

[37:16] that also means that we don't go after

[37:17] the $75 an hour jobs just yet. If we can

[37:20] get a job for $15 an hour, even if it

[37:23] doesn't make us much money, it's still a

[37:25] massive win because it's going to show

[37:26] on our Upwork profile. Let's move on to

[37:28] this guy. You can see also available now

[37:30] 100% job success. He's charging $62 an

[37:33] hour. He also has a booger consultation.

[37:35] So there's definitely a pattern here. He

[37:37] has worked 4,000 hours in total. He has

[37:40] perfect reviews, a portfolio, a bunch of

[37:42] skills, and a project catalog. Same with

[37:45] this guy. Also available now. 100% job

[37:47] success rate, 57.63

[37:50] an hour, 200k earned, also book

[37:53] consultation. So, we should probably add

[37:54] this. And then a bunch of work history,

[37:57] and a bunch of things that he has

[37:58] already worked on. It's always important

[38:00] when you're starting something new that

[38:01] you look at the people that are actually

[38:03] crushing it to steal all you can. What

[38:04] I've done is that I've actually taken a

[38:06] bunch of these high performing Upwork

[38:08] profiles. I've given them to Claude and

[38:10] created a Claude skill which means that

[38:11] we can literally just pass in our own

[38:13] Upwork profile and Claude can improve

[38:15] it. Let's do that. So now we can head

[38:17] back into Claude code. The Upwork skill

[38:19] should already be installed for you from

[38:21] all of my skills. Just write slash

[38:23] upwork and then we want to go to our own

[38:25] Upwork profile. Basically just Ctrl+ A

[38:27] to copy everything and say improve my

[38:31] Upwork profile.

[38:34] Paste this in. And because of the Upwork

[38:36] skill that is already trained in the

[38:38] best performing profiles in our space,

[38:40] Claude can now take that as a reference

[38:41] and then improve our Upwork profile. Of

[38:43] course, your Upwork profile should be

[38:45] with whatever you want to sell. I'm

[38:47] selling AI lead follow-up systems. So,

[38:49] I'm going to copy this one. Actually, I

[38:50] kind of like this one. AI automation

[38:52] engineer. Let's change this. Insert one

[38:54] of these as well. AI automation engineer

[38:56] lead follow-up systems and go high

[38:57] level. Instead of only lead follow-up

[38:59] systems, we can write AI agent systems

[39:01] and go high level and make save. Now you

[39:04] can see that claude is auditing our

[39:06] description as well. This is probably

[39:08] pretty good. I like this. So let's copy

[39:10] that and insert it.

[39:14] Click save. This is a pretty good call

[39:16] to action. Message me with whatever it's

[39:18] leaking. So let's include that at the

[39:20] end. I like this. And I'll tell you in

[39:21] 24 hours if I can fix it. That's like

[39:23] kind of a call to action on our Upwork

[39:25] profile. So we can write message me with

[39:27] your project

[39:30] and I'll tell you in 24 hours if I can

[39:32] help. Great. I think our Upwork profile

[39:35] is looking a lot better. Another tip

[39:37] that you want right now I just set the

[39:39] price to $15 an hour. You want to be

[39:41] more precise. Like let's set it to

[39:44] $16.73

[39:46] an hour for example and click save. The

[39:48] reason for that is that if you just set

[39:49] it to let's say $16 straight, it doesn't

[39:52] seem like a precise price. It's just

[39:53] like thrown out there. When you set it

[39:55] to something very precise, it makes the

[39:57] human brain think, "Okay, there's a

[39:58] reason that it's this precise, which can

[40:00] build credibility." Awesome. Our profile

[40:02] is starting to look a lot better with

[40:04] the help from Claude. Now, it's time to

[40:06] sign up to some jobs. Before we do that,

[40:08] you need to know about the two plans

[40:10] inside of Upwork. Right now, we are on

[40:12] the basic plan, right, which is the free

[40:13] plan. And here, we get 10 connects per

[40:16] month when we finish our Upwork profile,

[40:18] which if you are serious about building

[40:20] out your Upwork profile, probably won't

[40:22] be enough. They have an offer right here

[40:24] though. You can see the plus plan right

[40:25] now is $20 a month. But if you go to

[40:28] your profile, then you can actually get

[40:30] 50% off right here. Get freelancer plus

[40:33] for 50% off for one month. So that's $10

[40:36] for to try it for 1 month where you get

[40:38] a 100 connects instead, which is very

[40:40] very worth it. So I'm going to go and

[40:42] click upgrade now to the plus plan.

[40:44] Let's get started with that. There we

[40:46] go. You can now see we are on the

[40:47] freelancer plus plan. And the first

[40:49] thing you want to do is that you

[40:49] actually want to head back to your

[40:51] profile. And then you want to verify

[40:52] your identity. This was something that

[40:54] we didn't have before with the standard

[40:56] plan. And this basically just builds

[40:57] trust with both the Upwork and also with

[40:59] our clients. And this also cost 35

[41:02] connects right here. Right now we have

[41:04] 150. So that's definitely worth getting

[41:05] this done. So I'm going to click verify

[41:07] your identity. And now we scan it with

[41:09] our mobile device. And then we have to

[41:11] verify with some ID. I'll be back when

[41:13] I've done that. And there we go. That

[41:14] took about 5 minutes. So now you can see

[41:16] right next to my name, we now have a

[41:18] verified badge that says, "This

[41:20] freelancer's identity has been verified

[41:22] through a government ID check and a

[41:24] visual verification." Nice. On the left

[41:26] hand side of your profile right here,

[41:27] you also have the availability badge.

[41:29] That's the thing I showed you before

[41:31] where it says available now, which cost

[41:33] 14 connects per week. I would also say

[41:35] that that's very worth it. So I'm going

[41:37] to click turn on. You've added the

[41:38] badge. And now we can go to our public

[41:40] profile. We can see how it looks.

[41:43] You can see now it says this available

[41:45] now. A automation engineer $1673 an

[41:48] hour. There we go. The skills, the

[41:50] employment history, and the only thing

[41:51] that our profile is now missing is some

[41:54] jobs cuz that is really what matters.

[41:56] The thing that people really take a look

[41:57] at is this right here. Top rated plus

[41:59] and 98% job success rate. And then also

[42:02] total earnings, total jobs, total hours.

[42:04] And that's also why we set our price

[42:06] very low compared to these other people.

[42:08] You can see they're charging $62 an

[42:10] hour. We are only charging 1/4 of that.

[42:12] It's because we want to get jobs first

[42:14] so we can have 100% job success rate.

[42:16] Nice. It's time to go out and actually

[42:18] start bidding on jobs, which is what

[42:20] we're going to be using these connects

[42:22] for. So, let's get into that. All right,

[42:24] let's start actually applying to jobs.

[42:26] If you go to the find work tab right

[42:28] here and click find work, then it's

[42:30] going to show you the jobs that Upwork

[42:32] thinks that you will be interested in.

[42:34] We can see our connects right here. And

[42:36] the way that connects work is that we

[42:38] are basically paying a little bit every

[42:40] single time that we apply to a job. What

[42:42] does this mean? Well, it means that we

[42:43] need to be cautious with what we apply

[42:45] to. We cannot just apply to every single

[42:47] job. There's a couple of rules of funds.

[42:49] Like if a post, for example, has 50 plus

[42:52] proposals, it's probably not the best

[42:54] job for you to reach out to. 20 to 50 is

[42:57] in a much much better range. And it's

[42:59] really a speed to lead game. When

[43:02] someone makes a post, you want to be

[43:04] instant there. You can see this was

[43:05] posted yesterday. Posted yesterday. You

[43:07] want to be there instantly. This one,

[43:09] for example, right here was posted 3

[43:11] hours ago and the proposals are 5 to 10.

[43:14] This one would be a great one to reach

[43:16] out to. So, what you want to do is that

[43:17] you want to commandclick on this. So, it

[43:19] opens in a new tab. And you can see this

[43:22] opens up the job right here. We're

[43:23] looking for a skilled AI automation

[43:25] specialist who has hands-on experience

[43:27] with go high level. That's us, Cloud AI,

[43:29] and AI agent development to help

[43:30] streamline business operations, improve

[43:32] client communication workflows. There we

[43:34] go. This seems awesome. Remote,

[43:36] freelance, potential, long-term

[43:38] opportunity. There we go. There we go.

[43:40] This one seems like a really, really

[43:42] good job. And you see they are payment

[43:43] verified and phone number verified as

[43:45] well. And they have a 98% higher rate

[43:48] with 910 jobs posted. And again, what

[43:51] we're trying to do now is not to make

[43:53] the most amount of money. We need to

[43:54] build up our Upwork profile before we

[43:56] can do that. So, if we think we can do

[43:58] this, which we can, we can click apply.

[44:00] Now, you can see that this proposal

[44:01] requires 11 connects. And when we submit

[44:03] it, we have 103 left. How do we want to

[44:05] be paid? I'm going to choose by

[44:07] milestone. And then we can choose maybe

[44:09] something like 25 25.

[44:15] And we can do

[44:18] initial

[44:19] audit

[44:21] implementation.

[44:24] Verify that everything runs smoothly.

[44:28] And you can see when we do this, we are

[44:30] going to be making $67 like this. How

[44:33] long will the project take? Probably

[44:35] less than one month. And then right

[44:36] here, we can add our cover letter. And

[44:39] this is what is important to add. This

[44:41] is basically our application. What I've

[44:43] done is that I've taken some of the best

[44:44] proposals that I've seen. And I've

[44:46] created an Upwork proposal skill like

[44:48] this one. Write a high converting Upwork

[44:50] proposal for a job post the user

[44:52] provides. So let's write / proposal

[44:54] inside of our AI workspace. And again,

[44:56] you're also going to have access to this

[44:58] skill. It's inside of the old skills

[44:59] folder. We're then going to take this

[45:01] post right here that was submitted 3

[45:02] hours ago. I'm going to Ctrl A

[45:05] everything so we can see what it looks

[45:07] like. Going to paste this in and then

[45:08] say write a proposal, please. This job

[45:12] might not be the best one. You can see

[45:13] right here it says $10 as a fixed price,

[45:16] but I think what they're trying to do

[45:17] right now with this post is literally

[45:19] just to find someone that they could

[45:20] maybe hire. You can see they've spent in

[45:22] total 9.3K. This is also what our skill

[45:24] flags. You can see $10 fixed price.

[45:26] Client's average hourly paid is $7 and

[45:28] they have 67 open jobs. They're probably

[45:31] just using this for a hire. That's

[45:33] actually a good flag by claw. So, this

[45:34] is probably not the best job to spend

[45:36] your connects on. You can also see what

[45:38] country these people are making the

[45:39] postal from. Nigeria right here, which

[45:41] might not be the best type of client.

[45:42] You can see this one might be

[45:43] interesting. This is high level

[45:45] automation fulfillment partner. So, this

[45:47] is more of like a long-term job. You can

[45:48] see they've spent $50,000. It's inside

[45:51] of go high level as well. So, let's

[45:52] apply to this one instead. Maybe we can

[45:54] click apply now. Then we can copy this

[45:56] entire thing. I'll write clear here to

[45:58] clear it. And then choose the Upwork

[46:00] proposal generator again. Paste this in

[46:02] and say

[46:04] build a proposal for this. And now it'll

[46:07] ask for some information about ourself

[46:08] since it doesn't have any saved facts.

[46:11] Have you built high level voice agents

[46:12] before? Yes, high level. I know

[46:16] everything about go high level and I

[46:20] have built a lot of voice agents in 11

[46:24] labs not only

[46:28] in English

[46:30] but also

[46:34] in many other languages.

[46:40] Do you actually use high level in it

[46:41] today? Yes, all the time. Pricing

[46:44] approach. Do you want to quote each

[46:45] service line by line? Blah blah blah on

[46:48] hourly. One hourly rate is fine.

[46:54] Your unfair advantage angle for

[46:55] fulfillment partner. The US time zone

[46:57] overlap priulment experience in-house de

[47:00] custom calculators. Sign off name

[47:02] Albert. Yes. Your unfair advantage.

[47:05] Let's come up with an unfair advantage.

[47:08] prior agency fulfillment

[47:11] used to run

[47:13] my own agency. There we go. We give

[47:15] Claude some context and it's going to

[47:17] build out this entire proposal for us.

[47:19] Don't want to paste this in one to one.

[47:20] I'm going to copy this right here. You

[47:22] can see this would cost 27 connects.

[47:24] Schedule a rate increase. Let's do never

[47:26] for that. You can always tell that later

[47:28] if you want. And then a quick tip is

[47:30] that you can write this is not written

[47:32] by AI. Then I don't think this is good.

[47:36] We'll just write I used to run my own AI

[47:38] automation agency. So the usell I

[47:40] fulfill model the fulfill is not that

[47:42] good. So I'll just write so I so I know

[47:45] the usell I fulfill model and when it

[47:49] comes to these proposals you want to

[47:51] educate them. So I'll write it's

[47:52] important that we do good onboarding

[47:54] flow snapshot deployments of account

[47:56] hygiene flows and snapshot deployments

[47:58] and then add some information that they

[48:00] might not know already. We can actually

[48:02] use the go high level API

[48:05] through something like make it in to

[48:09] auto deploy sub accounts

[48:13] in both go high level and 11 labs.

[48:19] How start? Let's first order your

[48:22] current highle setup.

[48:24] Then we built the reusable template that

[48:27] also deploys through Nadm and go high

[48:32] level and 11labs

[48:36] through

[48:37] the API and then take over all. Let's do

[48:40] like this. This not like what I would

[48:42] write take all then I take over all

[48:46] clients world plus two should close

[48:47] deals rate $16 an hour is a quote fixed

[48:50] per client when I see

[48:54] the current onboarding process maybe

[48:58] just like current onboarding this is a

[49:00] pretty bad question there's also some

[49:01] weird spacing right here let's remove

[49:03] that too question is it custom build per

[49:05] project is custom or do you sell

[49:08] productized

[49:12] services will help me to plan it out.

[49:14] Elbert like this right here. Then we can

[49:16] add an attachment like we can for for

[49:18] example add that calendar right here

[49:19] that is completely full. And there we

[49:22] go. This is a really good upwork

[49:24] proposal. What really can make it better

[49:25] is that if we also at the end right here

[49:28] write made a loom as well that describes

[49:31] how we can set this up. And then in this

[49:33] loom you want to showcase that you

[49:35] actually know what it is that you're

[49:36] talking about. Let me let me show you an

[49:37] example of what Loom we can create for

[49:39] this. For example, I'm going to go to

[49:40] loom.com and then I'm also going to go

[49:42] to make.com, which is probably what I'll

[49:44] be using to set this up. And again, this

[49:46] just becomes very easy when you already

[49:48] know what it is that you are doing. And

[49:50] again, in these proposals, you kind of

[49:51] want to educate them on what is actually

[49:53] possible for them to do. So, what I'm

[49:55] going to do now is that I'm going to

[49:56] make a loom video right here. Very

[49:58] important that we show our face. So, we

[50:00] can click record a loom. We turn on our

[50:02] camera right here. There we go. Make us

[50:05] as big as possible. That's a good idea.

[50:07] So they see, okay, we are an actual

[50:08] person, not an AI. There we go. Make

[50:10] sure to make yourself big in the corner

[50:11] and then click start recording

[50:14] entire screen. Sup guys, just wanted to

[50:16] elaborate a bit about what I explained

[50:18] in the proposal. What you want to do is

[50:21] probably use either make.com or inn.

[50:24] It's possible in both, but many don't

[50:26] know that inside of go high level, and

[50:28] this is actually make specific. This

[50:29] doesn't exist in inn. you have this API

[50:33] call right here that says adds an

[50:35] account. And what this basically does is

[50:37] that this adds a sub account

[50:39] automatically. So you don't have to

[50:40] create one inside of um inside of go

[50:44] high level. And then what you can also

[50:46] do after you've used this is that you

[50:48] can use the go highle v2 API.

[50:52] This one right here. And you have a an

[50:56] account update right here. So you can

[50:57] basically do anything you can do

[50:59] manually inside of go high level. can do

[51:00] this through the highle API, not

[51:03] something that's inside of actual make.

[51:05] You would need to create an HTTPS

[51:07] request like this

[51:09] uh make a request, but then you can

[51:11] basically do all of these things. You

[51:12] can update every single thing inside of

[51:14] high level. You can connect with uh

[51:16] other services as well. We can basically

[51:18] automate the entire thing. So uh so

[51:20] yeah, just wanted to make sure that you

[51:22] guys know that this exists and uh yeah,

[51:24] let me know if you need help in setting

[51:26] it up. There we go. That was what a 1

[51:29] minute 15 second loom. Just copy this

[51:31] link right here and then paste it at the

[51:34] bottom.

[51:36] This is a perfect proposal. This is not

[51:39] written by AI. I used to run my own AI

[51:41] automation agency. So the usell

[51:44] model. So I actually really like

[51:46] the userfill model. It's important that

[51:49] we do good on flows in Snapchat

[51:51] deployments. We can actually use the go

[51:52] high level API through something like

[51:54] M.N and then in to auto deploy sub

[51:55] accounts in both go high level and a

[51:57] level apps how hard start let's first

[51:59] order your current high level setup then

[52:00] we built a reusable template it's

[52:02] sometimes good as well to write that we

[52:03] have a loom down here at the bottom this

[52:05] not written by AI made a loom below that

[52:08] explains but I used to run like this

[52:11] right here you want to make it seem like

[52:12] it's not written by AI and all of a

[52:14] sudden we have changed a lot about it

[52:15] ourselves but this is a perfect proposal

[52:18] we have added an image of a screenshot

[52:21] of a book calendar we don't want a bit

[52:24] contacts and then we click send for 27

[52:26] connects. There we go. Submit. We can

[52:28] even see if it has been opened yet. So

[52:30] far there's 19 proposals, six open, 13

[52:33] unopened, and all proposals are at $25

[52:36] an hour. The top rated one is also at

[52:39] $25 an hour. So, we actually have a real

[52:41] shot of landing this job. That's just

[52:43] what you want to do. You want to apply

[52:44] to a bunch of jobs and you want to stack

[52:46] up on connects. And you might think,

[52:48] well, I'm going to have to keep buying

[52:49] connects all the time. And that is

[52:50] honestly true. But there's no other

[52:52] platform really like Upwork where you

[52:54] can find so many interested hot leads.

[52:56] And honestly, it's very, very cheap. If

[52:58] you had to run Facebook ads, for

[53:00] example, for these type of leads, you'll

[53:02] probably paid $100 or $200 per lead.

[53:04] Right now, for 10 connects, we're paying

[53:06] $1.50.

[53:08] So, make this your routine. Go out, send

[53:10] a bunch of requests, actually make them

[53:12] good. Spend some time on them. Spend at

[53:13] least like 10 to 15 minutes on proposal.

[53:16] And of course, at least when you're

[53:17] starting out and then when you get

[53:19] better, you of course can be able to do

[53:20] it faster. Make a Loom video to every

[53:22] single one. This will help you land

[53:24] jobs. And you might think, Albert, well,

[53:25] already to this there was 20 applicants

[53:27] already. No one is going to choose me.

[53:30] Let me tell you why that's not the case.

[53:32] And the best way to show you that is

[53:34] actually not to hop on my freelancer

[53:36] profile right here, but to hop on my

[53:38] buyer profile, cuz I've also spent

[53:40] thousands of dollars on freelancers on

[53:42] Upwork, so I know what I like to look

[53:44] for when it comes to freelancers. Let me

[53:46] show you that perspective as well. So

[53:47] you actually know what your profile

[53:48] looks like and what buyers think about

[53:50] when they look for freelancers. This

[53:52] right here is now from my other profile.

[53:55] So from my company profile, not from the

[53:58] freelancer account, but from the company

[54:00] profile. And I made a post a couple of

[54:03] weeks ago about finding a video editor

[54:05] to basically cut up videos. And you can

[54:08] see I got 11 proposals on this post. And

[54:12] you can see who actually signed up. One

[54:14] from Pakistan, another one from

[54:16] Pakistan, a third one, fourth, fifth,

[54:18] sixth, seven, one from Nigeria, one from

[54:21] India, one from Pakistan, another one

[54:22] from Pakistan. And the reason that your

[54:25] competition is not as good as you think

[54:27] is that a lot of the freelancers already

[54:29] on Upwork are from third world

[54:31] countries, which means that if you are

[54:32] from Western countries, you definitely

[54:34] have an advantage. This doesn't mean

[54:36] that if you're from India that you

[54:37] cannot do it. You're just going to have

[54:39] harder competition because there's a lot

[54:40] more freelancers from those countries

[54:42] already on the platform. And let's take

[54:44] a look at the actual proposals. You saw

[54:46] how I reworked mine to not sound like

[54:48] AI. But what you'll find is that almost

[54:50] everyone right here just has basic

[54:52] proposals as a top rated plus video

[54:55] editor with over 6 years of retention

[54:57] retention focused end to end YouTube

[54:59] postp production like it just sounds AI

[55:02] made. Then he has inserted a bunch of

[55:04] videos that doesn't really concern us.

[55:05] In terms of AI, it's a core part of my

[55:07] workflows. I use tools like chat GBT.

[55:09] I'm working in EST. We can discuss the

[55:11] monthly rate. that said I love allow to

[55:13] connect blah blah blah blah you can see

[55:14] very AI generated not very personal this

[55:16] one as well portfolio at the end of the

[55:18] process like this is just like a

[55:20] template you can see that it's a

[55:21] template it's not personalized to the

[55:22] actual job that's just to show you from

[55:24] the buyer side as well how it actually

[55:26] looks so you know okay this is what I

[55:28] should write this is how it should catch

[55:29] their attention the most important part

[55:31] of your cover letter as well is this

[55:34] part right here that we wrote right this

[55:36] cover letter because this is the thing

[55:38] that they see and determine okay am I

[55:39] actually interested in this person and

[55:41] it also shows how important it is to

[55:43] actually just land your first couple of

[55:44] jobs on Upwork. And the thing is, when

[55:46] you're just starting on Upwork, you're

[55:47] going to have to do more effort, right?

[55:48] You're going to have to make a better

[55:49] cover letter. You're going to have to

[55:50] make it more personalized, send more

[55:52] looms, but what you can see right here

[55:54] is that this is definitely what moves

[55:55] the needle, like the job success rate,

[55:57] how many jobs completed, total hours,

[55:59] and also money earned. The bias on

[56:01] Upwork, they definitely look for that.

[56:02] And then to set some expectations,

[56:04] building an Upwork profile can take

[56:05] months, if not years. It's not a

[56:08] get-rich quick thing by no means. It

[56:09] also takes long work. But what you'll

[56:11] find out is that these clients that you

[56:13] work with on Upwork, all of a sudden

[56:14] they refer you to someone else and

[56:16] someone else and someone else. And

[56:17] Upwork, even though it is a freelancer

[56:19] platform, it's the perfect place to

[56:21] start. Awesome. I showed you this just

[56:23] to make you understand that the

[56:25] competition on Upwork isn't really that

[56:27] bad. And Upwork, even though that it's a

[56:29] freelancing platform, is in my opinion

[56:31] one of the best places to start. And

[56:33] here's why. You build experience. It's

[56:35] what I mentioned in the start, right?

[56:37] Get good before you get rich. And this

[56:39] by actually doing the work is the best

[56:41] way of getting good. And what also

[56:43] happens is that all of a sudden you have

[56:45] clients, you talk to people, and you'll

[56:47] be referred to other clients. It's a

[56:49] snowballing effect. And all of a sudden,

[56:51] when you have a bunch of clients, you

[56:52] can automate a lot of the work with

[56:54] Clawude Code. And boom, you're running a

[56:55] one-person business with Clawude Code.

[56:58] So, if you want to do any type of

[56:59] outreach, I recommend that you start

[57:00] with Upwork. And just to set some

[57:02] expectations, it might take a couple of

[57:04] weeks before you land your first

[57:06] freelancing client, but it's still going

[57:07] to be so worth it. However, if you want

[57:09] to speed that up and get even more

[57:11] volume, then I'm going to show you

[57:12] something insane that we can do with

[57:14] Cloud Code in the next part of the

[57:15] course. Let's get into it. Great. To be

[57:17] clear, now you just want to send as many

[57:19] connections as you want and land your

[57:22] first couple of jobs on Upwork. That's

[57:24] why you have the highest likelihood of

[57:26] getting your first client the fastest.

[57:27] If you want to push that further and

[57:30] send out a bunch of connections to every

[57:31] new job that is posted on Upwork that

[57:33] you're interested in and you want to do

[57:34] even more than that, then this part of

[57:36] the course is for you. If not, that's

[57:38] completely fine and you can skip to the

[57:39] next part of the course. Do you remember

[57:40] when we talked about this map right here

[57:43] where all of these countries marked with

[57:45] green, they are getting very very used

[57:47] to AI, especially like chat bots and

[57:49] voice agents. All the other countries

[57:52] however are a bit behind. So that is

[57:55] South America like for example South

[57:56] Africa is a very booming market. It's

[57:58] especially Europe and then also many

[58:00] parts of Asia. If you live in one of the

[58:03] blue countries then you should

[58:05] especially listen up in the next part of

[58:06] the course. This will also work in the

[58:09] green countries but it will have lower

[58:11] reply rates. What I've seen is that this

[58:13] next method is absolutely insane in the

[58:15] blue countries right here. And what

[58:17] we'll be doing is something called cold

[58:19] email. You have probably heard about it,

[58:20] which is where we can send a lot a lot

[58:22] of cold emails out to a lot of qualified

[58:24] prospects and then the people that are

[58:26] interested will reply back and we can

[58:29] automate this entire thing. My favorite

[58:31] code emailing software is instantly.ai.

[58:34] I'm not sponsored to say this. That is

[58:36] just my favorite platform. If you want

[58:37] to set this up and you want to support

[58:39] this channel, then make sure to check

[58:40] our affiliate link right below this

[58:42] video. Again, I'm not sponsored, but

[58:43] I've just created an affiliate link

[58:45] through their affiliate platform. But

[58:46] why is this software so good? Well, it

[58:48] allows us to get a bunch of email

[58:50] accounts very very easily and then send

[58:52] a bunch of bunch of emails to qualified

[58:55] leads and then they have an MCP that

[58:58] allows Claude to do 90% of the work. So,

[59:01] let's set that up first, shall we? The

[59:03] first thing I want you to do is actually

[59:04] search for the Claude desktop app if you

[59:07] don't have it already and then download

[59:08] it for whatever computer you have. So,

[59:10] I'm going to search Mac and then

[59:12] download Claude right here. Download for

[59:14] Mac OS. That's going to download the

[59:16] installer just like this. The next thing

[59:18] we want to do is actually not go inside

[59:19] of Claude code, but then go inside of

[59:21] the Claude desktop app that we have just

[59:23] installed. The reason we do this is

[59:25] because that's by far the easiest way to

[59:27] add MCPS. If you go in the bottom left

[59:29] corner right here and then click on

[59:31] settings and then go inside of

[59:33] connectors and then click this customize

[59:35] right here cuz they have now moved it,

[59:37] we can see all of the connectors that we

[59:39] have. So I can now click add connector

[59:41] right here. Add custom connector. call

[59:44] this instantly which is the platform and

[59:48] then what we need to do now is to go

[59:49] back to instantly. So now's the time if

[59:52] you haven't already then sign up for an

[59:53] instantly account again. If you want to

[59:55] use my affiliate link then that would be

[59:56] very appreciated. We can now click on

[59:58] the bottom left corner right here go to

[01:00:00] settings then click on integrations and

[01:00:02] then click on API keys right here. And

[01:00:05] now we want to create an API key. This

[01:00:07] is what allows claude to basically have

[01:00:09] access into instantly. So click create

[01:00:11] API key. I'm going to call this claude.

[01:00:14] I'm going to select all so I can do

[01:00:16] everything it needs through the API. And

[01:00:18] then we click create. And this is going

[01:00:20] to create this API key right here. So

[01:00:22] let me copy that one. And then I'm going

[01:00:24] to go to this docs right here. I'm going

[01:00:26] to leave this doc right below this

[01:00:28] video. But that's basically the MCP URL

[01:00:30] that we are using for instantly. So you

[01:00:32] can see it says your API key right here.

[01:00:34] You would copy this, change it with your

[01:00:36] own API key. And of course I'm going to

[01:00:38] rotate this API key after we're done. I

[01:00:40] can now copy this entire thing. Go back

[01:00:42] to CLA, add it right here, and then

[01:00:45] click add. And there we go. The

[01:00:46] instantly MCB has now been added. I can

[01:00:49] now just go back if I wanted to test it

[01:00:51] right quick. Go into cloud code, for

[01:00:52] example, and just ask it right quick. Do

[01:00:54] you have access to the instantly

[01:00:59] MCP? Now, this is just to verify that it

[01:01:01] actually has the access that it needs.

[01:01:03] You can see it says yes, the instant MCP

[01:01:05] is available. And I can see tools like

[01:01:07] blah blah blah blah. There we go. I'm

[01:01:09] also just going to ask it right now, can

[01:01:11] you pull campaigns just to check the

[01:01:14] connection works. This is to ensure that

[01:01:16] it both has the MCP. We just added that,

[01:01:19] but that it also can list all of the

[01:01:21] campaigns. You can see it's using this

[01:01:22] right here. List campaigns. Let's see if

[01:01:24] it returns. Connection works. One

[01:01:26] campaign returns blah blah blah like

[01:01:28] this. Great. The reason we installed the

[01:01:30] MCP inside of the claw app is because

[01:01:32] that makes it much much easier when we

[01:01:33] go to claw code. I'm going to click Ctrl

[01:01:36] C right here a couple of times. That's

[01:01:39] going to close. I'm going to write clear

[01:01:42] again and write cla again. And that's

[01:01:43] basically how we restart cloud. Now I

[01:01:46] can write slashmcp.

[01:01:48] Hit enter. And under cloud AI, we can

[01:01:51] now see the MCPS that we have added. And

[01:01:53] you can see right here claude.AI

[01:01:55] instantly is connected with 38 tools. I

[01:01:58] can hit enter on that as well. I can hit

[01:02:00] enter to view tools. And then we can see

[01:02:03] all of the tools that it has access to

[01:02:04] which was what 38 tools, right? They can

[01:02:07] basically reply to emails, list emails,

[01:02:09] create lead list. It can do it all. And

[01:02:11] you can see the status is already

[01:02:12] connected because we added the API key

[01:02:14] right here. Cool. We can write clear.

[01:02:17] That clears the claw code session. And

[01:02:19] then we can basically get claw to create

[01:02:20] these cold email campaigns for us.

[01:02:22] Before we do that, we need to do a very

[01:02:24] important step, which is we need to go

[01:02:25] into instantly and get some email

[01:02:27] accounts. Email accounts are basically

[01:02:29] the emails that we are sending from. And

[01:02:31] there's two different ways that you can

[01:02:32] do this. I can click add new right here.

[01:02:34] You can either choose pre-warmed

[01:02:36] accounts, which is basically accounts

[01:02:37] that instantly has warmed up for us.

[01:02:40] These accounts don't have our own name.

[01:02:42] You can see it's Arthur, it's Charlotte,

[01:02:44] and it's using these interesting domains

[01:02:46] right here. But they are pre-warmed,

[01:02:47] which means that when we send emails, it

[01:02:49] will have high deliverability. And those

[01:02:52] are the ones that I recommend that you

[01:02:53] start with because usually you would

[01:02:54] have to warm up your own domains, which

[01:02:56] takes a lot of work in 30 days. Here we

[01:02:58] can just pre- buy them. You could also

[01:03:00] connect existing accounts if you want

[01:03:02] to, but for now we just going to choose

[01:03:05] the pre-warmed accounts. We can click

[01:03:07] go. Now we can select some domains. We

[01:03:09] want to find something that seems very

[01:03:10] uh professional like land aiimagine.org.

[01:03:13] That's probably pretty good for an AI

[01:03:14] agency, right? I'm just going to choose

[01:03:16] landi imagine. It's $15 right here. And

[01:03:18] here you can see we get five emails. We

[01:03:20] will set the forwarding domain. And this

[01:03:22] is very important that you set it to

[01:03:23] your own website because if people then

[01:03:24] go to this website, it'll go to your

[01:03:26] own. So I'm going to go to my own

[01:03:28] website, shiny.ai. AI and to paste that

[01:03:30] in right here where it says enter valid

[01:03:32] UL and then click place order. And now

[01:03:34] we just need to check out right here.

[01:03:36] You can see it's $65 to get these

[01:03:38] domains, but that is very very worth it.

[01:03:40] It's $10 per domain. When you have

[01:03:42] bought your email domains, it's going to

[01:03:43] look like this. Now, you can see we have

[01:03:45] five email domains right here from this

[01:03:47] domain right here, frivegateway.com.

[01:03:49] And we can see the health score right

[01:03:51] here next to them. You can see 100%

[01:03:53] health score. Each of these, you can see

[01:03:55] it says email sent can send up to 20

[01:03:57] emails a day. So when we have five

[01:03:59] emails, then we can send 100 emails a

[01:04:01] day, which will be plenty because we're

[01:04:03] going to take a more targeted approach

[01:04:05] with our cold email. We want to get this

[01:04:07] right. So after you have your domains

[01:04:08] and after they are at 100% health score,

[01:04:10] we're ready to start and create our

[01:04:13] campaigns. And of course, we're going to

[01:04:14] be using Claude code for that. And I've

[01:04:16] actually already created a skill that's

[01:04:18] based on ours most high performing

[01:04:19] campaigns that you of course also have

[01:04:21] access to. So let's get started with

[01:04:23] that. Before I show you how to set it up

[01:04:24] with Claude, I wanted to show you this

[01:04:26] campaign right here, which is what you

[01:04:27] should strive towards. And this is also

[01:04:29] what the Claude skill is based on. You

[01:04:31] can see this campaign right here is

[01:04:33] going out to dentists. And we are 60%

[01:04:35] through. We have sent a,000 emails. This

[01:04:37] is both initial emails and then

[01:04:39] follow-ups. Follow-ups in cold email is

[01:04:41] very important. And from that, we have

[01:04:43] 50 people that replied with a 5.1%

[01:04:46] reply rate. 5.1 is extremely high in

[01:04:50] cold email. Usually when you send cold

[01:04:52] email campaigns, you maybe get a 0.5%

[01:04:54] reply rate with a lot of people that say

[01:04:56] [ __ ] you or I don't want to hear from

[01:04:58] you. But from those 50 replies, a lot of

[01:05:00] them will also take me off the list or

[01:05:02] don't talk to me again. That is just a

[01:05:03] part of cold email. But from those 50,

[01:05:05] we have six interested leads. These are

[01:05:08] people that said yes, I'm interested.

[01:05:09] Let's hop on a call. Let's talk more

[01:05:11] about this. Or as a follow-up questions

[01:05:13] that we can then answer and start a

[01:05:14] conversation with. And you can see if

[01:05:16] you can send let's say a 100 emails a

[01:05:18] day, you get five that replies. That

[01:05:20] means that every other day you get an

[01:05:22] interested lead which is extremely good

[01:05:24] and you can run a business on that. So

[01:05:26] let's get into it. Let me show you how

[01:05:27] we create this campaign. Awesome. If we

[01:05:29] open a new Claude code chat, write

[01:05:32] instantly

[01:05:33] campaign like this. This is the

[01:05:36] instantly campaign creation skill. We

[01:05:38] can then write I want to create a

[01:05:41] campaign for let's do Rufus in Sweden.

[01:05:47] And if you are from one of these blue

[01:05:49] countries that I showed earlier, then

[01:05:50] usually what a lot of beginner do is

[01:05:52] that they still go after the English

[01:05:54] market. So after the US, the UK,

[01:05:56] Australia, but because AI is being

[01:05:58] adopted so massively everywhere around

[01:06:00] the world, I would actually say that

[01:06:02] it's a much better advantage that you

[01:06:04] speak some unique language. For example,

[01:06:07] if you speak Swedish, please go after

[01:06:09] Sweden. Just make sure that it's legal

[01:06:11] to reach out with cold email in the

[01:06:13] country that you want to go after. But

[01:06:15] please go after those countries. Why?

[01:06:17] Well, you're going to have a much higher

[01:06:18] reply rate. For example, I don't really

[01:06:21] speak Swedish. I understand a little

[01:06:22] bit, but someone that speaks Swedish

[01:06:24] have an unfair advantage over me by

[01:06:27] knowing the language. Same if you're

[01:06:29] from Greece or from Portugal or from

[01:06:31] Belgium, then you speak a language that

[01:06:34] not the entire world speaks, which

[01:06:36] basically just means that you have way

[01:06:37] less competition. So, just to do it as

[01:06:39] an example and to show you that Claude

[01:06:41] is going to do all of the creation of a

[01:06:43] campaign for us, I'm going to create a

[01:06:45] campaign for roofers in Sweden while not

[01:06:47] speaking Swedish, only understanding a

[01:06:49] little bit. And that's all we need to

[01:06:50] write. Now, we hit enter and Outlaw Code

[01:06:52] is going to build out the campaign using

[01:06:55] the instantly MCP that we have connected

[01:06:57] it to. And you can see it says, "What

[01:06:59] are you selling to Swedish roofers?

[01:07:00] Describe the offer in one paragraph.

[01:07:02] What's what does it do for them?" Let's

[01:07:04] do new website just to show you how it

[01:07:06] looks. What's the primary pain? Slow

[01:07:07] coding loses jobs, few Google reviews.

[01:07:10] Let's do few Let's do a couple like few

[01:07:12] Google reviews, ranking on SEO,

[01:07:17] and also missing out on leads because

[01:07:21] they don't have the right information.

[01:07:25] What tangible asset will email free

[01:07:27] offer? This becomes the main subject

[01:07:28] line. Let's do the lost revenue

[01:07:30] calculator so they can see, okay, how

[01:07:32] much they're saving. What should the

[01:07:33] campaign be called? This doesn't really

[01:07:34] matter, but let's just call it roofer

[01:07:36] Sweden. And then submit answers. And now

[01:07:38] PL code is going to build out this

[01:07:40] entire campaign for us. And here we go.

[01:07:42] We have a draft that goes over

[01:07:44] everything. This means I was thinking

[01:07:46] about calling, but I wanted to write to

[01:07:48] you first. And then saying right here

[01:07:49] that I'm thinking that you're losing

[01:07:51] jobs because your website is not showing

[01:07:52] high on Google. Is this something that

[01:07:54] you would need some help for? I really

[01:07:56] like this. This all absolutely awesome.

[01:07:58] So I'm going to approve. Let's create

[01:08:00] campaign. And now plot is going to build

[01:08:02] out the instantly campaign for us. We

[01:08:04] don't have to do a thing. Now you can

[01:08:05] see it is calling instantly. If I click

[01:08:08] control O, we can see what it does. It's

[01:08:10] making this request right here with the

[01:08:12] information. There we go. Campaign is

[01:08:14] created and called Rufus Sweden. If we

[01:08:17] go into instantly right now, hit to

[01:08:19] campaigns. We can see that the new

[01:08:20] campaign has now been created. If we

[01:08:22] click in on this, we can go to

[01:08:24] sequences, which is basically where we

[01:08:25] write the emails. But it looks

[01:08:26] completely blank for some reason. It

[01:08:28] looks completely blank. I don't see the

[01:08:29] emails. Let's see what happened here.

[01:08:31] Oh, I just had to refresh it. Do a hard

[01:08:34] refresh. Shift command R. That half

[01:08:36] refreshes the page. Now, we have the

[01:08:37] entire campaign inside of instantly that

[01:08:40] we just wrote out. So, we need to make

[01:08:42] sure that we click save. There we go.

[01:08:44] Click save on everyone. Now, we have our

[01:08:46] campaign ready to go. The last thing we

[01:08:48] need is just leads to send to. and I'm

[01:08:51] going to show you the fastest and what I

[01:08:52] found to be one of the cheapest way to

[01:08:54] get really qualified and verified leads.

[01:08:57] Let's get into it. To get qualified

[01:08:59] leads, we're going to use a platform

[01:09:01] called Apollo. But don't worry because

[01:09:04] Apollo itself is actually extremely

[01:09:06] expensive. I'll show you a cheaper way

[01:09:08] to get the same quality leads as Apollo,

[01:09:11] but using a different site. So, let's

[01:09:13] say we want to find roofing and building

[01:09:15] companies in Sweden, for example. We

[01:09:17] then make sure to go to people right

[01:09:20] here. We go to location and we search

[01:09:22] Sweden. Then we go to industry and

[01:09:24] keywords and we could search for that

[01:09:27] would probably be construction. Let's

[01:09:29] maybe also do real estate. If it's like

[01:09:30] home builders, then real estate is

[01:09:32] probably also going to fit. You can see

[01:09:33] now we have a 102,000 people. You

[01:09:36] probably want to narrow it further. So

[01:09:38] you can also do things like job titles.

[01:09:40] You of course want to do go for someone

[01:09:42] like the owner of the business or the

[01:09:45] founder of the business. Add a lot of

[01:09:48] keywords to people that would fit your

[01:09:49] description for your ideal client like

[01:09:51] CEO, founder, CEO, and founder.

[01:09:55] Add all of these because you can see

[01:09:57] that increases the list size right here.

[01:09:59] You could also exclude titles if you

[01:10:00] want to, but now you can see we have a

[01:10:02] list of around 3,000 people right now.

[01:10:04] What's very important that you go into

[01:10:05] email status and then make sure to click

[01:10:07] this verified right here. That's going

[01:10:09] to get you the best quality of emails.

[01:10:12] And you can see our list is now at

[01:10:13] 2.1,000 people. You can also choose if

[01:10:15] it should be B2B only, B2C only, if it

[01:10:18] should be e e-commerce companies,

[01:10:20] fintech companies, nonprofits, SAS,

[01:10:22] consulting. You have a bunch of things

[01:10:24] to choose from inside of here. This one

[01:10:25] is also one of the most important

[01:10:27] filters. You can choose what range you

[01:10:29] want to reach out to. So, how big the

[01:10:31] company should be. If you're just

[01:10:33] starting out, you're probably not going

[01:10:35] to close an enterprise deal right off

[01:10:36] the bat. That would be insane. We could

[01:10:38] probably choose like from 1 to 10 to 11

[01:10:40] to 20, maybe up to 21 to 50. And that

[01:10:44] gives us 1.8,000 people that we could

[01:10:47] reach out to. That way, we leave out the

[01:10:49] biggest companies because you can reach

[01:10:51] out to those in the future, but you need

[01:10:52] some experience first. And there we go.

[01:10:54] Now, we have a bunch of Swedish

[01:10:56] companies and we can access all of their

[01:10:58] emails. But usually, if you're using

[01:11:00] Apollo, you do it directly through the

[01:11:01] Apollo platform, which is extremely

[01:11:03] expensive. What you can do instead is

[01:11:05] that you can use this platform right

[01:11:06] here that's called trustedleads.io

[01:11:08] enterprisegrade B2B lead list starting

[01:11:10] at 0.005

[01:11:13] per lead. So go in there, create a user.

[01:11:16] Again, I'm not affiliated to say this.

[01:11:17] And then you can click new order right

[01:11:19] here. Here you write in your name, your

[01:11:21] email address, write in a phone number,

[01:11:23] and then you want to choose the Apollo

[01:11:25] URL right here and the number of leads

[01:11:27] that you want to scrape. Very important

[01:11:29] that you turn on the verified leads that

[01:11:31] I showed you earlier. Otherwise, you're

[01:11:32] going to get bad lead quality, and

[01:11:34] you're going to be paying for leads that

[01:11:35] are not good enough. So, let's say a

[01:11:36] total of how much money do we want to

[01:11:38] scrape? Like 1,500 of these leads. And

[01:11:41] you can see that's going to cost us $25,

[01:11:44] which is 67% less of what we would use

[01:11:47] on Apollo. So, now we can click

[01:11:49] continue. Now, we want to take this URL

[01:11:52] from Apollo right here and paste it in.

[01:11:54] It says that the email is not verified,

[01:11:56] but I'm almost 100% sure that we set it

[01:11:58] to verified. You can try and refresh

[01:12:00] this page just to make sure that you

[01:12:02] have the newest URL like this. Let's

[01:12:04] copy this one. Paste this in and see if

[01:12:05] it says now that is fine. Let's just

[01:12:07] verify that we have the right search. We

[01:12:09] do. Then click continue. You could if

[01:12:11] you wanted to pay additional to actually

[01:12:13] verify the emails, but I'm going to show

[01:12:15] you a better way of doing that. So,

[01:12:16] we're going to click no additional

[01:12:18] verification. Do a signature and then

[01:12:20] click proceed to payment. After you're

[01:12:22] done with that, then it can take a

[01:12:24] couple of hours, but then you receive an

[01:12:25] email like this. Your order is ready.

[01:12:27] Thank you for your order of B2B leads.

[01:12:28] download the lead lead list right here.

[01:12:30] And this is a previous list that I

[01:12:32] bought and this is actually 10,000

[01:12:34] leads. You can see how many leads we

[01:12:35] have right here. When you buy leads from

[01:12:37] inside of here, it is very smart to do

[01:12:39] it in bulk. You can see here we have a

[01:12:41] lead list of 10,000 leads. What I'm

[01:12:43] going to do now is that I want this as a

[01:12:45] CSV file. So I can click file. I can

[01:12:47] click import, click upload. I can upload

[01:12:50] this CSV file right here. That's going

[01:12:52] to take some time because it's a very

[01:12:53] big file. Replace spreadsheet and import

[01:12:55] the data. And here we go. Here we have

[01:12:57] our big list. We have to do one more

[01:12:59] step. For now, I don't need all of the

[01:13:00] rest of these leads. So, I'm just going

[01:13:02] to delete those like this. This is just

[01:13:04] to show you how you import those into

[01:13:06] instantly. I'm going to call them leads

[01:13:08] and maybe just the date May 16th. And

[01:13:11] I'm going to click file, click download

[01:13:13] as CSV. And then before we put those

[01:13:15] into instantly, we want to run them

[01:13:17] through million verifier first. This is

[01:13:19] basically finding all the emails that

[01:13:20] are old. When you're sending out cold

[01:13:22] email campaigns, you don't want to send

[01:13:23] to email domains that doesn't exist.

[01:13:26] That's like a tell that you are spamming

[01:13:28] with emails which the email providers

[01:13:30] don't like. Instead, you want to run it

[01:13:32] through something like million verifier.

[01:13:33] First, you click select file and we can

[01:13:36] select this one right here. And you can

[01:13:37] see we're going to be using 969 credits

[01:13:39] for this and we're going to be removing

[01:13:41] the duplicates. We click start verify

[01:13:43] and this will now create a job right

[01:13:44] here. You can see it says takes 15

[01:13:46] minutes. This is basically going to find

[01:13:48] all of the duplicates and it's going to

[01:13:49] remove all of the old emails. It's

[01:13:51] basically verifying the emails to make

[01:13:53] sure that they are good to send to. You

[01:13:55] don't want to skip this step. That's

[01:13:56] very important because if you skip it,

[01:13:58] then you're going to be sending emails

[01:14:00] and your domains will instantly be

[01:14:01] backlisted and then you're going to have

[01:14:02] to buy new domains. So, make sure to use

[01:14:04] Million Verify as well. Again, you can

[01:14:06] find links to all of these platforms in

[01:14:08] the description of this video. There we

[01:14:10] go. That took around 10 minutes. And you

[01:14:13] can see now we have a total of 64% of

[01:14:16] emails that were good emails, 17% that

[01:14:18] are risky, and 18% that are bad emails

[01:14:21] that doesn't exist anymore. You

[01:14:23] definitely don't want to be sending to

[01:14:25] these bad emails cuz those emails are

[01:14:27] going to bounce and it's going to hurt

[01:14:29] your domains. Risky emails you can

[01:14:31] sometimes send to, but I would still not

[01:14:33] recommend sending to these inboxes. And

[01:14:36] then we have the 64% that are good

[01:14:37] emails. Those are the ones that we'll be

[01:14:39] sending to. I can now click download

[01:14:41] report right here. And I want the good

[01:14:43] emails only. So I'm going to click right

[01:14:45] here. That's going to give me this file

[01:14:47] right here of only the good emails.

[01:14:49] There we have it. And now it's time to

[01:14:51] add these emails to instantly. So, I'm

[01:14:53] going to go back into instantly. I'm

[01:14:54] going to click on the leads tab under my

[01:14:56] campaign. I'm going to click add leads

[01:14:58] and then from a CSV file. And then I'm

[01:15:00] going to choose the CSV file that we

[01:15:02] just downloaded, which is this one right

[01:15:04] here fromverify.com.

[01:15:06] And you can see we now have to map the

[01:15:08] fields with first name is going to be

[01:15:09] first name, last name, last name. Title

[01:15:12] is going to be job title. LinkedIn URL,

[01:15:14] email. All of these we don't want to

[01:15:16] import. Let's see if there's any one we

[01:15:18] need. Probably not. Company name we

[01:15:20] want. Website we want. industry probably

[01:15:22] not. And we don't want to verify our

[01:15:24] leads because we just did that

[01:15:25] ourselves. Now we can click upload all.

[01:15:28] This will upload 625 contacts to your

[01:15:30] campaign. Let's click yes. And now you

[01:15:33] can see all of our leads are uploaded to

[01:15:35] this campaign right here. These emails I

[01:15:37] got done probably a couple of months

[01:15:39] ago. So we of course not going to send

[01:15:40] emails in Swedish to these US leads. And

[01:15:43] then when we have our leads, last thing

[01:15:45] we need to do is go inside of options

[01:15:47] and then just add our email accounts

[01:15:49] like this. As you can see, these email

[01:15:51] accounts right now are inactive. But

[01:15:52] when you've just gotten them, they're

[01:15:54] going to be good to go. And then you can

[01:15:55] click launch and publish like this. When

[01:15:58] you have your sequence in place, when

[01:16:00] you have your leads uploaded, when you

[01:16:02] have your pre-warmed email accounts,

[01:16:03] then you can click start campaign up

[01:16:06] here in the right corner. And that's

[01:16:07] going to send out all of these emails.

[01:16:09] Another thing that you can change if you

[01:16:10] want to, you can see this is set from

[01:16:12] 7:00 a.m. to 400 p.m. by default. And

[01:16:14] the time zone right here, this is a

[01:16:16] pretty good window, but if you wanted to

[01:16:18] change that, you could, of course, do

[01:16:19] that. You can also send on weekends if

[01:16:21] you want to, but I probably wouldn't

[01:16:22] recommend that. Do it in business hours.

[01:16:24] And here's what's going to happen. When

[01:16:25] you start sending out, you're going to

[01:16:26] start receiving replies. And if you're

[01:16:28] sending to these countries that only you

[01:16:29] speak the where you speak the language

[01:16:31] and not that many else do, then you're

[01:16:33] going to get a very high reply rate. If

[01:16:35] you have made a good campaign with a

[01:16:36] good offer, you'll find those responses

[01:16:38] inside of Unibox. You can click on

[01:16:40] interested and then you can see the

[01:16:41] responses that are interested. And you

[01:16:43] can see this is from a real campaign

[01:16:45] that we sent around a month ago. And

[01:16:46] what you can see is that we are

[01:16:47] receiving interested responses. This

[01:16:49] says, "Yes, I'm interested in a

[01:16:51] conversation." This person right here

[01:16:53] says that they're interested in an offer

[01:16:54] to see what we provide. And in this

[01:16:56] campaign, we have only sent to 278

[01:16:59] leads. I hope you start to see the power

[01:17:01] of coing, especially to countries where

[01:17:04] you speak a unique language. If you want

[01:17:05] to absolutely maximize you get as many

[01:17:07] clients as possible, I recommend running

[01:17:09] Upwork as your main priority. every

[01:17:11] single day, go out and send connection

[01:17:13] requests because those are the hottest

[01:17:15] leads and the easiest way to gain that

[01:17:16] initial experience, which is so crucial.

[01:17:18] And then if you want to maximize, then

[01:17:20] also run a campaign in the background

[01:17:21] cuz this is just going to be running

[01:17:23] sending out emails as you go and

[01:17:24] probably book you a couple of meetings a

[01:17:26] day if you have a good offer and

[01:17:27] offering an interesting service in your

[01:17:29] market. These two things when you're

[01:17:31] just starting out is the highest

[01:17:32] leverage thing that you can do. I

[01:17:33] honestly recommend getting started on

[01:17:35] that right now. But you of course also

[01:17:38] need to know how you can actually

[01:17:39] convert meetings and how you take this

[01:17:41] information that you're getting on

[01:17:42] meetings when people are actually

[01:17:43] interested in proceeding. How you take

[01:17:45] all of this information and give that to

[01:17:47] Claude so Claude can build out and do

[01:17:49] the majority of our service delivery.

[01:17:51] This is going to be extremely valuable

[01:17:52] for you to watch. So let's get into it.

[01:17:54] Great. Now you have outreach going out

[01:17:56] on both Upwork and on Instantly. The

[01:17:59] next part of this course is one of the

[01:18:01] most important because this is where

[01:18:02] most beginners they mess up. And they

[01:18:04] mess up in three places. Both before the

[01:18:06] call, under the call, and after the

[01:18:08] call. So, let's go through some ground

[01:18:10] rules that you need to understand before

[01:18:12] you start booking meetings and closing

[01:18:14] clients. The first one is that you

[01:18:16] never, and I mean it, never mention

[01:18:20] price before the call. And here's the

[01:18:23] reason why you don't do that. If you

[01:18:24] mention the price, let's say that the

[01:18:26] client asks for, "Okay, what is the

[01:18:27] price before I hop on a call?" And you

[01:18:29] say, "Well, it's $2,000 a month." What

[01:18:32] then happens is that the client judges

[01:18:33] you on the price and not the value. The

[01:18:37] client doesn't know enough about your

[01:18:38] service. They don't know what work

[01:18:40] you'll be doing and why it'll give them

[01:18:42] a bigger return than the price that

[01:18:44] they'll pay. So, when they hear a $2,000

[01:18:47] price, what they think about is, "Okay,

[01:18:49] do I want to pay $2,000 right now?"

[01:18:51] Probably not. I'm going to stop replying

[01:18:53] to this guy. And the only reason they do

[01:18:55] that is because they don't know your

[01:18:56] value yet. You only mention the price

[01:18:58] after you're on a call with them, after

[01:19:00] they know how you can help them, because

[01:19:02] it's a way better equation, a way better

[01:19:04] thought process if you show, let's say,

[01:19:06] $10,000 worth of value and then show the

[01:19:09] $2,000 price point. All of a sudden,

[01:19:11] that becomes a no-brainer deal. So many

[01:19:13] beginners make this mistake and they

[01:19:14] throw out a bunch of meetings because

[01:19:16] they don't understand this key

[01:19:18] principle. That is rule one. Rule two,

[01:19:20] that you always want to send a

[01:19:22] confirmation and reminders. It happens

[01:19:25] more often that you think that people

[01:19:26] simply just forgets that they book a

[01:19:28] meeting with you and then they get kind

[01:19:30] of like embarrassed that they forgot the

[01:19:31] meeting and you never hear from them

[01:19:33] again. You always want to send both a

[01:19:35] confirmation and reminders to every

[01:19:38] single call that you hop on. So the

[01:19:39] process looks like this. You have people

[01:19:41] interested either over cold email or

[01:19:43] over Upwork. You decide to okay, let's

[01:19:45] hop on a meeting. And then as soon as

[01:19:47] they have agreed to a time, you send

[01:19:49] them the confirmation. Right? This is

[01:19:51] 400 p.m. Monday. This is where we meet.

[01:19:53] And you send this on as many channels as

[01:19:55] possible. So if you have them on email

[01:19:57] and their phone number, then you send it

[01:19:59] on both as many channels as you can. You

[01:20:01] do this instantly. And then depending

[01:20:03] how long the meeting is out in the

[01:20:04] future, you then also send reminders. I

[01:20:06] recommend that you send reminders at

[01:20:08] least every 3 days up to the meeting.

[01:20:10] Then do it one day before and then do it

[01:20:13] 1 hour before. and then do it 5 minutes

[01:20:16] before. And the way that these messages

[01:20:18] look is basically just like, "Hey, first

[01:20:20] name, just reminding you that we have

[01:20:22] our call tomorrow. Excited to see you

[01:20:23] there." If you're doing like a demo for

[01:20:25] them already, it's a very good idea to

[01:20:27] mention that you've also put worked in

[01:20:28] before the meeting because that then

[01:20:30] makes them feel guilty if they don't

[01:20:31] show up. So, if you build a demo or a

[01:20:34] test website for them or something ahead

[01:20:36] of time, then say, "I've prepared a demo

[01:20:38] for you based on your own company." So,

[01:20:40] you can see exactly how this will look

[01:20:41] for your business. So, if they then

[01:20:43] don't show up, they're going to feel

[01:20:44] guilty. You want to include that in the

[01:20:46] reminders. And then the last reminder,

[01:20:48] like 5 minutes before, you should just

[01:20:50] write, "Hey, first name. I'm just

[01:20:52] finishing up another call. I'll see you

[01:20:53] in our call in like 5 minutes." That's a

[01:20:55] really good message for this last

[01:20:56] reminder. Something I also see, and this

[01:20:58] is connected to never mention price, is

[01:21:01] that some people want to skip this step

[01:21:04] of hopping on a call. The reality is

[01:21:05] that's just making your life 10 times

[01:21:07] harder. Not everyone's going to be ready

[01:21:09] to buy right in that moment. and they

[01:21:11] might need more information. So you need

[01:21:12] to hop on a call with them. You want to

[01:21:14] give them limited amount of information.

[01:21:15] So where they really get the information

[01:21:17] is on the call with you. And it's very

[01:21:19] simple. If you have an issue right now

[01:21:20] where people don't show up to the calls,

[01:21:22] it's just an equation of do they think

[01:21:24] that this call would be valuable enough

[01:21:27] for their time. Do they think it can

[01:21:28] actually help them? And if you have an

[01:21:30] issue where people don't show up to the

[01:21:31] calls, the reality is they don't think

[01:21:33] that it's going to be worth their time.

[01:21:35] That is why you need to include the

[01:21:36] reminders. You need to tell them, I've

[01:21:37] built a demo specifically for your

[01:21:39] business. I've prepared this and this

[01:21:40] and this. We're going to see this and

[01:21:42] this and this. And you of course need to

[01:21:43] solve a painful problem. Okay. So,

[01:21:45] before the call, you never mention the

[01:21:47] price. You always want to be judged on

[01:21:48] value. You always send confirmations and

[01:21:51] reminders on as many platforms as you

[01:21:53] possibly can. And then you always never

[01:21:55] skip the call. You don't sell straight

[01:21:57] away. You hop on a call with them before

[01:21:59] you start selling. Okay. Now, on the

[01:22:01] call, a lot of beginners make the

[01:22:02] mistake of talking too much. The best

[01:22:05] way I found to sell these AI services is

[01:22:08] not by being the one talking. It's by

[01:22:10] taking what I call the doctor approach.

[01:22:12] When you go to the doctor, they don't

[01:22:14] just hand out medicine the moment you

[01:22:16] come in the door. The first thing they

[01:22:17] do is that they ask you a bunch of

[01:22:18] questions like where does it hurt? How

[01:22:20] long has it been hurting? And then they

[01:22:22] try to diagnose what's wrong. You want

[01:22:23] to take the same approach when you are

[01:22:25] selling. And the way you do that is that

[01:22:28] you only talk 20% of the time and they

[01:22:31] should talk 80% of the time. And when

[01:22:34] you talk, it should mainly be questions.

[01:22:36] And when they talk, they should answer

[01:22:38] your questions. And here's why that

[01:22:40] works so well. When you're the one

[01:22:42] asking the questions, you're the one

[01:22:44] leading the conversation. And one thing

[01:22:46] is that you mention why your service is

[01:22:48] valuable. An even better thing is when

[01:22:50] the client themselves mention why your

[01:22:53] service is valuable. When you get people

[01:22:55] to say themselves that they need

[01:22:57] something, they convince themselves way

[01:22:59] better. And the way to do that is by

[01:23:00] asking questions. Let me give you an

[01:23:02] example. If someone has, let's say, only

[01:23:04] a couple of reviews on Google Maps, you

[01:23:06] can then ask a question like, "Okay,

[01:23:08] what do you think happen when a customer

[01:23:10] compares you to your competitor? Your

[01:23:12] competitor has 150 reviews. I can see it

[01:23:15] right here. You only have five reviews

[01:23:17] and one of them is a freestyle review.

[01:23:19] Who do you think your client is going to

[01:23:20] go with?" They're going to say, "Well,

[01:23:22] based on this only, they're probably

[01:23:24] going to go with that client." Do you

[01:23:25] see what we do there? We get them to

[01:23:27] mention themselves that they have a

[01:23:29] problem. And when we take this doctor

[01:23:30] approach, it becomes a much stronger

[01:23:33] cell. And it doesn't even feel like

[01:23:34] selling. You're just saying, well, what

[01:23:36] do you think happens when when this and

[01:23:37] this and this in your current situation,

[01:23:39] they mention a problem. And then it

[01:23:41] makes it much easier later in the call

[01:23:43] to say, well, I actually have a solution

[01:23:44] to your problem. That's the entire point

[01:23:47] of the call. It's to position a problem

[01:23:49] that they actually do have. If someone

[01:23:51] doesn't have a lot of reviews on Google,

[01:23:52] they have a problem that's hurting their

[01:23:54] business. And then positioning you as

[01:23:56] the solution. That way, you don't need

[01:23:58] to convince them. You don't need to be

[01:23:59] overly salesy or talk a bunch. You're

[01:24:01] literally just asking good questions.

[01:24:03] And by asking good questions, you

[01:24:04] control the conversation. You can steer

[01:24:06] it in any way that you want. And that

[01:24:08] leads me to the second rule, which is

[01:24:09] that you need some structure on your

[01:24:11] calls. And the structure that I like to

[01:24:13] use is initially just these doctor

[01:24:15] questions. Initially, just asking if

[01:24:17] they even have a problem. And what

[01:24:19] you'll run into sometimes, let's say

[01:24:21] that you are providing more Google

[01:24:22] reviews with automation and you run into

[01:24:25] someone that already has 500 reviews and

[01:24:27] are on top of everyone else. They

[01:24:29] probably don't need that specific

[01:24:30] service. You might sell them another

[01:24:31] service, but you don't want to sell

[01:24:32] something to someone that doesn't

[01:24:34] actually need your service. That's

[01:24:36] unethical. So, the structure that I like

[01:24:37] to use is one, ask a bunch of good

[01:24:39] questions that positions a pain. Then

[01:24:42] two, tell them that well, we can take

[01:24:45] this pain and we might have a solution.

[01:24:48] Let's see if it even makes sense. And

[01:24:49] then you can show a demo or you can

[01:24:51] explain well we can actually do this to

[01:24:53] fix this problem. And three is that you

[01:24:55] then explain further about this solution

[01:24:58] and say well the price of this solution

[01:25:01] is and only when they've seen the

[01:25:03] solution when they have seen the value

[01:25:05] that you can provide that's when you

[01:25:06] mention the price and that's when you

[01:25:08] shut up. They now get themselves to

[01:25:10] decide okay this pain that I have and

[01:25:12] this solution right here is this worth

[01:25:14] it or is it not worth it? And what will

[01:25:17] usually happen is that they will then

[01:25:19] have some kind of objections and those

[01:25:21] come at the end. Objections are usually

[01:25:22] like questions. Well, what happens if we

[01:25:24] try this for a month and it doesn't work

[01:25:26] or well what happens with my current

[01:25:29] website if we're changing it over? How

[01:25:31] would I do that? It's basically just

[01:25:32] questions from their side that show some

[01:25:34] kind of concern. So, what you then have

[01:25:36] to do is just answer those questions in

[01:25:38] a good way and then make sure that that

[01:25:40] that concern is squashed and then at the

[01:25:42] end that's when you close. That is the

[01:25:44] structure that I like to use where it

[01:25:46] doesn't really feel like a sales call at

[01:25:48] all. This is what you should do. Let me

[01:25:50] show you what you then shouldn't do. A

[01:25:52] lot of beginners, they make the mistake

[01:25:53] of basically just hopping a call and

[01:25:55] just yapping. So, a lot of beginners

[01:25:57] structure look like this. They just

[01:25:59] explain

[01:26:01] their product and usually they basically

[01:26:03] just explain features and then they ask,

[01:26:06] "Do you want to buy after they've done

[01:26:07] that, right?" So, what does this look

[01:26:09] like? This is maybe if you create a

[01:26:11] slideshow that's just about your

[01:26:12] product. this is what we do and you just

[01:26:14] sit there and and go through it. The

[01:26:15] problem with this approach is that you

[01:26:17] don't position a pain. You don't ask

[01:26:19] good questions. You don't make sure that

[01:26:20] they actually need your service. You

[01:26:22] basically let them make the decision

[01:26:23] themselves if okay, do I need this or do

[01:26:25] I not need this? And if you have a

[01:26:26] boring slideshow, chances are that

[01:26:28] they're just going to zone out. All of a

[01:26:30] sudden, they hear a $2,000 price point

[01:26:32] and they say no. All right? So, don't do

[01:26:36] this. Do this structure instead. Right?

[01:26:39] Questions first that position a pain.

[01:26:41] This is questions like, well, your

[01:26:44] competitor has way more Google reviews

[01:26:45] than you. They have a good website as

[01:26:47] well. You have almost no reviews. You

[01:26:49] have a trash website. What do you think

[01:26:50] happens? Well, don't say trash website.

[01:26:52] Say, your website is quite old. Who do

[01:26:55] you think your client is going to go

[01:26:56] with? And just like a first impression.

[01:26:58] You want to say themselves, well, well,

[01:27:00] in this scenario, they're probably going

[01:27:01] to go with my competitor. You position a

[01:27:03] pain that hurts their business. Then you

[01:27:04] position your solution. Well, we can

[01:27:06] install this AI that reaches out after

[01:27:08] every single job and gets you Google

[01:27:09] reviews. And we can also install this

[01:27:11] website. I can build it in 2 days, which

[01:27:13] looks way better than what you currently

[01:27:14] have. Do you think that would help

[01:27:16] increase your first impressions with new

[01:27:18] customers? Again, you even position that

[01:27:20] as a question to make sure that they

[01:27:22] understand the solution is worth the

[01:27:23] price. Great. Great that you now

[01:27:25] understand this. So many beginners mess

[01:27:26] up right here. You don't want to do

[01:27:28] that, but you now understand this, which

[01:27:30] is good. Great. So now you're starting

[01:27:32] to get sales. What do you do? Well, you

[01:27:33] follow rule number three, which is

[01:27:36] always book next call

[01:27:40] on this

[01:27:43] call. Always book next call on this

[01:27:46] call. Here's the reason. When you're

[01:27:47] just starting out and you don't do this,

[01:27:49] there's going to be a bunch of times

[01:27:50] where someone agrees, "Yeah, let's go.

[01:27:52] Let's do it." And then you don't book

[01:27:53] the next call on this call. And what

[01:27:55] then happens is that they ghost you. If

[01:27:57] they have something in their calendar,

[01:27:58] it increases the chances that they

[01:28:01] actually hop on a call with you. Again,

[01:28:02] this rule is extremely important and

[01:28:04] it's going to save you a bunch of deals.

[01:28:05] So, what you basically just do on the

[01:28:06] call is that you say, "Great, happy that

[01:28:08] you are on board. Let's book it into

[01:28:10] calendar already the next time we're

[01:28:12] going to talk and then I'll have this

[01:28:13] and this ready for you." Again, you kind

[01:28:14] of sell the next call. You hype up the

[01:28:16] next call and you make sure that they

[01:28:18] see that you've made an investment now

[01:28:19] in them, that you've spent time so they

[01:28:21] feel bad if they don't show up. And the

[01:28:23] last rule, and this honestly surprises

[01:28:24] me that people still do this to this

[01:28:26] day, is that they don't take payment on

[01:28:28] the call. You always want to have some

[01:28:30] initial commitment. And even if you're

[01:28:32] doing like a free trial, you still take

[01:28:34] the payment information for after just

[01:28:37] like you would with a software. If you

[01:28:38] get a free trial, you still give your

[01:28:40] payment information first. Do that here

[01:28:42] as well. The reason for that is that

[01:28:43] they need to make a commitment and they

[01:28:45] need to make a commitment on that call.

[01:28:47] And you can even write it into the

[01:28:48] contract that if they then know your

[01:28:50] next call that you then charge a $500

[01:28:52] fee for wasting your time. Like you can

[01:28:54] write that into a contract, but you

[01:28:55] always want their payment information on

[01:28:57] file. And the way you would usually do

[01:28:59] that is with Stripe. And what we usually

[01:29:01] do is that we just get like a $1 payment

[01:29:03] link inside of Stripe. If you create a

[01:29:05] $1 a year payment link, what that does

[01:29:08] is that it saves the payment

[01:29:10] information. You can just say that's

[01:29:11] like a $1 down. We'll refund it to you.

[01:29:12] Don't worry. But then you have that

[01:29:14] payment method on file, which means that

[01:29:15] you can charge it any time. And that

[01:29:16] also removes friction later on because

[01:29:18] now you have that payment method on

[01:29:20] file. So let's say that you're doing a

[01:29:21] commission deal or you're just charging

[01:29:23] $2,000 a month. You can just use this

[01:29:24] payment information to set it up. they

[01:29:26] don't have to do anything on their side,

[01:29:27] which again removes friction. So, you

[01:29:30] always want to make sure to take some

[01:29:32] kind of commitment on the call. Even if

[01:29:33] you're doing a free trial, get the

[01:29:35] Stripe $1 payment link. Let me actually

[01:29:37] show you how to do that. You go inside

[01:29:39] of your Stripe dashboard. Then you click

[01:29:41] on payment links right here and click

[01:29:43] create payment link. Then you create a

[01:29:44] new product. You just call it checking

[01:29:47] card, something like that. You do it

[01:29:49] recurring and you do $1. And then you do

[01:29:52] a year like this. And then you click add

[01:29:54] product. You don't need to collect tax

[01:29:56] automatically on this cuz you're going

[01:29:58] to be refunding the payment anyways.

[01:29:59] Create link. And what will then happen

[01:30:01] is that when someone signs up on this

[01:30:02] link, you get this payment link right

[01:30:04] here, which says $1 a year. When someone

[01:30:06] fills out the information right here,

[01:30:08] you get their email and you also get

[01:30:09] their payment information. And then it

[01:30:11] allows you to later to just create a

[01:30:13] subscription for whatever you're

[01:30:14] charging. Let's say you're charging

[01:30:16] $2,000 a month. You can then go in and

[01:30:17] set up that payment without involving

[01:30:19] them. So that's how you create Stripe

[01:30:20] payment links. And trust me, that's

[01:30:22] going to help you get commitment, which

[01:30:23] is going to help you close more deals.

[01:30:24] If you have a problem closing, you're

[01:30:26] using this format right here and you get

[01:30:28] to the point where you mention the price

[01:30:30] and people don't buy it, then use this

[01:30:33] last tip, which is kind of a bonus tip

[01:30:35] cuz it's not always required. Sometimes

[01:30:37] you have such a good offer that this is

[01:30:38] not needed, but I call it the proof of

[01:30:41] concept offer. What is this? Well, this

[01:30:44] is an offer that has made us tens of

[01:30:46] thousands of dollars. Here's how it

[01:30:47] works. It's basically a satisfaction

[01:30:49] guarantee. And you can say this that

[01:30:50] well, this is our satisfaction

[01:30:51] guarantee. What you say is that you get

[01:30:53] payment up front. You always want to get

[01:30:55] commitment, right? So, let's say that

[01:30:56] you are selling a website. You still get

[01:30:58] your payment or whatever you agree on.

[01:31:00] Let's say it's $1,000 first and $1,000

[01:31:02] later. You still get your $1,000

[01:31:05] upfront. And you take that payment on

[01:31:06] the call via Stripe, but then you give a

[01:31:09] satisfaction guarantee, which means that

[01:31:11] I'm going to build this entire thing

[01:31:12] out. So, if we use the website example,

[01:31:15] I'm going to build out the entire

[01:31:16] website for you. So, you can see

[01:31:18] everything that's going to be on your

[01:31:19] website. You can see the finished

[01:31:20] product. If you're not happy with your

[01:31:22] website at that point, then you'll get a

[01:31:25] full refund. So, if you don't even want

[01:31:27] to see how it performs when we give it

[01:31:29] to customers and when we actually launch

[01:31:31] it, you can get a full refund, no

[01:31:34] questions asked. We have given this

[01:31:35] offer a bunch of times. And the great

[01:31:37] thing about this is that one, it removes

[01:31:39] all risk because they can actually make

[01:31:41] an informed decision. They can see

[01:31:42] exactly what they're getting. And if

[01:31:44] they don't like it, they can get a full

[01:31:45] refund. And two, it makes you seem very

[01:31:47] confident and good at what you do. all

[01:31:49] of a sudden they say, "Okay, this guy

[01:31:51] can put money on the line. He will

[01:31:52] refund me the entire thing if I don't

[01:31:54] like what he creates." That makes you

[01:31:55] seem very confident, which again is

[01:31:57] going to increase your close rate. We

[01:31:58] have given this offer so many times, and

[01:32:00] we have never had to refund anyone

[01:32:02] because people have been very happy with

[01:32:03] what we built. So, if you're good at

[01:32:05] what you do, this is such a no-brainer

[01:32:07] offer, and adding this will probably

[01:32:08] just bump up your close rate with like

[01:32:09] 10% instantly. And you can use this for

[01:32:11] any AI build really. It doesn't matter

[01:32:13] if you're building websites, it doesn't

[01:32:14] matter if you're building automations,

[01:32:16] you just want to hop on a call with them

[01:32:17] and show it face to face. Okay, here's

[01:32:19] how it works. Look at this. It does

[01:32:21] this. It does this. It does this. Get

[01:32:23] them to try it. Get them to see how good

[01:32:24] it actually is. And then say, "Great.

[01:32:26] Let's launch it." You don't even need to

[01:32:28] mention the satisfaction guarantee. Just

[01:32:30] mention it on the call when you close

[01:32:31] them. And they're going to come back to

[01:32:32] you and say, "Well, if they don't want

[01:32:34] it, then they can get it refunded." But

[01:32:35] as soon as they commit, and as soon as

[01:32:36] they see, okay, how good of a job that

[01:32:38] you've actually done, they're not going

[01:32:39] to want the refund. They're going to

[01:32:40] want to actually implement it. Great.

[01:32:42] So, on the call, you only talk 20% of

[01:32:44] the time. They talk 80% of the time

[01:32:46] because you use the doctor approach. You

[01:32:49] use this structure right here where you

[01:32:50] start with questions. You position a

[01:32:52] pain. You then position that pain to

[01:32:54] your solution. Exactly. You handle any

[01:32:56] objections and you close the sale. You

[01:32:57] don't want to just have a slideshow

[01:32:59] where you just explain your product

[01:33:00] because you don't position a pain.

[01:33:02] They're going to judge you on your

[01:33:03] price, not on your value, and they're

[01:33:05] not going to feel like your solution is

[01:33:06] a fit to their exact problem. Then you

[01:33:08] always always book the next call on your

[01:33:12] current call. So, they need to know when

[01:33:14] they're going to hear from you again.

[01:33:15] You don't make the mistake of not taking

[01:33:17] payment on the call. You always take

[01:33:19] payment on the call. You get a payment

[01:33:21] information. You get the $1 Stripe link.

[01:33:23] And if you're struggling with closing

[01:33:24] deals, then use the proof of concept

[01:33:26] offer because this removes all risk. You

[01:33:28] still get commitment up front. But if

[01:33:29] they don't like the build afterwards,

[01:33:30] you basically just have a satisfaction

[01:33:32] guarantee where they can get a full

[01:33:33] refund. Awesome. After the call, then

[01:33:36] this shouldn't come as a surprise, but

[01:33:38] you need to do a good job. When you're

[01:33:40] just starting out, your reputation and

[01:33:41] building trust and building up a client

[01:33:43] base is absolutely everything. Make sure

[01:33:45] that you do a hell of a job that you go

[01:33:47] all out for this client and do a hell of

[01:33:49] a job. Number two, a mistake that many

[01:33:51] beginners make as well is that they

[01:33:53] don't continue the calls. Let's say that

[01:33:55] you've set up an AI system. A lot of

[01:33:57] agencies, they just set it up and then

[01:33:58] they never communicate with the client

[01:34:00] again. That's not a good way to do it.

[01:34:02] I'd recommend at least bi-weekly calls.

[01:34:05] So, hop on a call with them every two

[01:34:06] weeks. Go over performance. go over

[01:34:08] things that could be improved. Why is

[01:34:10] that? Well, that's going to decrease

[01:34:11] your churn a lot. Let's say you're

[01:34:13] charging $2,000 a month. Doing

[01:34:15] continuous calls by a weekly will

[01:34:17] probably decrease your turn to something

[01:34:18] like 5% or something where many agencies

[01:34:20] have upwards of 20 to 30% churn. If they

[01:34:23] have 30% churn, it means that a client

[01:34:25] on average only stays for 3 months.

[01:34:27] There's another reason to why we do

[01:34:28] bi-weekly calls. This is two steps that

[01:34:30] almost every beginner's day miss. They

[01:34:32] don't upsell. Let's say that you've just

[01:34:34] helped someone build a very, very cool

[01:34:36] website. you don't stop providing

[01:34:38] services there. Then you ask, well, do

[01:34:40] you also want an AI to pick up the phone

[01:34:42] calls from the people that call you now

[01:34:43] because now you have a good website? Or,

[01:34:45] well, I just helped you build this

[01:34:46] website. Do you also want to send

[01:34:48] additional traffic with Google ads to

[01:34:50] this site to get more customers? You

[01:34:52] always keep upselling. And if you don't

[01:34:54] have it already, then come up with ideas

[01:34:56] that could help that business further.

[01:34:58] Upselling is a way to increase your

[01:35:00] lifetime value, also called LTV, which

[01:35:03] is the total amount of money that your

[01:35:04] client pays you over their lifetime as a

[01:35:06] client. Upselling is the best way to

[01:35:08] increase LTV without increasing the cost

[01:35:10] of acquiring a client. You've already

[01:35:12] acquired this client. It doesn't cost

[01:35:14] you anything more just to hop on a call

[01:35:15] with them and say, "Well, do you also

[01:35:16] want this?" And then charge for that as

[01:35:18] well, of course. And it doesn't stop

[01:35:20] there. And again, another reason to why

[01:35:21] we do bi-weekly calls. You also want to

[01:35:24] ask for referrals. Referrals is another

[01:35:26] way to get more customers without paying

[01:35:28] for anything. You're not paying for

[01:35:29] additional code emails. You're not

[01:35:31] paying for additional connections on

[01:35:32] Upwork. This is free clients if you just

[01:35:34] ask for referrals. So many beginners

[01:35:36] when they start out don't ask for

[01:35:37] upsells and they don't ask for

[01:35:39] referrals. But that's also because that

[01:35:40] they don't do it the right way. Here's

[01:35:42] how you do it. You don't say, "Do you

[01:35:43] know anyone that could use this as

[01:35:45] well?" Instead, you're very specific.

[01:35:46] You want to mention the exact people

[01:35:48] that you want to work with, like, "Do

[01:35:49] you know any plumbers in your area that

[01:35:52] could use this as well?" or do you know

[01:35:53] anyone in this industry or in this

[01:35:55] industry that can use this like do you

[01:35:56] know any plumbers, electricians, people

[01:35:58] that own a landscaping company? You

[01:36:00] basically just name a list of potential

[01:36:02] clients that you want to work with. The

[01:36:03] reason that you do this is if you just

[01:36:05] ask well do you know anyone else that

[01:36:06] could use this service? They might come

[01:36:08] with someone but usually they're just

[01:36:10] going to say no because it's harder for

[01:36:11] the human brains to just come up with

[01:36:13] something from scratch. If you give them

[01:36:15] a specific list like do you know any

[01:36:16] plumbers, electricians, landscapers? Do

[01:36:19] you have any other friends in business?

[01:36:20] If you already give their brain that to

[01:36:22] think about, usually they're going to

[01:36:24] come up with more referrals to you.

[01:36:25] That's rule number one when asking for

[01:36:27] referrals. Rule number two is that you

[01:36:29] always want to give a commission. If

[01:36:31] someone refers you to a very, very big

[01:36:33] company, you can give them 20% of your

[01:36:35] revenue. That is fine. You're going to

[01:36:37] get way more referrals if you actually

[01:36:38] incentivize people to refer you, saying,

[01:36:41] "Well, you're going to get 20% of

[01:36:43] whatever they pay me. I'm going to pay

[01:36:44] that to you." Do that to all your

[01:36:46] clients, and you're going to get way

[01:36:47] more referrals. I hope you see how all

[01:36:49] of these are connected. You do a hell of

[01:36:51] a good job. You need to make the client

[01:36:53] happy. And a part of making the client

[01:36:55] happy and treating them well is also to

[01:36:57] giving them the time every two weeks.

[01:36:59] And only if you do a good job and they

[01:37:01] also want to refer you to their friends.

[01:37:02] It all stems from actually being good at

[01:37:04] what you do and doing a good job. Great.

[01:37:06] I hope you found this section of the

[01:37:07] course valuable. I would honestly take

[01:37:08] all of these things, write them down so

[01:37:10] you have them and so you remember them.

[01:37:12] look over them daily cuz these rules are

[01:37:14] literally going to save you thousands of

[01:37:16] dollars if not tens of thousands of

[01:37:17] dollars through your AI agency career.

[01:37:19] All right, the next part of the course

[01:37:20] is also very important because that's

[01:37:22] how you actually do the service delivery

[01:37:23] and I have an insane trick that's going

[01:37:25] to save you so much time prompting

[01:37:26] claude and basically have the product

[01:37:28] delivery ready before you even leave the

[01:37:30] call with the client. It's insane. Let's

[01:37:32] get into that.

[01:37:35] What most agencies do when they are done

[01:37:37] after a call is that they get to work.

[01:37:39] And if you're using something like

[01:37:41] Claude code, you're then taking the

[01:37:43] information that you yourself learned

[01:37:45] from the call and you're passing that to

[01:37:46] Claude code, right? You're spending time

[01:37:48] giving it all the context that you

[01:37:50] gained on the call. But human mistakes

[01:37:52] happen. You might have missed something.

[01:37:54] You might have forgotten something.

[01:37:55] Sometimes you don't start on the project

[01:37:57] right after the call is finished, which

[01:37:58] means that you might lose out on some

[01:38:00] information. Let me show you both a

[01:38:02] smarter way and a way where you never

[01:38:03] miss information ever again where you

[01:38:05] pass all the context effectively to

[01:38:08] claw. To do this, we're going to go back

[01:38:09] into the claw app that we installed

[01:38:11] earlier. If you don't have it already,

[01:38:13] then install claw for desktop. You go to

[01:38:15] this settings tab right here. Then you

[01:38:18] go to connectors and now they moved it.

[01:38:20] So you go into customize. That's going

[01:38:21] to show all your MCPS right here. Click

[01:38:23] this plus right here and click browse

[01:38:25] connectors. and then search for Fathom

[01:38:28] or the noteaker that you're using. On my

[01:38:30] calls, I'm using something called that's

[01:38:31] called Fathom which basically records my

[01:38:33] calls and it gives me a transcript. This

[01:38:36] makes it very very easy for us to just

[01:38:38] say, "Hey Claude, I just finished the

[01:38:40] call. Go in and check the transcript and

[01:38:42] get all of the context. I'll recommend

[01:38:43] that you click this from needs approval

[01:38:45] to always allow." And now the Fathom MCP

[01:38:47] is connected. So if we go down and

[01:38:50] restart Claude because every time you

[01:38:52] install a new MCP, you need to restart

[01:38:54] Claude. and we then write slashmcp. You

[01:38:57] can now see that the fathom MCP has been

[01:38:59] connected. And if I click in on it and I

[01:39:01] click view tools, you can see we can get

[01:39:02] list teams, list meetings, find person.

[01:39:06] And these two are the probably the most

[01:39:07] important. Get meeting summary and get

[01:39:09] meeting transcript. So let's test the

[01:39:11] connection. So let's go in and find this

[01:39:13] call that I had with the team on March 6

[01:39:16] and say I had this call. Please use

[01:39:20] Fathom MCP and give me a summary. There

[01:39:24] we go. You can now see it says

[01:39:26] composing. And now you can see it's

[01:39:27] calling Fathom right here. And there we

[01:39:29] go. Now we get a summary of the meeting

[01:39:31] and we can get all of the context and we

[01:39:33] can ask specific questions to parts

[01:39:35] about the meeting. How we build this,

[01:39:36] how we built that. This is extremely

[01:39:38] useful if you're just starting out

[01:39:40] because then Claude have insights into

[01:39:42] everything and can help you build

[01:39:43] anything that you want that you say on

[01:39:45] the call that you can build. So if a

[01:39:47] client wants a website that can do some

[01:39:49] specific thing, you can have Claude read

[01:39:50] through this transcript, build the

[01:39:52] entire plan for you, and then build

[01:39:53] everything out from you directly based

[01:39:55] on the call. Very, very useful to know.

[01:39:57] You should use this for every single

[01:39:58] meeting that you finish. All right, let

[01:40:00] me show you how to do the service

[01:40:01] delivery so you can start building cool

[01:40:02] stuff with Claude. If we do a small plot

[01:40:05] search and we basically just ask how

[01:40:06] many businesses in the US are still

[01:40:08] missing a website, what we will find is

[01:40:11] approximately 30% of US small businesses

[01:40:14] don't have a website in 2026. Right now

[01:40:17] there's approximately 36 million small

[01:40:19] businesses in the US, which means that

[01:40:22] we have around 10 million businesses

[01:40:24] without a website. And that's also why I

[01:40:26] previously showed you these three levels

[01:40:27] of things that you can build right off

[01:40:29] the bat where website is level one. The

[01:40:32] reason for that is that what this entire

[01:40:34] business model is about is that you

[01:40:36] learn skills, you learn how to do

[01:40:37] something well, where businesses would

[01:40:39] rather pay you to do it than do it and

[01:40:42] find out how to do it themselves. And

[01:40:43] building websites is one of those things

[01:40:45] that still have a lot of perceived

[01:40:46] value. If people see a very nice

[01:40:48] website, they think, "Okay, that must

[01:40:50] have cost a lot." Even to this day where

[01:40:52] AI is so good at building websites,

[01:40:54] there's a lot of perceived value in

[01:40:56] websites. And that is also why it's what

[01:40:58] we're going to start off with building.

[01:41:00] Here's an example of the website that

[01:41:01] you're going to learn how to build. And

[01:41:02] I'll even give you this template right

[01:41:04] here. So, you can actually build exact

[01:41:06] websites like this from one prompt. You

[01:41:09] can see we have this navbar right here

[01:41:11] that when we scroll that becomes like a

[01:41:12] sticky nav bar at the top. I really like

[01:41:14] this design, the choice of colors, the

[01:41:16] choice of fonts. It looks very, very

[01:41:18] professional. And what you'll also be

[01:41:20] learning in this next part of the module

[01:41:22] is how also make these websites actually

[01:41:24] high converting because that's one of

[01:41:25] the places where a lot of website

[01:41:27] designers, they mess up. It's because

[01:41:29] they create websites that might look

[01:41:30] good, but they are not high converting.

[01:41:32] You can see there's a very bright button

[01:41:33] right here that you would want to click

[01:41:35] on. And you also have a number right

[01:41:37] here that you can call if you want to

[01:41:38] call them straight away. You'll also

[01:41:39] learn how to create these very, very

[01:41:41] clean animations directly with code and

[01:41:44] plot code. You can see this little

[01:41:45] plumbing animations right here from the

[01:41:47] pipe and this calendar animation right

[01:41:49] here. All of these you'll learn how to

[01:41:51] create and you'll learn how to prompt

[01:41:52] claude to get a result that looks this

[01:41:54] good. You'll also learn how to create

[01:41:56] these type of animations which are kind

[01:41:57] of like these cards where they stack on

[01:41:59] top of each other which looks very

[01:42:00] professional. Then you'll learn how to

[01:42:02] split this into a dark mode that creates

[01:42:04] like a very nice contrast. And at the

[01:42:06] end you're going to have this form right

[01:42:07] here and I'm going to show you how you

[01:42:08] can hook this website up to any CRM that

[01:42:10] you want. This is a very nice website

[01:42:13] and usually web agencies can charge a

[01:42:15] lot. If I create a new chat and I ask

[01:42:17] Claude, "How much does web agencies

[01:42:20] charge for a nicel looking website?" The

[01:42:23] result will probably surprise you. Look

[01:42:24] at this. Depends on a lot what nicel

[01:42:27] looking means, but rough market ranges.

[01:42:29] For a solo freelancer on a template

[01:42:31] where it takes 1 to 2 weeks, you can

[01:42:32] charge 1 to 3k. If it's a small

[01:42:34] boutique, then it's 3 to 10k. If it's a

[01:42:37] proper agency, it's 10 to 30k. If it's a

[01:42:40] mid-tier agency, it's 30 to 80k. And if

[01:42:44] it's a ward tier, like a big web agency

[01:42:46] where they really spend a lot of the

[01:42:48] time on the design, they can charge 80

[01:42:50] to $250,000.

[01:42:52] And if it's some of the very well-known

[01:42:54] brands, they can charge $250,000.

[01:42:57] And you can even see for my agency

[01:42:58] recommend to $10,000 for websites. So

[01:43:01] you can see how much you can actually

[01:43:02] charge. And again, to set your

[01:43:04] expectations, you're not going to go out

[01:43:06] and close your first website to $10,000.

[01:43:08] But what you might go out is close a

[01:43:10] website for $500, which would still be a

[01:43:12] massive win with users starting out. And

[01:43:14] we can build all of this from inside of

[01:43:16] Claude Code. I'm going to show you how

[01:43:17] to build it. Of course, I have a skill

[01:43:19] that creates these websites in literally

[01:43:20] just like 5 minutes. I'll show you how

[01:43:22] you hook this up to any CRM so the

[01:43:25] information that is captured when people

[01:43:26] click get an offer is actually being

[01:43:29] sent to somewhere. And I'll show you how

[01:43:30] to make your website responsive so it

[01:43:32] looks good on both desktop and mobile.

[01:43:34] And I'll also show you how you can host

[01:43:35] and deploy a website for free up to a

[01:43:38] certain amount of traffic. But usually

[01:43:40] if you're doing it for smaller

[01:43:41] companies, the hosting would be free.

[01:43:42] You're going to learn a lot in this next

[01:43:44] part. So let's get into it. To build

[01:43:46] clean websites like this, you're going

[01:43:47] to be using the build premium website

[01:43:49] skill. And again, you can find all of my

[01:43:50] claw skills inside of our completely

[01:43:52] free community. Just go inside of

[01:43:54] classroom, then learning hub, scroll

[01:43:57] down until you find my claw skills right

[01:43:59] here. I'll also leave a link to this

[01:44:00] right below this video. This drive

[01:44:02] folder includes all of my claw skills

[01:44:05] because after you've installed this

[01:44:06] skill and you write slash build premium

[01:44:09] website, you can see this skill builds a

[01:44:11] premium animated marketing website with

[01:44:14] reactiv and tailwind CSS for any

[01:44:16] industry. Use when the user has to build

[01:44:19] a website blah blah blah. So we're going

[01:44:21] to use it and write /bu premium website.

[01:44:24] I want to build a website for a

[01:44:26] landscaping company. Let's just call it

[01:44:28] cityscape. Then we hit enter because

[01:44:31] what this skill will now do is that it's

[01:44:33] going to force Claw to ask follow-up

[01:44:35] questions about exactly what side it is

[01:44:37] that we want. So the colors, etc. There

[01:44:39] we go. Now it's going to ask us what the

[01:44:41] theme should be. I think a bold modern

[01:44:43] theme looks good. The brand color

[01:44:45] direction, urban slate, m forest, rust,

[01:44:48] olive, brass. I think this dark green

[01:44:50] one, Mars green could look pretty cool.

[01:44:51] So let's go with that. Let's go with

[01:44:53] design. Heartscaping lawn trees lighting

[01:44:55] commercial for the services. I think

[01:44:57] that's pretty good. And then I'm going

[01:44:58] to ask it to build it in this folder

[01:45:02] that we are currently

[01:45:05] in. I'm going to write submit answers.

[01:45:07] And now it's going to build out the plan

[01:45:09] for this website. You can see it's

[01:45:11] reading all of these files from inside

[01:45:12] of the skill that it uses as a reference

[01:45:15] for how it should design this. Building

[01:45:16] this out can take a couple of minutes.

[01:45:18] And there we go. That took 8 minutes and

[01:45:21] 41 seconds. And now you can see site is

[01:45:24] live at localhost 5173.

[01:45:27] So, I'm going to copy this. Then I'm

[01:45:29] going to go into our browser, paste this

[01:45:32] in, hit enter, and there we go. Now we

[01:45:34] have it. Cityscape, home, services,

[01:45:37] approach, process, contact. Then we have

[01:45:39] request a quote right up here. Landscape

[01:45:41] architecture for the city. We design,

[01:45:43] build, and maintain. And now we have a

[01:45:46] clean website in literally just 8

[01:45:48] minutes. That's based on this template.

[01:45:50] And you can see we have this animation

[01:45:52] right here with these stock images,

[01:45:53] which looks extremely good. We're

[01:45:56] missing one right here. So we need to

[01:45:57] fix that. Every discipline under one

[01:45:59] roof from a single tree prun to halfacre

[01:46:01] estate planned from scratch. This looks

[01:46:04] extremely good. Request a quote. When

[01:46:06] they click on this, it scrolls to this

[01:46:07] section right here where they can fill

[01:46:09] in this information. They can even

[01:46:10] attach photos if they want to and then

[01:46:13] they can send the inquiry. Absolutely

[01:46:15] insane website. You can see this website

[01:46:17] right here is built from the same

[01:46:19] template that we used for the plumbing

[01:46:20] company. I'm really happy with how this

[01:46:22] site turned out. We can also right click

[01:46:26] and click inspect and then turn it to

[01:46:29] mobile right here. And then we can see,

[01:46:30] okay, it also looks good on mobile,

[01:46:33] which is also very important for

[01:46:34] websites these days. Great. You've

[01:46:36] learned how to create insane looking

[01:46:38] websites. And you can sell this for

[01:46:40] probably around 500 bucks. But let me

[01:46:42] show you how to also deploy it. You

[01:46:44] deploy it by first going to GitHub. If

[01:46:46] you don't have GitHub already, then

[01:46:48] create a profile. Then click new. And

[01:46:50] I'm going to call this cityscape right

[01:46:53] here. You're going to set it to public

[01:46:55] because you don't want to give everyone

[01:46:56] your code to your website. And then you

[01:46:58] click create repository. There we go.

[01:47:01] And now you can take this URL right here

[01:47:03] and you can pass it to cloud code. And

[01:47:04] you can say push this code for the

[01:47:07] website only to this repo. Then you

[01:47:10] don't want to have it in auto mode

[01:47:12] because it won't be able to push to repo

[01:47:14] in auto mode. So you're just going to

[01:47:15] set it like this. Hit enter. And now

[01:47:17] claude is going to push all of this code

[01:47:19] into the cloud into GitHub. And you can

[01:47:21] see it's first checking. Hit enter. When

[01:47:23] you do this the first time, it'll pop up

[01:47:25] with an authentication where you have to

[01:47:27] authenticate. In order to do this, it'll

[01:47:29] basically just open a window on your

[01:47:31] GitHub where you just have to click

[01:47:33] accept. But because I've done this

[01:47:34] plenty of times, it already has a

[01:47:36] connection. Now, it is creating a

[01:47:39] commit. And you can see pushed code is

[01:47:41] live at this GitHub UL right here. If we

[01:47:44] then hit enter on the GitHub repo, then

[01:47:46] we can see our code from this website.

[01:47:50] And hosting it is actually extremely

[01:47:52] easy. when we have it on GitHub. I'm

[01:47:54] going to use the platform that's called

[01:47:55] visil.com which is a hosting platform.

[01:47:58] Then if you don't have it already,

[01:47:59] create a user. It's free. Go in the top

[01:48:02] right corner, click add new, and then

[01:48:03] project. And then you can paste in this

[01:48:06] GitHub repo right here when you have

[01:48:09] your GitHub account connected to your

[01:48:11] Visel account. This makes it very very

[01:48:13] easy because VIL can now see

[01:48:15] automatically. Okay, this is a V

[01:48:18] project. It's called Cityscape. And all

[01:48:21] we have to do is just click deploy once.

[01:48:24] And there we go. Now you can see it's

[01:48:26] being deployed. This should take usually

[01:48:27] around something like 30 seconds before

[01:48:29] the website is deployed. This actually

[01:48:31] only took 9 seconds this time. You can

[01:48:33] see it says congratulations. You've now

[01:48:35] deployed a new project. We can then

[01:48:37] click continue to dashboard. And here we

[01:48:38] have our project. We actually already

[01:48:40] have a domain that's given by VIL. If we

[01:48:43] copy this domain right here and we open

[01:48:45] it in a new tab, this is the domain that

[01:48:47] we've been given. Then we actually have

[01:48:49] a preview of what the site will look

[01:48:50] like when it's live. And it's actually

[01:48:52] live right now. We can go to this

[01:48:53] website from anywhere. And you can see

[01:48:55] it looks extremely good. Looks exactly

[01:48:57] like it did on development. If you

[01:48:59] wanted to add a domain, you would come

[01:49:01] down here onto domains on the left hand

[01:49:03] side. And you can see we have our

[01:49:05] standard domain right here. You would

[01:49:07] just click add existing. And here you

[01:49:08] can add any domain that you want from

[01:49:11] GoDaddy, from NameCheep, from

[01:49:13] Cloudflare, it doesn't matter. It will

[01:49:14] take you through the review process of

[01:49:16] setting up the DNS. And when you have

[01:49:18] done that, then you have deployed your

[01:49:19] website to the web. If you wanted to

[01:49:21] sell these kind of websites, I would

[01:49:22] create five different templates and then

[01:49:24] create a portfolio website where you

[01:49:26] show these websites that you have built.

[01:49:28] That makes it much easier for a

[01:49:29] potential client to make a decision.

[01:49:31] Okay, should I work with this company or

[01:49:33] not? And if you're using the method that

[01:49:35] I showed previous in this course, which

[01:49:36] is signing up for Upwork, then you can

[01:49:38] find plenty of work. For example, this

[01:49:41] one right here. Posted yesterday website

[01:49:42] for a construction company, logo and

[01:49:45] website design specialist, website

[01:49:47] design and launch looking to make a

[01:49:49] website. There are plenty of jobs in

[01:49:51] doing this and this is one of the best

[01:49:53] ways to land your first couple of

[01:49:54] clients. The great thing about this is

[01:49:56] that when you have built up a portfolio

[01:49:58] and when you have built a bunch of

[01:49:59] skills like this, I'll show you later in

[01:50:01] this course how you can create something

[01:50:03] and then create your own skill so you

[01:50:05] can replicate the process much much

[01:50:06] easier. your work becomes almost

[01:50:08] automated. Claude can do 95% of your

[01:50:11] product delivery and that's the entire

[01:50:13] point of this business model. It's to

[01:50:15] build stuff with Claude and then

[01:50:17] creating skills from it to replicate the

[01:50:19] process and making the service delivery

[01:50:21] so much easier. All right, that is

[01:50:23] websites complete. The next thing you're

[01:50:25] going to learn is how do you actually

[01:50:26] automate things? If a client comes to

[01:50:28] you and says, well, I want my invoicing

[01:50:30] process automated or I want my

[01:50:33] onboarding process automated. How do you

[01:50:35] actually go about doing that? How do you

[01:50:36] create automations with cloud code?

[01:50:38] Usually, you would use a platform like

[01:50:40] make.com or like nadm. But with cloud

[01:50:44] code, there's a much better way of doing

[01:50:46] it. Now, we can use the platform that's

[01:50:47] called trigger.dev. Trigger is an open-

[01:50:50] source platform for building and hosting

[01:50:52] automations. You can see that it's open

[01:50:54] source right here. It has 15,000 GitHub

[01:50:57] stars. And we can see the entire

[01:50:58] codebase right here. Why is it smart to

[01:51:00] use trigger.dev over something like

[01:51:02] Naden or make? It's a very simple

[01:51:05] reason. The reason is that trigger.dev

[01:51:07] is built with code and cloud code is

[01:51:10] really good at coding. We actually

[01:51:11] already have a skill that's called

[01:51:13] trigger.dev and this skill is going to

[01:51:15] be very important when we build out

[01:51:16] those workflows. What I'm going to do is

[01:51:18] that I'm going to clear the session

[01:51:19] inside of cloud code and let's build out

[01:51:22] a couple of automations with trigger.dev

[01:51:24] so you can see how it works. I'm going

[01:51:26] to write I want to automate my invoicing

[01:51:29] process. I want to be able to fill out a

[01:51:32] form with a bit of information and then

[01:51:35] it should build the invoice

[01:51:39] export as PDF

[01:51:42] and then send via Gmail to the client.

[01:51:46] Use two skills and you will have access

[01:51:48] to both of these skills as well. First

[01:51:51] one is the composio skill. I want to use

[01:51:55] composio

[01:51:57] for authentication. Skill two is

[01:52:01] trigger.dev.

[01:52:02] I want to build and host this with

[01:52:05] trigger.dev.

[01:52:09] I'm going to give it an ultraink. Put it

[01:52:10] in plan mode. And let's build this out.

[01:52:12] Remember what I showed you before with

[01:52:14] the process of connecting claw directly

[01:52:17] with your AI noteaker like Fathom via

[01:52:19] the MCP. Right now I explained what I

[01:52:22] wanted, right? But if you're talking to

[01:52:24] a client, I want this and this and this

[01:52:25] automated. You can take that directly,

[01:52:27] that context directly from the call into

[01:52:29] cloud code and then you don't even have

[01:52:31] to explain it like I did here. What

[01:52:33] fields should the invoicing form

[01:52:34] capture? It should capture it all. So,

[01:52:38] business information, contact

[01:52:41] information, line items,

[01:52:46] text and discounts,

[01:52:49] everything that we usually

[01:52:53] have on an invoice. It should create its

[01:52:57] own invoicing number and it should take

[01:53:01] the date automatically. Then we hit

[01:53:04] enter and there we go. It's now going to

[01:53:06] plan out the process of automating this

[01:53:09] process of creating invoices. For how we

[01:53:11] trigger the automation, I think we

[01:53:12] should use a simple web form from

[01:53:14] Nex.js. So I'm going to choose simple

[01:53:16] web form and hit enter. For the

[01:53:17] environmental value, it's very important

[01:53:19] that they're not hardcoded but in av

[01:53:22] file. these let's use react for the PDF

[01:53:26] and for the invoice number let's do it

[01:53:27] date based and then we hit enter if you

[01:53:30] know how to build with cloud code it

[01:53:31] almost doesn't matter what it is that

[01:53:33] you're building if you know the process

[01:53:34] of first planning it out making sure

[01:53:36] that you actually give it the right

[01:53:37] context so it knows exactly what you

[01:53:39] want then the opportunity is literally

[01:53:41] endless you can build basically anything

[01:53:42] that you want if you know the actual

[01:53:44] formatting and the method of building

[01:53:46] stuff here it's asking about the invoice

[01:53:47] counter I think it's a bit overkill to

[01:53:49] set up the invoice counter in a database

[01:53:51] I'm just going to say let's skip the

[01:53:53] invoice counter for now

[01:53:58] and just do it date based the naming

[01:54:03] for the invoicing I think both attaching

[01:54:06] the PDF and storing it in Google Drive

[01:54:08] is going to be best who writes the

[01:54:10] subject fixed template with placeholders

[01:54:12] let's actually do it AI generated so the

[01:54:14] email is AI generated and then we hit

[01:54:17] enter now it's going to plan out exactly

[01:54:19] how we can do this and there we have it

[01:54:21] we have an entire plan for how we should

[01:54:23] build this out. What's worth noticing is

[01:54:25] that we are using Composio. If you don't

[01:54:27] know what Composio is, it's a tool for

[01:54:29] authentication. So, we can authenticate

[01:54:31] to a bunch of different things very,

[01:54:33] very easily. I made a full guide for

[01:54:35] Composio on my channel. So, if you want

[01:54:37] to, you can go and watch that later. And

[01:54:39] then it's using trigger.dev for building

[01:54:41] and hosting this automation. The great

[01:54:44] thing about trigger.dev is that it's

[01:54:45] open source. And when it's open source,

[01:54:47] we can host this anywhere we want. The

[01:54:50] easiest thing is probably just to host

[01:54:51] it with trigger.dev and that's how they

[01:54:53] make money as well. But if you wanted to

[01:54:55] or if your client wants this on their

[01:54:57] own servers, then you can literally just

[01:55:00] build these workflows and host them

[01:55:02] wherever you want. That's the power of

[01:55:04] open source. So now we have a plan.

[01:55:06] Let's take yes and use auto mode and

[01:55:09] let's build out this automation. There

[01:55:11] we go. Now we can see that it is built.

[01:55:13] Let me open another terminal. The first

[01:55:15] thing we do is that we need to fill out

[01:55:17] this envir environmental file right

[01:55:19] here. I'm going to click new file. Add

[01:55:22] add ain file. Copy this example over.

[01:55:26] And then we need to fill out all of this

[01:55:28] information. So the first thing we need

[01:55:30] is a trigger secret key. If you don't

[01:55:32] have it already, then make sure to

[01:55:33] create a user on trigger and log in. I'm

[01:55:36] going to create a new project. So in the

[01:55:38] top left pan, I'm going to create new

[01:55:40] project. I'm going to call this one

[01:55:41] invoicing and just click create. Now we

[01:55:44] can head under API keys and then copy

[01:55:46] this secret key right here. So we paste

[01:55:48] that in where it says secret key. And

[01:55:50] then we also need to go in and find the

[01:55:54] project ref right here. Copy this one as

[01:55:57] well. Now we need the Composio API key.

[01:55:59] So we go to Composio and Composio is

[01:56:02] basically what manages our

[01:56:03] authentication. If you don't have a

[01:56:05] Composio account then sign up for one.

[01:56:07] It is free up to a certain amount of

[01:56:09] tool calls. Copy the composure API key

[01:56:13] and insert that as well. Now we need the

[01:56:15] enthropy API key. This is used for the

[01:56:17] AI generation of the email. So we're

[01:56:19] going to platform.plot.com.

[01:56:22] Go into API keys. Click create key. Call

[01:56:26] this invoicing.

[01:56:29] Click add. Copy this key. Paste that in

[01:56:33] as well. We need the Google Drive folder

[01:56:36] ID. We get that by going to drive. We

[01:56:39] just create a new folder. I'm going to

[01:56:40] call this invoices. If we click in on

[01:56:42] that, we get this ID right here in in

[01:56:45] the header. So, I'm going to copy this

[01:56:46] one. Then, we write all of this

[01:56:48] information, which is our business info.

[01:56:50] So, let's just try and see if it works.

[01:56:52] I'm going to write test street one two

[01:56:55] three building 4,

[01:56:58] business phone number. Just write in

[01:57:00] some random phone number. Business logo

[01:57:02] URL. We can go to our shiny website. We

[01:57:06] can copy this image URL right here. That

[01:57:09] way we don't need to host it, which is

[01:57:11] smart. Business bank details. Let me

[01:57:14] just write a random eban. Default

[01:57:16] currency USD. Default tax rate zero.

[01:57:20] Default payment terms net 14. Default

[01:57:22] due days. This looks good. Awesome. Then

[01:57:25] we close down this environmental file

[01:57:27] and we say great I have filled out the

[01:57:31] invent file. What now? How do we or

[01:57:37] now? Run the one time composure or

[01:57:39] script. It'll print out two or URLs.

[01:57:41] Gmail drive. You click through both and

[01:57:43] it waits until each connection flips

[01:57:44] active. Let me just create a new API

[01:57:46] key. Maybe it was an old one that's

[01:57:48] already rotated. I'm going to write try

[01:57:50] again. We still had a bug. This is a new

[01:57:52] bug. So, Claude can just fix that. There

[01:57:54] we go. Now, you can see it says Gmail

[01:57:56] oorthth URL. So, it's going to give us

[01:57:58] this URL right here that I can copy and

[01:58:00] put into our browser. And this is how

[01:58:03] composio works. This means that we don't

[01:58:05] have to set up the Google credentials

[01:58:07] ourselves. We can use composio that just

[01:58:09] gives us this link right here. And then

[01:58:11] click continue. And you can see now

[01:58:13] composio is connected to Gmail. I can

[01:58:15] now say done. Come with all URLs please.

[01:58:19] So we can connect. And for each service

[01:58:21] that we connect, it's going to give us a

[01:58:23] new URL. So now we have the Google Drive

[01:58:26] URL right here. Paste that in once

[01:58:27] again. Log in with our Google account.

[01:58:29] There we go. And now you can see. Then

[01:58:31] open localhost 3001 server is still not

[01:58:35] running. So I'm going to ask please run

[01:58:38] the dev server for me. And there we go.

[01:58:41] Now the two servers are started both the

[01:58:43] back end which is trigger.dev and our

[01:58:45] front end. We can go to localhost 3001.

[01:58:48] And now we have send an invoice

[01:58:49] generates a PDF drafts with email with

[01:58:51] claude and ships it via Gmail from

[01:58:53] default currency USD text 0% terms net

[01:58:56] 14 due in 14 days. So I can write the

[01:58:58] plan's name is let's do Albert testing

[01:59:01] and let's do the company as let's just

[01:59:03] call it agent and the address we maybe

[01:59:06] don't need to I can do like

[01:59:08] king street 420

[01:59:12] line items we can do consulting services

[01:59:16] one quantity price let's do 1500 we need

[01:59:19] another which is upfront fee let's do

[01:59:22] one call that 500 currency money and

[01:59:25] date currency which be will be USD tax

[01:59:27] rate zero

[01:59:28] Discount type none. Discount value none.

[01:59:31] Issue date, let's do that. The 18th. Due

[01:59:34] date, we can just leave that. Payment

[01:59:36] terms net 14. Any notes? No. And then we

[01:59:39] can click generate and send. Of course,

[01:59:41] we need an email right here. Let me just

[01:59:43] write in my own email just as a test.

[01:59:46] And then click generate and send. You

[01:59:47] can see it says cute run ID and then

[01:59:50] this run right here. Let's see if it

[01:59:52] works. Let's ask cloud now to check the

[01:59:54] logs. Did my test work? You can see it

[01:59:57] checked the logs and found a couple of

[01:59:58] issues. So, let's let Claude fix those.

[02:00:00] Keep doing runs and you can see Claude

[02:00:03] fixed one issue. Let's try and check the

[02:00:06] logs again. Ran into another bug. We'll

[02:00:09] fix that. Try and submit once again.

[02:00:11] Check now. And we just keep doing that

[02:00:13] until Cloud has fixed all of the bugs.

[02:00:15] And you can see Composia won't us

[02:00:18] depending on the tool. Submit again.

[02:00:21] Let's do that. Generate and send now.

[02:00:24] There we go. Finish. Success. Let's see

[02:00:27] how it looks. If we go to my Gmail and

[02:00:30] now hit a refresh, you can see we just

[02:00:32] received an invoice from me. Hi Alba, I

[02:00:35] hope things are going well at build my

[02:00:36] agent. Please find attached invoice for

[02:00:38] $2,000 dub by payment at terms of 14. So

[02:00:40] let us know if you have any questions,

[02:00:42] need anything from my end. Thanks so

[02:00:43] much for continue trust and working with

[02:00:45] shiny. Really means a lot. And here we

[02:00:47] have an invoice that is generated 100%

[02:00:49] by our workflow. It has all of our

[02:00:51] information invoice number like this and

[02:00:54] this and this. net 14 payment details.

[02:00:57] Boom. If a company now comes to you and

[02:00:59] say, "Well, we spent a lot of time

[02:01:00] sending out invoices." You know how to

[02:01:02] fix it. And we've just successfully

[02:01:04] built a trigger. Workflow without even

[02:01:06] being inside of a workflow builder or

[02:01:09] anything. Claude did it all. But let's

[02:01:11] say we want to deploy this, right? Right

[02:01:12] now, it's running from our computer.

[02:01:14] It's doing all of this from our

[02:01:16] computer. We can then write let's deploy

[02:01:17] this to production on trigger dev. Thank

[02:01:22] you. The way that trigger.dev dev works

[02:01:24] in the cloud is that you have these

[02:01:26] projects right here and inside of here

[02:01:27] you have different environments like

[02:01:28] right now I'm in development but we want

[02:01:30] to be on production so now you can see

[02:01:33] cloud is deploying this it's pushing it

[02:01:34] into the triggerdevcloud but we have to

[02:01:37] do is just set the environmental values

[02:01:39] we do that by going into environmental

[02:01:41] values right here and then clicking add

[02:01:43] new then making sure that we have it on

[02:01:46] production right here a quick tip is

[02:01:48] that you can go inside of environmental

[02:01:50] value copy this then you can go back and

[02:01:53] paste all of this in. We're just going

[02:01:55] to show all of the environmental values

[02:01:56] the right way. We can click save. And

[02:01:59] there we go. All of the environmental

[02:02:01] values has now been added. And there we

[02:02:03] go. Plot has now pushed our task. You

[02:02:05] can see our task inside of here now.

[02:02:07] Send invoice has now been added. We

[02:02:09] don't have any activity yet because we

[02:02:10] just pushed it to production, but we can

[02:02:12] run it again if you want. So, let's test

[02:02:15] it. Before we test it, we need to take

[02:02:16] the API key from inside of production

[02:02:19] cuz that's going to be a different

[02:02:20] secret key. And we need to go in and

[02:02:22] fill out this new information. So the

[02:02:23] trigger secret key, change that to prod.

[02:02:26] Same with the project reference right

[02:02:28] here. That's actually the same. So we

[02:02:30] don't need to do anything there. Let's

[02:02:32] test it now by running npm rundev. I'm

[02:02:34] going to write let's test it in

[02:02:36] localhost 3000 first

[02:02:39] form.

[02:02:40] I change the environmental values. Let's

[02:02:44] try and test it now. So I'm going to

[02:02:45] write elot testing 2. Going to send it

[02:02:48] to my own email once again. I'm going to

[02:02:50] call it build my agent 2 and addresses

[02:02:52] we don't want I'm just going to call it

[02:02:54] testing for let's say 1 1500 USD blah

[02:02:57] blah blah all of this we don't need

[02:02:58] generate and send now you can see it

[02:03:00] says sending cute run if you now go into

[02:03:04] trigger.dev and to our task you can now

[02:03:08] see one run was just cued right here we

[02:03:11] can also see run is now being executed

[02:03:13] that's one of the good things about

[02:03:14] trigger.dev dev as well is that all of

[02:03:16] this dashboard right here where it's

[02:03:18] executing all of the runs, we have a

[02:03:19] very clean dashboard where we can see

[02:03:21] what's going on. And now we can see

[02:03:23] right here the run succeeded in 13

[02:03:25] seconds. If we go to our invoice and

[02:03:28] give it a refresh, we now have a new

[02:03:30] invoice with this new information and

[02:03:32] our invoice generator is now deployed in

[02:03:35] the cloud. So that is the back end,

[02:03:37] right? The backend is now deployed on

[02:03:38] trigger.dev. If we wanted to deploy the

[02:03:41] front end as well, we can also do that.

[02:03:43] We head to Vessel and then we write

[02:03:46] great let's push the front end to

[02:03:50] Vessel.

[02:03:52] First GitHub though we're going to go to

[02:03:55] GitHub and just like we did before we're

[02:03:57] going to create a new repo. I'm going to

[02:03:58] call it invoice front end. We can choose

[02:04:00] an owner. We're going to choose

[02:04:02] ourselves. We can click private and then

[02:04:04] we can create the repository just like

[02:04:06] this. Paste this into claw. Hit enter.

[02:04:09] And now it's going to push our front

[02:04:10] end. Our front end is this page right

[02:04:13] here, right? The invoice generator. It's

[02:04:15] going to push that. So, we have it

[02:04:16] online. The crazy thing about this is

[02:04:18] that you don't have to remember, okay,

[02:04:19] how exactly did Elbert do this. You just

[02:04:21] have to remember, you can use Composio

[02:04:23] for authentication trigger.dev for the

[02:04:26] actual automation. When you tell that to

[02:04:27] Claw, Claude can figure out the rest.

[02:04:29] There we go. It is now pushed. We can

[02:04:31] now, just like we did before, go to

[02:04:33] Visil, click add new project, paste this

[02:04:36] in, click deploy, choose next, and just

[02:04:39] click deploy again. And that will deploy

[02:04:41] our project to the web. There we go. It

[02:04:43] is now deployed. We can click continue

[02:04:45] to dashboard. Click on the domain right

[02:04:47] here. And now you can see the invoice

[02:04:48] generator is live on the web. The last

[02:04:51] thing we need to do is just set the

[02:04:52] environmental values. You do that by

[02:04:54] going in environmental variables right

[02:04:55] here. Click add environmental variables.

[02:04:57] And just take the same one that we used

[02:04:59] before, paste that into the cell like

[02:05:01] this. Click save. You're going to have

[02:05:03] to redeploy. And now the invoice

[02:05:05] generator is live on the web. What's

[02:05:07] important to know is that right now we

[02:05:08] have hooked our enthropic API key up to

[02:05:11] this invoice generator. It generates

[02:05:13] this text right here with AI when it

[02:05:15] sends the email. The issue with that is

[02:05:17] that if someone got access to this

[02:05:19] invoice generator, they can send maybe a

[02:05:21] thousand emails which will burn a lot of

[02:05:23] credits on Claude. So you want to add

[02:05:26] some authentication. You don't want to

[02:05:27] just deploy it like this. And in the

[02:05:28] next part of the course, I'm going to

[02:05:30] show you exactly how you do that because

[02:05:32] you're usually not only going to give

[02:05:33] them one of these like automations.

[02:05:36] Usually when you automate something

[02:05:37] within a business, you're automating

[02:05:39] maybe 5, 10, 20 different processes.

[02:05:42] That adds up to a bunch of time. This is

[02:05:45] just one of them. In the next part of

[02:05:46] the course, I'm going to show you how

[02:05:47] you can add authentication as well. So

[02:05:49] only the actual company can access this.

[02:05:52] And I'm going to show you how you can

[02:05:53] build almost sort of a mini app with a

[02:05:55] bunch of individual small tools like

[02:05:57] this one where all of the automations

[02:05:58] are in one place that you can then give

[02:06:00] to the client which makes you look very,

[02:06:02] very professional. And a collection of

[02:06:03] these is something that you can charge

[02:06:05] 10, 15, $20,000 for if you're doing it

[02:06:07] for big companies. There we go.

[02:06:08] Deployment has been created. If we go to

[02:06:10] it now, you can see it has all of our

[02:06:12] information from before. And now it will

[02:06:14] work from the web. Throughout this

[02:06:15] course, we have built a lot of things

[02:06:17] using skills. For example, the website

[02:06:19] that we created was heavily carried by

[02:06:20] this build premium website skill. This

[02:06:22] is probably one of the most important

[02:06:24] parts about this entire course. And the

[02:06:26] reason for that is that one of the most

[02:06:27] important skills that you want to learn

[02:06:29] if you want to do this oneperson AI plot

[02:06:31] business is to create a process. Spend a

[02:06:34] lot of time building something out like

[02:06:35] this invoicing for example and then

[02:06:37] building a skill so you can cut a lot of

[02:06:40] time if you had to build it again. So

[02:06:41] learning how to build generic skills

[02:06:43] that builds out a process exactly like

[02:06:45] how you want it will save you a bunch of

[02:06:47] time in the long run. And the way you

[02:06:49] want to do this is that you want to use

[02:06:51] my skill that's called create skill.

[02:06:53] This skill is also inside of our drives.

[02:06:55] this one right here called create skill

[02:06:57] which instructs Claude in how to create

[02:06:59] a good skill. I'm going to say nice,

[02:07:03] well done. This was a success.

[02:07:07] I want to

[02:07:09] create a skill for creating simple

[02:07:13] automations like this where I have some

[02:07:17] kind of form

[02:07:20] and that

[02:07:23] is on next.js. That's what we used and

[02:07:27] the back end is on trigger.

[02:07:30] Authentication if possible should be on

[02:07:33] composio. I want you to take what you

[02:07:37] learned from this session and add that

[02:07:40] information into a skill so we can

[02:07:43] replicate it in in the future.

[02:07:47] Very important.

[02:07:49] use the create skill skill that you have

[02:07:53] access to. The user should in the start

[02:07:57] answer a couple of questions about what

[02:08:01] it is that they want to build.

[02:08:04] The skill should be generic for any mini

[02:08:07] automation/tools

[02:08:09] that they want to build. So it shouldn't

[02:08:12] be invoicing

[02:08:15] specifically

[02:08:17] in the skill include that

[02:08:21] claude should research before building.

[02:08:26] Then you want to hit it with an

[02:08:27] ultraink. You want to hit it with plane

[02:08:29] mode. And sometimes something I also

[02:08:31] like to do is to use sub agents to go

[02:08:34] into our current app as the example and

[02:08:39] get all context

[02:08:42] you need. This is something that I like

[02:08:44] to write this use sub agents because

[02:08:46] Claude is then going to speed up the

[02:08:48] process of finding context. Hit enter

[02:08:49] and now Claude is going to build out a

[02:08:51] skill for how to create automations like

[02:08:53] this. This will take a couple of

[02:08:54] minutes. There we go. Now it's asking

[02:08:56] some question like what should the skill

[02:08:58] be named? I think mini automation fits

[02:09:00] pretty well. Should the skill bundle is

[02:09:01] start a template directory? Yes. Where

[02:09:03] should it live? Let's do in the personal

[02:09:06] skills and then click submit answers.

[02:09:09] Because we use plan mode, it's now

[02:09:11] asking us if we want to accept this

[02:09:13] plan. I think it sounds good. So, let's

[02:09:15] accept it. And Claude is now going to

[02:09:17] build up this skill for us. And there we

[02:09:19] go. After a couple of minutes, the skill

[02:09:21] is now built. Something that you always

[02:09:23] want to ask is, is this a standalone

[02:09:26] skill that doesn't reference the invoice

[02:09:30] app? Answer this, please. The reason for

[02:09:32] that is that sometimes Claude likes to

[02:09:34] reference files, but if you want it to

[02:09:37] be its own skill, we don't want it to

[02:09:39] reference this folder. We wanted to

[02:09:41] include all of the information in that

[02:09:42] one skill without needing any other

[02:09:45] context. But you can see it says yes,

[02:09:47] fully standalone, no path references to

[02:09:50] any other skill. Awesome. Now we have

[02:09:52] created a standalone skill. When we

[02:09:54] write Ctrl C and then click plot right

[02:09:57] here and we now write mini automation,

[02:10:01] you can see we have a skill that's

[02:10:02] called / mini automation. Build a mini

[02:10:05] automation in XJS form on the front end,

[02:10:07] a trigger.dev background task on the

[02:10:09] back end with composio for third party

[02:10:11] authentications,

[02:10:13] etc. blah blah blah blah. And now you

[02:10:14] can replicate the automation that we

[02:10:16] just created with that front-end form

[02:10:18] for anything that you want. and you want

[02:10:20] to get a habit out of doing this. So,

[02:10:22] anytime you build something that you

[02:10:23] think, well, I might build this for

[02:10:24] another client in the future, please

[02:10:27] make a skill for it. It's going to save

[02:10:29] you so much time. And especially if

[02:10:31] you're building out websites and you

[02:10:32] want like a couple of different

[02:10:33] templates to choose from, building out a

[02:10:35] custom website and then making it into a

[02:10:37] skill is going to be so valuable for

[02:10:39] you. I personally believe that in the

[02:10:41] future with these Claude businesses, the

[02:10:43] value, the IP is going to be the SOPs

[02:10:46] and processes that Claude knows just

[02:10:48] like in a regular business. Now, what's

[02:10:49] really valuable is all of the documents

[02:10:51] that businesses have that explains how

[02:10:53] they do things. The value in a Claude

[02:10:55] code company like this will be the

[02:10:58] processes, the skills that Claude has

[02:10:59] access to, the information that you have

[02:11:02] given it. So, please make sure to make

[02:11:03] those skills. They're going to become

[02:11:04] valuable and save you a bunch of time in

[02:11:06] the future. Make a habit out of it. It's

[02:11:08] one of the most important parts of

[02:11:09] running a cla code business like this.

[02:11:11] All right. Now we have gone over how you

[02:11:13] build websites for clients, how you also

[02:11:15] build automations and how you can

[02:11:16] include AI into that to build agents.

[02:11:19] The third thing we're going to cover is

[02:11:20] kind of like the next level to that,

[02:11:22] which is how do you then build out full

[02:11:24] AI systems? How do you take a business

[02:11:26] and then automate as many processes that

[02:11:28] you can and then deliver that to the

[02:11:30] client? That's what we're going to cover

[02:11:31] now. A full AI system is basically just

[02:11:33] a bunch of automations combined. So just

[02:11:35] like the invoicing automation that we

[02:11:36] just created, when you provide a full

[02:11:38] system, you usually automate maybe like

[02:11:41] 5 to 10 to 25 depending on depending on

[02:11:44] how big the project is. So how do you go

[02:11:46] about delivering this? This varies from

[02:11:49] agency to agency how you actually do

[02:11:51] this. Some agencies just deliver a bunch

[02:11:54] of init flows where if the client wants

[02:11:57] they can look at the executions of the

[02:11:58] inflows, but the reality is that the

[02:12:00] client is never going to do that. So the

[02:12:02] client just kind of trusts the agency

[02:12:03] that it's set up correctly. But that is

[02:12:05] how most agencies use to do it. When we

[02:12:07] are building with clawed code, we want

[02:12:09] to give Claude access to everything as

[02:12:11] well. So we'll do this a little

[02:12:13] differently. We're going to be building

[02:12:14] a dashboard for the client so they can

[02:12:17] see exactly what's going on, exactly how

[02:12:19] many executions, maybe even exactly how

[02:12:21] much time these automations have saved

[02:12:23] them so far. That also allows us to

[02:12:25] provide them with tools like we just

[02:12:27] created with the invoicing tool where if

[02:12:29] they have some process that requires

[02:12:31] like something to trigger it like a form

[02:12:34] for example like we just showed you can

[02:12:36] put those tools inside of this system as

[02:12:38] well. This also allows us to have a

[02:12:40] front end where they can see all of this

[02:12:42] and then also have a separate backend

[02:12:44] where everything is running and

[02:12:46] depending on the size of the project and

[02:12:48] the client that you're working with.

[02:12:49] Sometimes the client wants to run

[02:12:51] everything on their own servers and this

[02:12:53] system is great for that because the

[02:12:55] front end, the actual visuals and the

[02:12:58] back end is all going to be running as

[02:13:01] code and specifically the back end we

[02:13:03] using trigger.dev exactly like we did in

[02:13:05] the previous automation. And because

[02:13:08] trigger.dev is open source, it means

[02:13:10] that we can host it anywhere. We can

[02:13:11] host it on a virtual machine like a

[02:13:14] machine that's just in the cloud. We can

[02:13:16] host it on the servers that they are

[02:13:17] already running if they're using

[02:13:18] something like Azure, which is the most

[02:13:21] normal one. That is like a cloud

[02:13:23] provider it's called, or if they're

[02:13:25] using AWS, Amazon Web Services, or if

[02:13:28] they're using Google Cloud. Having a

[02:13:30] system like this means that you can

[02:13:32] basically deploy it anywhere you want

[02:13:33] with a very nice user experience. Let me

[02:13:36] be clear from the start, sometimes it's

[02:13:38] overkill building out a full AI system.

[02:13:41] Let's say that you're delivering a

[02:13:42] project for a client and all they need

[02:13:44] is just like one automation done. What

[02:13:46] you could do is literally just provide

[02:13:47] the tool like we just did connected to

[02:13:49] something like a type form where you

[02:13:51] don't have your own front end. This is

[02:13:52] for bigger projects where you are

[02:13:54] auditing the entire business and then

[02:13:55] say okay these things are what we can

[02:13:57] improve and I'm going to build this out

[02:13:59] like this and this and this and you

[02:14:00] basically get your custom dashboard,

[02:14:02] custom app where you have access to all

[02:14:03] of this. All right, let's get into how

[02:14:05] we actually build this. The first thing

[02:14:06] we're going to build out is the

[02:14:08] structure of what is really going to be

[02:14:10] a full app. We're going to have the

[02:14:11] front end and this is going to be very

[02:14:13] similar to what we already built. It's

[02:14:14] going to be next.js as the framework.

[02:14:17] It's going to be tailwind CSS for the

[02:14:19] design. And then I like to use something

[02:14:21] that's called Next O for the

[02:14:23] authentication that works well with

[02:14:25] Nex.js. And then we're going to use

[02:14:26] recent together with Nex off. And resent

[02:14:28] was basically the service that we can

[02:14:30] use to send emails like magic links. And

[02:14:33] this is the most secure and easiest way

[02:14:35] to set up a login. That way you don't

[02:14:36] have to manage passwords. You don't have

[02:14:38] to manage usernames. All you're doing is

[02:14:40] just sending a magic link. Every time

[02:14:42] someone wants to log in, they click on

[02:14:43] that link and they're in. So it's

[02:14:44] basically like forcing a two-factor

[02:14:46] authentication every single time, which

[02:14:48] is the most secure thing that you can

[02:14:49] do. Then we have the backend. This is

[02:14:51] where all of our automations are running

[02:14:53] and we're going to be building that in

[02:14:55] trigger.dev. We're going to start by

[02:14:56] hosting it on trigger.dev. just know

[02:14:59] later on if you wanted to host the

[02:15:00] backend yourself, if you wanted it on

[02:15:02] something like a virtual machine or

[02:15:04] hosting the back end on the client's

[02:15:06] actual servers, then trigger.dev is the

[02:15:08] perfect thing to build with because it

[02:15:09] allows you to just take the code and

[02:15:11] just host it somewhere else as well. And

[02:15:13] then because we're building out kind of

[02:15:15] like a full app, we also need a

[02:15:17] database. But the database I like to

[02:15:19] launch fast and the fastest way that I

[02:15:21] found to launch is by using [ __ ] DB,

[02:15:25] which is a database provider. And the

[02:15:27] database is where we are basically going

[02:15:28] to be storing all of this data. So all

[02:15:31] of the invoices for example if we have

[02:15:33] an invoice generator, all of the clients

[02:15:35] information like the email and user

[02:15:37] information. And what's worth knowing as

[02:15:39] well is that we also going to be hosting

[02:15:40] MongoDB in the cloud just to launch

[02:15:42] fast. If you at one point wanting to

[02:15:44] build this entirely on the client's

[02:15:46] platform, then you can also do that

[02:15:48] because MongoDB also has a community

[02:15:51] edition that allows you to self-host it

[02:15:53] just like you can with trigger.dev. And

[02:15:54] while it's not entirely open source,

[02:15:56] their community edition is fine for what

[02:15:58] you're building as well. So that is kind

[02:15:59] of the stack that we are going to be

[02:16:00] using. And don't worry if you don't know

[02:16:02] what all these things are yet. After

[02:16:03] watching this part of the course, you'll

[02:16:05] know exactly how to set up full AI

[02:16:06] systems like this. Right, let's get into

[02:16:08] it. The first thing we're going to do is

[02:16:09] that I'm going to build a new folder

[02:16:11] right here. I'm going to call this one,

[02:16:15] let's just call it AI system. And then

[02:16:16] I'm going to go out of pl

[02:16:20] AI system, which is basically just going

[02:16:22] to get me into that folder. and then

[02:16:24] write cla. The reason I do this is

[02:16:26] because I don't want to give it the

[02:16:27] context of all the other things that we

[02:16:28] have built as well. We want it to be

[02:16:30] completely focused on this right here.

[02:16:32] What I'll then do is that I'm going to

[02:16:33] write I want to build out a full app

[02:16:40] for a client that includes a bunch of

[02:16:44] small automations. And then I'm going to

[02:16:46] write the only domain

[02:16:49] that should be allowed

[02:16:52] into this app

[02:16:55] should be at.shiny

[02:16:59] AI domains. The reason we do this is

[02:17:01] because if we deliver this for a client,

[02:17:03] we want to add some additional security

[02:17:04] measures that only emails with their own

[02:17:07] domain should be allowed into it. This

[02:17:09] is basically just an additional layer of

[02:17:11] security. Then I'm going to write the

[02:17:12] stack next.js. We want to use next off

[02:17:16] for authentication

[02:17:18] with no Google login because we don't

[02:17:20] need that. We want to be using the magic

[02:17:22] links, but again, this is something that

[02:17:24] you could add later if you wanted to.

[02:17:26] Then we want to use Tailwind CSS recent

[02:17:29] for magic links. For the back end, I'm

[02:17:32] not going to write yet. I'm going to

[02:17:33] write backend. Let's

[02:17:36] not create this yet, but I want to use

[02:17:40] trigger.dev

[02:17:42] at one point. So just create

[02:17:46] a structure for its and then for the

[02:17:49] database we want to use [ __ ] I'm going

[02:17:51] to give it some additional information.

[02:17:54] This app will be a dashboard for the

[02:17:59] automations they have running.

[02:18:02] So keep as much as possible in the front

[02:18:07] end, only automations in the back end,

[02:18:11] only automations and processes in the

[02:18:14] back end. For database, we're going to

[02:18:17] use MongoDB. And then I don't want to

[02:18:19] build anything out yet. I don't want you

[02:18:21] to build out any automations

[02:18:26] yet. I want you to create the structure

[02:18:30] of the app. I want a dashboard when they

[02:18:35] land inside a simple

[02:18:39] sidebar where they will

[02:18:42] have all of the automations and a

[02:18:45] settings tab

[02:18:48] where they can log out etc. Please build

[02:18:52] this out. Then we're going to use

[02:18:54] ultraink. I'm going to put it inside of

[02:18:57] plan mode to force it to create the full

[02:18:59] plan before it starts building. And then

[02:19:01] we are going to hit enter. It's going to

[02:19:03] take a little longer now the first time

[02:19:04] that we are building it out. But just

[02:19:06] like we have done previously with the

[02:19:08] automations, what we'll later do is that

[02:19:09] we're going to create a skill out of

[02:19:11] this. And I'm going to give that skill

[02:19:12] to you as well. And then we can

[02:19:13] basically build out full AI systems like

[02:19:15] this in a single prompt. Which next or

[02:19:17] version? Use the latest one. For

[02:19:19] components, we can use shared CN

[02:19:21] components. How should the shiny AI

[02:19:23] domain restriction work? only allow add

[02:19:25] shiny.ai domains, MongoDB persistence

[02:19:28] for all. Let's use the JWT session

[02:19:31] tokens and submit answers. There we go.

[02:19:33] Now it's going to plan out this entire

[02:19:35] thing. And there we go. Claude has now

[02:19:37] written a plan. I'm also going to tell

[02:19:39] it to please use the super

[02:19:42] powers skill. And I approve of plan.

[02:19:46] Again, if you don't know what the

[02:19:47] superpower skill is, superpowers is

[02:19:50] basically a plugin that includes a

[02:19:51] couple of skills for like writing plans,

[02:19:53] writing skills, executing plans. It's

[02:19:55] basically forcing Claw to think like a

[02:19:58] developer, which is very useful when

[02:20:00] it's building out a full app. So now

[02:20:02] it's going to build out and scaffold

[02:20:04] this entire project. That took 13

[02:20:06] minutes to build out. Now we want to

[02:20:08] click into the folder where we have

[02:20:09] created all of this. We want to go into

[02:20:11] the envo example, copy this, paste this,

[02:20:14] and then rename it and just call

[02:20:16] it.local

[02:20:19] like this. Cool. Now, we need to fill

[02:20:21] out all of this information. The first

[02:20:22] is the O secret. To create this, we can

[02:20:25] use this command right here that it was

[02:20:26] so nice to show, which is called open

[02:20:29] SSL rand and then B 6432.

[02:20:32] Hit enter. That's going to give us a

[02:20:34] random string of letters. We can paste

[02:20:37] that in instead of the O secret right

[02:20:39] here, just like this one. Then we have

[02:20:40] the next O URL. For now, we're just

[02:20:42] going to keep that at localhost 3000.

[02:20:44] And then we go under API keys, create an

[02:20:47] API key. I'm going to call this one AI

[02:20:49] system. It should have full access.

[02:20:51] Click add. Copy it. Go back. Paste it in

[02:20:54] right here. And then I'm just going to

[02:20:55] write shiny portal. That is going to be

[02:20:58] the sender. And for the email, I'm going

[02:21:00] to use this email that I have already

[02:21:02] set up, which is just going to be this

[02:21:03] one. So, it's going to be no reply at

[02:21:05] mail.clipip.dev.

[02:21:07] This is the domain that I'm just going

[02:21:08] to use because I've already set it up

[02:21:09] once. Then, we're using the MongoDB URI.

[02:21:12] We get that by going to MongoDB if you

[02:21:14] don't have one already. Then create a

[02:21:16] user. Create a database and a cluster if

[02:21:19] you don't have one already. I'm now

[02:21:20] going to click connect. Click drivers.

[02:21:22] And this gives me this string right

[02:21:24] here. What we then also want to do is

[02:21:25] that we want to click done right here.

[02:21:27] Then go into database and network

[02:21:29] access. Click edit. And then we can edit

[02:21:31] the password. also generate a secure

[02:21:33] password. Copy this and paste it in

[02:21:36] right here where it says DB password

[02:21:38] just like this. And the trigger.dev key

[02:21:41] is not set up yet. We're going to set

[02:21:42] that up later when we set up the

[02:21:43] automations and the back end. Now we can

[02:21:45] click update user right here. And now it

[02:21:47] should work. The last thing we need to

[02:21:49] do is that we need to insert the name of

[02:21:51] the database. It's not called shiny

[02:21:53] right here. It is called newsify right

[02:21:55] here. It's from a pre previous project.

[02:21:57] So I'm going to fill this out as well.

[02:21:59] Cool. Let's try and run it. So, we need

[02:22:01] to go inside of AI system by using cd

[02:22:05] and AI system. And then we're going to

[02:22:07] write npm rundev to start the dev

[02:22:10] server. Hit enter. This is going to open

[02:22:12] it up. And you can see it starts on

[02:22:14] localhost 3000. Let's see what it build

[02:22:17] out. We definitely need to work on the

[02:22:18] design of this. Let me sign in with

[02:22:20] elot@shiny.ai.

[02:22:22] Send magic link. There you go. Check

[02:22:23] your inbox. Can then go inside of our

[02:22:25] inbox. Elbert shiny.ai. Click sign in.

[02:22:29] And there we go. Now we have our

[02:22:31] dashboard right here that shows total

[02:22:33] automations, how many active, how many

[02:22:35] drafts, recent activity, nothing yet.

[02:22:37] Welcome back. Your automations

[02:22:39] dashboard, blah blah blah. And then it

[02:22:40] says shiny up here. Then we have an

[02:22:42] example automation right here where

[02:22:43] we're going to have all of the

[02:22:44] automations here on the left side. And

[02:22:46] then inside of settings, you can see

[02:22:47] you're signed into shiny email and then

[02:22:50] you can sign out. Great. This is just

[02:22:51] the structure of the dashboard. We

[02:22:53] definitely need to make this look a lot

[02:22:55] better before we do anything else. The

[02:22:57] way we do that is that we use the front

[02:22:59] end design skill. I'm going to write

[02:23:00] that looks good.

[02:23:04] I want to give

[02:23:07] our app a makeover.

[02:23:12] Use the front end design skill. I want

[02:23:15] it in a linear light mode type of vibe.

[02:23:20] We should have both light mode and dark

[02:23:24] mode. Let's build this out. Thanks.

[02:23:28] Ultra think hit it with a plane mode so

[02:23:30] it plans out how we should do it and

[02:23:32] then hit enter. I like to build the

[02:23:34] structure first where it doesn't care at

[02:23:36] all about the design first so it builds

[02:23:38] it out the right way and then later we

[02:23:40] give it a makeover so it doesn't look

[02:23:42] this bad. That will probably take a

[02:23:44] couple of minutes to fix. Great. Now we

[02:23:46] have something that looks more like this

[02:23:47] which looks good. But I can see we

[02:23:49] already have an error right here. So I

[02:23:51] can click on this little icon to copy

[02:23:53] the error. Go back into claw, paste it

[02:23:56] in, and say, "Please fix this.

[02:23:59] Thank you." And now it's going to fix

[02:24:01] this error right here. It's running a

[02:24:04] lint. Go. See if it There we go. The

[02:24:08] error disappeared. If we write at

[02:24:10] elot@shiny.ai, you can see it says check

[02:24:12] your inbox. We can then go to our inbox.

[02:24:16] Click sign in. Now it says something

[02:24:18] went wrong, which is weird. That's

[02:24:19] because I clicked on the one that was 20

[02:24:21] minutes ago. You got to click on the

[02:24:22] newest one right here. 0 minutes ago.

[02:24:25] And there we go. Welcome back. Now it

[02:24:27] looks like this. And the light mode.

[02:24:30] This looks a lot better. There's some

[02:24:31] weird thing going on with the gradient.

[02:24:33] What I'll do is that I'm going to take a

[02:24:35] screenshot of this and say this title

[02:24:39] and other text has a weird gradient.

[02:24:42] That is too much. Please fix it. Thanks.

[02:24:47] I like this theme a lot better. And now

[02:24:49] we have both a light mode and a dark

[02:24:51] mode. It fixed this elbow text right

[02:24:53] here, but it still does that weird

[02:24:55] thing. I'm going to take a screenshot of

[02:24:57] it again and say, see how in the sides

[02:25:01] it's less bright than in the middle. We

[02:25:05] don't want that. Please fix. Also, this

[02:25:08] light mode button up here, we probably

[02:25:09] don't want that. So what I'm going to do

[02:25:12] is that I'm going to say nice the light

[02:25:15] mode button top right corner. Please

[02:25:18] remove that

[02:25:20] so we only have it inside of settings.

[02:25:25] Looks a bit brighter but it still has

[02:25:26] that weird shadow on the left hand side

[02:25:28] right here. Let's try and write on the

[02:25:30] sides. There's a weird shadow. Can we

[02:25:32] please remove that? Actually I think

[02:25:34] it's because these dots in the

[02:25:36] background they probably would need this

[02:25:37] shadow. So, I'm going to write they

[02:25:39] should

[02:25:41] only have the effect on the background

[02:25:44] dots, not on the headlines

[02:25:47] and text and

[02:25:51] components, etc. You can see it's doing

[02:25:53] it on every single page kind of like

[02:25:55] dark to the sides, but then in the

[02:25:56] middle it's like bright. We want it to

[02:25:58] be white everywhere. There we go. Now,

[02:26:00] it fixed it. So, the text is completely

[02:26:02] white. I think this looks a lot better.

[02:26:04] What we're going to do now is we're

[02:26:06] going to install our first app. So I'm

[02:26:08] going to write let's install our first

[02:26:12] app which will be this invoicing app

[02:26:16] that also uses trigger.dev as back end.

[02:26:21] I want you to one install this front end

[02:26:26] form to our AI system folder

[02:26:31] and two create the trigger.dev

[02:26:35] back end. We're going to call this the

[02:26:37] AI system back end right here inside

[02:26:41] this folder that will include the back

[02:26:43] end for this automation.

[02:26:48] also make it so we later can add more

[02:26:53] tasks. That's what it's called inside of

[02:26:56] trigger.dev. Uh so more autoations

[02:26:59] later. For now, let's add this one. I'm

[02:27:03] going to set it to plan mode and write

[02:27:05] use sub aents to get all context you

[02:27:09] need

[02:27:11] and

[02:27:12] implement

[02:27:14] this

[02:27:15] automation. So now we're going to take

[02:27:17] the invoice app that we just built

[02:27:18] previously and we're going to install it

[02:27:20] into our AI system. And there we go. Now

[02:27:23] you can see inside of our app, the send

[02:27:25] invoice automation has now been set up.

[02:27:27] And now they have the same form that's

[02:27:29] now in this new branding where they can

[02:27:30] generate and send an invoice much

[02:27:32] faster. And then when you're building

[02:27:33] out this AI system, it will consist of

[02:27:35] let's say five to 10 different

[02:27:37] automations that are automating

[02:27:38] something. And all the employees would

[02:27:40] need access to is just this dashboard

[02:27:42] right here that would allow them to use

[02:27:44] the automations. The next thing I want

[02:27:46] to do is to handle the authentication.

[02:27:47] If it's not off yet, I want to add a off

[02:27:50] button to this right here. For this

[02:27:52] automation,

[02:27:54] we need to handle the off. So, please

[02:27:59] add if we don't have authentication set

[02:28:03] up when we trigger it

[02:28:08] that a modal pops up where they can

[02:28:12] authenticate. When we created this

[02:28:14] standalone automation, we handle the

[02:28:15] authentication using the link inside of

[02:28:17] the terminal via cloud code. But of

[02:28:19] course, our clients will not be able to

[02:28:21] do that. So, we need to add a nice

[02:28:23] button where they can just authenticate

[02:28:24] with whatever they're using. Cool. I

[02:28:26] hope you see the idea of how you can

[02:28:27] build fully custom AI systems, how you

[02:28:30] can make it domain specific so only

[02:28:32] people from this company can use it, and

[02:28:34] how you now have this structure to build

[02:28:36] anything that you want inside of here.

[02:28:37] Your automations doesn't have to be a

[02:28:39] form where then something happens. You

[02:28:41] can also set up automations that doesn't

[02:28:43] require anyone to do anything where it

[02:28:45] just runs on autopilot. Let me show you

[02:28:46] how to do that, too. But before we do

[02:28:48] that, let's make sure that this actually

[02:28:49] works. So, the first thing I'm going to

[02:28:51] do is that I'm going to go to

[02:28:52] trigger.dev. I'm going to create a new

[02:28:55] project that I'm going to call AI

[02:28:57] system. Then I want to push this

[02:28:59] straight to production. So, I am going

[02:29:01] to get the API key right here, the

[02:29:05] secret key, paste that into

[02:29:06] environmental value right here. And we

[02:29:08] also need to do that right here. Then we

[02:29:10] need to go down and we need to get

[02:29:11] inside of general the project ref. So

[02:29:14] I'm going to insert that here too. Close

[02:29:15] this down like that. Just like before,

[02:29:17] we need to set the environmental

[02:29:18] variable. So I'm going to go inside of

[02:29:20] here, click add new, and then go inside

[02:29:22] of our environmental file. Copy all of

[02:29:24] this and then paste this into the

[02:29:27] production like this and click save. And

[02:29:30] now I'm going to wait for it to set up

[02:29:31] the connection so everything works. And

[02:29:33] then we're going to test it. Cool. Let's

[02:29:34] try and push this to production. Going

[02:29:37] to give it the project ref right here.

[02:29:39] I'm going to say we need to set up the

[02:29:42] back end. I have installed

[02:29:47] all the invigrigger.dev.

[02:29:50] This is the project ref. Can you please

[02:29:53] set this up?

[02:29:56] So, it works with everything

[02:30:00] authentication,

[02:30:02] etc. Thank you. So now we're basically

[02:30:05] just going to push this backend code

[02:30:07] right here, which is the AI system

[02:30:08] backend. We're going to take these tasks

[02:30:11] that are the trigger.dev tasks. We're

[02:30:13] going to push it to this new trigger.dev

[02:30:16] that we just set up. So we have

[02:30:17] everything in one place. You can see

[02:30:19] there's no tasks inside of here yet. And

[02:30:21] there we go. I now deployed it to

[02:30:24] production. Now you can see the task is

[02:30:26] inside of our system on trigger.dev in

[02:30:28] the production. If I then just make a

[02:30:30] test, we can call it Elbert test.

[02:30:33] company name BMA just write consulting

[02:30:35] set it to $1,500

[02:30:38] and click generate and send you can see

[02:30:40] it says there we go cute run ID run like

[02:30:43] this and now if you go inside of runs

[02:30:45] you can see it's now executing and you

[02:30:47] can see invoice from shiny here we have

[02:30:49] it cool so now our send invoice is

[02:30:52] working what I'll do now just like I

[02:30:54] explained before is that now we have the

[02:30:56] start right this is a good starting

[02:30:58] point for every new client that you get

[02:30:59] on as well this overview dashboard where

[02:31:01] you have a company name and one

[02:31:02] automation is the perfect starting point

[02:31:05] for every new client that you sign on as

[02:31:06] well. So, we want to give Claude this

[02:31:08] information. What I'll do is that I'm

[02:31:10] simply just going to create a new

[02:31:11] session. Then I'm going to use the

[02:31:12] create skill and I'm going to drag these

[02:31:14] in

[02:31:16] and I'm going to write I want to create

[02:31:18] a skill that creates this kind of front

[02:31:22] end

[02:31:24] and backend system to when I close a new

[02:31:29] client. It should be a generic skill.

[02:31:33] So first it should ask what the client

[02:31:37] name is, what the domain is

[02:31:42] for their emails, etc. It should gather

[02:31:47] all of the context first. Then I wanted

[02:31:50] to build out this template with the

[02:31:54] front end and back end first. After we

[02:31:58] have built out this structure, we should

[02:32:02] not have any automations yet, we will

[02:32:06] then later build automations,

[02:32:10] but the entire structure should be

[02:32:13] there. So it's very easy to do. This

[02:32:17] skill should work stand alone. Should

[02:32:21] not reference these folders.

[02:32:24] It should have everything included

[02:32:28] with references etc

[02:32:34] inside its own scope. I'm going to give

[02:32:36] it an ultra think and hit enter.

[02:32:38] Actually before I do that I'm going to

[02:32:40] write use sub aents to gather all the

[02:32:43] context for how this works. Thank you.

[02:32:47] I'm also going to write include

[02:32:50] everything

[02:32:52] like design, how it works,

[02:32:56] uh domain, the ST, etc. Use sub agent to

[02:33:01] gather all the context for how this

[02:33:02] works. Thank you. There we go. Bit of a

[02:33:05] long prompt, but we're going to use the

[02:33:07] create skill now. And again, this skill

[02:33:09] is inside the claude skills folder,

[02:33:10] which is inside the free community. So

[02:33:12] to get this skill, just go inside of

[02:33:14] classroom, go inside of learning hub and

[02:33:16] inside of claw skills right here, you

[02:33:18] have all of my claw skills in this drive

[02:33:20] folder and it's completely free to get.

[02:33:22] There we go. Now it's going to start

[02:33:23] building out this and it's calling it

[02:33:24] new client system skill. This will

[02:33:26] probably take some time because it needs

[02:33:27] to gather all of the context first and

[02:33:29] then build the skill. But you really

[02:33:30] want to learn how to do this because

[02:33:32] every time you build something new for a

[02:33:33] client that potentially could be

[02:33:34] replicated in the future, you want to

[02:33:36] build a skill around it. If you don't do

[02:33:37] this, you're going to start from scratch

[02:33:39] every time you land a new client. When

[02:33:40] you do this, you're going to make your

[02:33:42] life much easier. And there we go. Now

[02:33:44] you can see the skill is ready. If I go

[02:33:46] into a new claude session, let's restart

[02:33:49] cla right quick. And we then write new

[02:33:51] client system like this. It says

[02:33:53] scaffold new clients full stack next.js

[02:33:55] 16 front end dashboard trigger.dev bag

[02:33:57] and worker from the shiny automations

[02:33:59] template. Use when the user says new

[02:34:01] client onboarded client scaffold new

[02:34:02] client system, etc. And of course, I'm

[02:34:04] also going to drop this skill together

[02:34:05] with the rest of the skills inside of

[02:34:08] the Google Drive. Now you know how to

[02:34:09] create full agent dashboards and you can

[02:34:11] of course customize this to however you

[02:34:13] want. If you want graphs right here

[02:34:15] showing you the task, you can ask Claude

[02:34:17] to build it. If you want to add a

[02:34:19] specific automation, you can ask Claude

[02:34:21] to build it. If you want to add

[02:34:22] something inside of settings where they

[02:34:24] can add the users themselves, you can

[02:34:25] tell Claude to build it. But what I want

[02:34:27] to do now is build some more automations

[02:34:29] inside of this dashboard to show you how

[02:34:31] that would look as well. Let's get into

[02:34:32] it. All right, let me show you how to

[02:34:34] build out some more automations inside

[02:34:36] of our own automation system. What I'll

[02:34:38] write is I'm going to take this folder

[02:34:41] and this folder right here which is

[02:34:42] basically all the files that this system

[02:34:44] includes and then I'm going to write

[02:34:46] read how this works use sub agents

[02:34:52] and then I'm going to say specifically

[02:34:54] look at how the automations are created

[02:34:57] serum then return to me. We just want to

[02:35:00] give the context for how this AI system

[02:35:02] works before we then try and build out a

[02:35:04] new automation inside of this session.

[02:35:07] So it'll probably take a couple of

[02:35:08] minutes. All right. So let's automate

[02:35:10] some more business processes. Let's say

[02:35:13] the onboarding for example. A lot of

[02:35:15] businesses right now they do manual

[02:35:16] onboarding. So they send out emails

[02:35:18] manually. They send out the contract

[02:35:20] manually. They do all of this stuff

[02:35:21] manually. We can create an automation

[02:35:23] for this as well and display it inside

[02:35:25] of here. So we can now write I want to

[02:35:28] create a new automation that is called

[02:35:32] start onboarding. This should let's

[02:35:34] write up what it should do. Send over

[02:35:37] contract via Gmail using Composure of

[02:35:41] course. Send another email with a

[02:35:44] Calendarly link where they can book an

[02:35:47] onboarding call. Send an email with an

[02:35:51] onboarding

[02:35:52] survey for example. Send another email

[02:35:54] with Kennet link. So the connection we

[02:35:56] need

[02:35:59] is probably just Gmail. The user will

[02:36:02] provide

[02:36:04] onboarding

[02:36:06] link and K link because it already now

[02:36:09] knows the structure of our system. This

[02:36:11] should be extremely easy to set up. I'm

[02:36:14] also going to write please add this task

[02:36:16] to trigger.dev of course like the other

[02:36:21] ones straight to prod. And there we go.

[02:36:25] Now it's going to build out this

[02:36:26] automation just like the other one.

[02:36:28] There we go. That was extremely fast. We

[02:36:30] can actually see that building out this

[02:36:32] automation actually took only 3 minutes.

[02:36:34] And it was also a pretty simple

[02:36:35] automation, but we're going to start

[02:36:37] here before we build something a little

[02:36:38] more advanced. But let's say that we

[02:36:40] just closed me. This is my email. This

[02:36:44] is my company. Then we can say send a

[02:36:47] contract URL. This could be something

[02:36:49] like a docu sign. But we might want to

[02:36:51] change this to an actual PDF. So maybe

[02:36:53] let's do that. Let's copy this and say,

[02:36:55] can we instead here

[02:36:58] upload a PDF

[02:37:01] on the new tool.

[02:37:03] Thank you. Then it's going to change the

[02:37:04] automation to do that. Then we insert

[02:37:06] the onboarding survey and then the

[02:37:07] calendarly URL. So the calendarly URL

[02:37:09] could be something like this. The

[02:37:11] onboarding survey URL be something like

[02:37:13] this. And then we need the contract URL,

[02:37:15] but that's going to change. Now we can

[02:37:17] instead upload a PDF, but I actually

[02:37:19] want to generate the PDF instead. Can we

[02:37:21] actually instead generate the PDF? For

[02:37:24] now, we just generate a dummy contract.

[02:37:30] But when we generate, we insert the

[02:37:33] actual name and business information,

[02:37:37] etc. Please update. And again, if the

[02:37:39] client comes back and says, well, I want

[02:37:41] this change, so can we do this instead?

[02:37:43] Then you have everything in code. B can

[02:37:45] literally change anything inside of

[02:37:46] here. So, it's fully customizable. There

[02:37:48] we go. So now we don't have a field to

[02:37:50] upload the contract, but now it's

[02:37:52] basically just going to send out a dummy

[02:37:53] one. So let's test if it works. How

[02:37:55] about a shiny

[02:37:58] AI BMA? Insert our onboarding survey.

[02:38:02] Insert our calendarly form.

[02:38:05] Send this in. Click start onboarding.

[02:38:07] There we go. You can see it cued this

[02:38:09] run right here. If we go into

[02:38:11] trigger.dev and look at runs, we can now

[02:38:14] see that it's now executing this run.

[02:38:16] There we go. It took 12 seconds. Let's

[02:38:19] see if I received something in my email.

[02:38:20] Now I get the book your onboarding call,

[02:38:22] onboarding survey, and your contract.

[02:38:25] Let's get started. This is all

[02:38:27] automatically sent. You can see we even

[02:38:29] generated this draft or placeholder

[02:38:31] contract that we can always change later

[02:38:32] to exactly what we want. But now it

[02:38:34] sends out this contract as well that

[02:38:36] includes my information. Absolutely

[02:38:38] insane. We just automated the onboarding

[02:38:40] process as well. And if the client says,

[02:38:42] "Well, we want a new lead updated inside

[02:38:44] of our CRM or we want something else

[02:38:45] done." Literally, you just tell Claw to

[02:38:47] build it and it's going to build it out

[02:38:49] for you. But let's build out some more

[02:38:51] advanced automations. Let's get into it.

[02:38:54] All right. Now, you know the basics of

[02:38:56] creating these automations. But let's

[02:38:58] say that we want to build something a

[02:39:00] little more impressive. And this is

[02:39:01] usually also the things that you'll be

[02:39:03] selling to clients. For example, let's

[02:39:05] imagine that a client that you're

[02:39:07] working with, they have five support

[02:39:09] reps that are answering emails all day

[02:39:11] long. And you want to automate, let's

[02:39:13] say our goal is 60% of that. So all of

[02:39:16] the questions that are the ones that are

[02:39:18] asked all the time that an AI can

[02:39:20] literally just answer those questions

[02:39:22] almost instantly. We can also build a

[02:39:24] full AI agent system like this inside of

[02:39:27] our dashboards. The great thing about

[02:39:29] this is that it is custom code. It means

[02:39:32] that there are literally no boundaries.

[02:39:34] We can build whatever we want. So I'm

[02:39:36] going to go back into claw code. I'm

[02:39:38] again going to tell it to read the

[02:39:42] structure of this full AI system. Then

[02:39:46] return back to me. Use sub agents.

[02:39:51] Understand how it works. Hit enter. It

[02:39:54] will gather all of the context that it

[02:39:55] needs and then it's going to come right

[02:39:57] back. Great. It returns back after 2

[02:39:59] minutes and has all the context. Then we

[02:40:01] can write I want to build another

[02:40:05] more advanced automation. This should be

[02:40:09] a full AI support email ticket response

[02:40:14] system that can respond to around 60% of

[02:40:18] all emails. In this

[02:40:23] automation, I want the front end to show

[02:40:27] the AI responses. I want it to show if

[02:40:31] any has been escalated. I want to be

[02:40:34] able to upload a full knowledge base.

[02:40:38] So, it should use rag to search through

[02:40:41] it

[02:40:43] and answer questions based on this

[02:40:46] knowledge base. It's getting a little

[02:40:48] more advanced here, but luckily we do

[02:40:49] have claw that can build out all of this

[02:40:51] for us. We will use composio to answer

[02:40:55] the emails uh using Gmail. I want to

[02:41:00] trigger the trigger.dev every 10 minutes

[02:41:04] to check for new emails and then respond

[02:41:08] to every single one that has come in. If

[02:41:14] the ticket requires

[02:41:17] human attention, we should be able to

[02:41:21] escalate it

[02:41:23] and see the escalated emails in the

[02:41:26] front end. and be able

[02:41:30] to respond to them from within the front

[02:41:34] end as well. I also want to be able to

[02:41:38] see the AI responses inside the front

[02:41:41] end, too. Let's see if there's anything

[02:41:43] else that we need. Let's plan this

[02:41:46] out. Please use the superpower skill and

[02:41:51] ask me if you have any questions. Let's

[02:41:55] fire away that prompt. Then you can see

[02:41:56] it's going to load the superpower skill

[02:41:58] and ask us any questions if it needs

[02:42:00] more context. We can all of a sudden

[02:42:01] start building some really cool advanced

[02:42:03] stuff based on the current structure

[02:42:05] that we have. How should the AI reply

[02:42:07] behave for the 60% it's confident about

[02:42:09] it should also send immediately. Of

[02:42:11] course, if you're building this out for

[02:42:12] a client, you would of course have

[02:42:14] talked with the client and asked how it

[02:42:15] should do it. But you're going to want

[02:42:17] to send immediately. What format will

[02:42:18] the knowledge base come in? Let's do

[02:42:21] let's do PDF files. Let's actually just

[02:42:23] do text files. That is a lot easier.

[02:42:26] Well, we can actually turn on all of

[02:42:27] them to be the answers. Where should we

[02:42:29] store the vector embeddings for rag?

[02:42:32] We could use something like pine cone or

[02:42:34] postgress, but we actually already have

[02:42:36] vector search inside of MongoDB. So,

[02:42:38] let's use that. Let's do if any human

[02:42:41] asks for like low confidence from AI and

[02:42:44] tone and refund. And let's also do

[02:42:46] specific clarifiers like this. Submit.

[02:42:49] You can see it's asking all of the right

[02:42:50] questions that it needs in order to

[02:42:52] build out this plan. Which email account

[02:42:54] does this monitor reply from? One shared

[02:42:56] inbox. Yes. And there we go. Now you can

[02:42:58] see it's exploring the project to get

[02:43:01] context. And then it's going to ask a

[02:43:02] couple of clarifying questions.

[02:43:05] Come with a couple of approaches and

[02:43:07] then design the plan for building this

[02:43:09] out. How should we identify a ticket?

[02:43:11] What groups emails together? A Gmail

[02:43:14] thread is one ticket. That sounds good.

[02:43:16] When you escalate and reply from the

[02:43:17] front end, should the AI help draft the

[02:43:19] human reply? Yep. If we can get AI to

[02:43:21] pre-draft the reply but just not send. I

[02:43:23] think that could be pretty cool. Where

[02:43:24] do we store tickets, messages, KB chunks

[02:43:26] and embeddings? Reuse existing MongoDB.

[02:43:29] Let's do that. Anything important I

[02:43:31] should add to the spec. Pick anything

[02:43:32] that apply. Per ticket tags, labels,

[02:43:36] analytics view. Could do an analytics

[02:43:38] view and a per ticket tag. Submit this.

[02:43:41] Submit the answers. Got it. And off to

[02:43:43] design. Let me present it in sections.

[02:43:45] Confirm after each section one. This is

[02:43:48] this. That sounds good. Then it's going

[02:43:49] to use trigger.dev to pull every single

[02:43:52] 10 minutes. This also looks good. There

[02:43:54] we go. That was all of the sections. Now

[02:43:55] it's saying writing out the spec ducks.

[02:43:57] Now that's probably going to take a

[02:43:59] couple of minutes. Now we have the full

[02:44:01] design spec, which looks good. So I'm

[02:44:04] going to tell it to continue. Now it's

[02:44:06] going to use this design spec to build

[02:44:07] out the entire plan so it can build this

[02:44:09] out. It's going to ask us if we want to

[02:44:11] use sub aent driven or inline execution.

[02:44:13] We definitely want to use sub agent to

[02:44:16] speed it up. This means that different

[02:44:17] agents are going to be working on

[02:44:19] different tasks. You can see I dispatch

[02:44:21] a fresh sub agent per task review

[02:44:23] between task fast iteration usually the

[02:44:26] fastest and also what's recommended in

[02:44:28] this superpower skill. So now it's going

[02:44:30] to use the sub aent driven development

[02:44:33] loaded this superpower skill and now

[02:44:34] it's going to build out this entire

[02:44:36] thing for us based on our instructions.

[02:44:39] This will probably take a bit to build

[02:44:40] up because it is a pretty extensive AI

[02:44:42] system that we are building out. There

[02:44:44] we go. The next thing we need to do is

[02:44:45] that we need to go in and get our

[02:44:47] MongoDB secret. This one right here.

[02:44:50] Then we need to go to our production

[02:44:52] trigger.dev. Go inside of environmental

[02:44:54] variables. Click add new. Click on

[02:44:56] production. Make sure that we have added

[02:44:58] both the MongoDB URI and the DB name.

[02:45:01] Click save. There we go. Now I'm going

[02:45:04] to write we need this to update on

[02:45:07] prodrigger.dev.

[02:45:10] So it will work.

[02:45:13] Can we please do that? Hit enter. And

[02:45:15] now it's going to push the changes it

[02:45:17] has made to our back end. It's going to

[02:45:19] push those trigger.dev so everything

[02:45:21] will run smoothly. All right, there we

[02:45:23] go. If I now spin up the server, you can

[02:45:25] see on localhost 3000, we now have

[02:45:27] another automation that's called AI

[02:45:29] support. And here we have a couple of

[02:45:31] tabs for escalated support request, AI

[02:45:33] replied, human applied, and all. We also

[02:45:36] have this analytics tab right here that

[02:45:38] shows us how many tickets we have. And

[02:45:40] we have this knowledge base tab where we

[02:45:42] can either paste text into the knowledge

[02:45:44] base. We can add PDFs or URLs. Let's see

[02:45:46] if this actually works. I can take our

[02:45:49] shiny.ai website for example. Paste this

[02:45:51] in. Click add to knowledge base. It is

[02:45:54] processing now. Let's see if it actually

[02:45:56] works. Says error though. So this didn't

[02:45:59] work for some reason. Let's try and see

[02:46:01] if we can fix this. I tried to add

[02:46:04] knowledge base

[02:46:07] but ran into this issue. Can we check

[02:46:10] what happened? In the meantime, let's

[02:46:12] try and check the other things like

[02:46:14] paste text. If I just like copy this

[02:46:16] entire page and paste it in right here,

[02:46:18] add to knowledge space and call it basic

[02:46:22] info.

[02:46:24] Let's see if it wants to add this. But

[02:46:26] then give it a refresh again. Still says

[02:46:28] failed. So there's obviously something

[02:46:29] wrong with the knowledge base, but

[02:46:31] should of course be able to fix this.

[02:46:33] Ah, now I see why it fails. We need an

[02:46:35] openi API key. So I'm going to set this

[02:46:38] value as well inside our environmental

[02:46:40] value right here at the end. Openi API

[02:46:43] key and then I'm going to go to

[02:46:45] platformi.com

[02:46:48] login with my Google account. Let's just

[02:46:50] call it AI system. Create the secret

[02:46:52] key. Copy it. Paste it in right here.

[02:46:55] And then we also need to go into

[02:46:56] trigger.dev inside of the environmental

[02:46:58] v variables here. Click add new and add

[02:47:01] it here too. Let's do a refresh of this

[02:47:03] right here. Go to support. We're

[02:47:05] actually getting emails now that is

[02:47:07] escalated to us. No emails has been

[02:47:09] replied to yet. Let's try and add the

[02:47:10] knowledge base once again. Copy our

[02:47:12] entire website right here. Paste it in.

[02:47:14] Call it basic info. Add to knowledge

[02:47:17] base. Now it is processing. Refresh.

[02:47:19] Still processing now. So it didn't fail

[02:47:21] instantly. That's good. Let's delete

[02:47:22] these others while we are at it. And

[02:47:24] there we go. Now it says ready. So if we

[02:47:26] go back to inbox, we can see it has no

[02:47:28] subject line right now. So I'm going to

[02:47:29] copy this and I'm going to paste this in

[02:47:32] and say that seems to work. issue now is

[02:47:35] that the subject lines don't show even

[02:47:39] though we do have subject lines.

[02:47:42] Can you fix please hit enter? And then

[02:47:45] hopefully we can fix this subie line not

[02:47:48] showing but this isn't actually email

[02:47:49] from my inbox which is kind of cool.

[02:47:51] Same with this one. This is also an

[02:47:52] actual email. So it's actually fetching

[02:47:54] the emails which is kind of cool. And it

[02:47:56] is escalating them. Great. I think I

[02:47:58] also want to add a refresh button

[02:48:01] somewhere on this page. So, I'm going to

[02:48:03] give it this page right here and say,

[02:48:05] can we add a refresh button here as well

[02:48:09] that triggers

[02:48:11] the polling, please? That way, we can

[02:48:13] also do it manually just to get all the

[02:48:15] newest ones. And there we have it. Now,

[02:48:17] we have a little poll inbox button right

[02:48:19] here. If we pull it, it says refreshing

[02:48:22] just like this. And then it shows us if

[02:48:24] we have any new emails in our inbox. So,

[02:48:27] let's actually try it. Let me go on this

[02:48:29] email account right here. Write to my

[02:48:32] personal email right here. Basically,

[02:48:33] write what does shiny AI do? We can

[02:48:36] basically just ask, "Hi, saw your

[02:48:40] website. What is it that Shiny AI does?"

[02:48:44] We can click send. And now I receive my

[02:48:46] email inside of my inbox. Let's see if

[02:48:48] we also receive it or if we hit little

[02:48:50] polling right here. For some reason, it

[02:48:52] doesn't show. Maybe it's because it's

[02:48:54] under AI replied. H

[02:48:57] check the latest poll.

[02:49:01] I ran it after sending an email to

[02:49:05] myself to check it, but

[02:49:09] the email doesn't show up.

[02:49:13] Why is that? Check the latest logs.

[02:49:15] Let's see if I can figure out what

[02:49:17] happened. There we go. The issue was

[02:49:18] apparently that I clicked in on it, so

[02:49:21] it became red. Of course, we only want

[02:49:23] to take the unread emails. So yeah, the

[02:49:25] system is actually working like it's

[02:49:26] intended. Cool. But it's still being

[02:49:28] escalated. So we need to fix some

[02:49:30] prompting. I think I can write, see this

[02:49:34] email.

[02:49:35] It's a simple question and we have it in

[02:49:39] our knowledge base.

[02:49:43] Why did it get escalated?

[02:49:47] Hit enter and see what it says. All

[02:49:49] right. This was another one of these

[02:49:50] issue where it wasn't really an issue. I

[02:49:51] just had to refresh. But now I've added

[02:49:53] auto refreshes every 30 seconds inside

[02:49:56] of this tab. And if we go to AI replied

[02:49:58] and we click in on this, we can actually

[02:50:00] see, hey, so your website blah blah

[02:50:02] blah. And then the AI actually replied

[02:50:03] with this. Hi, thanks for reaching out.

[02:50:05] China AI helps businesses boo sales.

[02:50:07] It's using the information that it got

[02:50:09] from the website. And if we click why

[02:50:11] this reply, we can even see the

[02:50:13] confidence score of how smart it thinks

[02:50:15] it was. We can see the citations that it

[02:50:17] used from our knowledge base. This is

[02:50:21] extremely cool and it works. So now we

[02:50:23] basically have an AI inside of our

[02:50:24] support email that can reply to

[02:50:26] questions. When this confidence score

[02:50:28] isn't as high, it's going to put it in

[02:50:30] the escalator tab so we can reply to it.

[02:50:32] I hope you start to see the power of

[02:50:33] these insane automations and agent

[02:50:36] systems that you can all build inside of

[02:50:38] your own kind of dashboard right here

[02:50:40] and provide to a client. And the great

[02:50:42] thing is that as you build these systems

[02:50:44] out, you could then sell this support

[02:50:46] inbox to another client and it wouldn't

[02:50:48] really take you that much work to set it

[02:50:50] up with a new knowledge base on a new

[02:50:52] client. And the crazy thing is the

[02:50:54] amount that you could sell this for. If

[02:50:56] you sold this to a big company, you

[02:50:58] could probably charge anywhere from $2

[02:51:00] to $5,000 a month depending on how big

[02:51:03] the team is. Cuz think about it, right

[02:51:05] now their support rep is probably being

[02:51:06] paid maybe like $2,000 a month. If it's

[02:51:09] in the US, then probably three to $4,000

[02:51:11] a month. So if a team has, let's say, 10

[02:51:14] support reps and all of a sudden they

[02:51:15] only need five, you're saving them five

[02:51:18] times, let's say $3,000, that's $15,000

[02:51:22] a month you're saving them. And if

[02:51:24] you're charging $5,000, you're still

[02:51:26] saving them $10,000 a month. And this is

[02:51:28] just one of the automation systems that

[02:51:30] you can build out. You can automate all

[02:51:32] processes inside of a business. Smack

[02:51:34] all of these automations here on the

[02:51:35] lefth hand side and you have a killer

[02:51:38] product. And you can manage all of this

[02:51:40] with cloud code. What I'll do is that I

[02:51:41] will take the code for this entire

[02:51:43] dashboard right here with the trigger.

[02:51:45] Back end. So these folders right here,

[02:51:47] AI system and AI system back end. I'm

[02:51:49] also going to take the cityscape

[02:51:51] website. So you have that as well as

[02:51:52] reference. And I'm going to put the

[02:51:53] GitHub repos right below this video. The

[02:51:55] last thing I want to do is that I want

[02:51:57] to host the front end. Right now we are

[02:51:59] still in localhost 3000. The back end is

[02:52:02] hosted on the cloud.trigger.dev,

[02:52:04] but the front end isn't hosted yet. And

[02:52:07] it's honestly quite easy to do. It's the

[02:52:09] exact same thing we did with the other

[02:52:10] website. We just open visil.com and

[02:52:13] click add new project. Then we want to

[02:52:15] push this to GitHub. So I'm going to say

[02:52:17] I want to push only the front end. So

[02:52:20] the only this AI system folder right

[02:52:22] here. That's the front end, not the back

[02:52:24] end, only the front end to GitHub.

[02:52:29] Then we're going to open a new tab.

[02:52:30] We're going to go to github.com. We're

[02:52:32] going to create a new repo. I'm going to

[02:52:34] call it AI system front end. We're going

[02:52:38] to make this private. Create repository.

[02:52:41] Then we can copy the link right here. We

[02:52:43] can paste it in and say push to this

[02:52:46] repo. You won't be able to do this in

[02:52:48] auto mode. So I'm going to set it to

[02:52:50] just like the default mode. Hit enter.

[02:52:52] And then we will have to accept the

[02:52:53] commands. It's asking us what to push.

[02:52:55] I'm going to ask it. I want to push this

[02:52:57] to separate repo so I can host the front

[02:53:01] end. So only the AI system standard push

[02:53:04] it's it's empty brand new and then we

[02:53:06] push it. Something that I always like to

[02:53:08] do is that I write of course don't push

[02:53:11] any secrets

[02:53:14] etc. Just so we don't push any of these

[02:53:16] secrets to GitHub for security. Always

[02:53:18] good to remind Claude not to do that.

[02:53:20] Sometimes it gets ahead of itself. It is

[02:53:22] running all of these commands. We'll

[02:53:24] just let it do that. There we go.

[02:53:25] Created a little mini plan for itself.

[02:53:27] Now it's going to push this frontend

[02:53:29] repo to GitHub. There we go. Now it is

[02:53:31] pushed. If we go onto GitHub, then hit

[02:53:34] enter. Now you can see we have all of

[02:53:36] the files inside of here. Let's host

[02:53:38] that by going to VIL, copying this

[02:53:41] GitHub repo, pasting it in, clicking

[02:53:44] deploy. It's going to see that it's a

[02:53:46] NextJS application right here. Click

[02:53:48] deploy again. And let's see if we have

[02:53:50] any build errors. See, we already have a

[02:53:53] build error. So what I'll do is that I'm

[02:53:55] going to copy this. better thing to do

[02:53:56] actually is to go inside of here where

[02:53:59] we are running it and then instead of

[02:54:01] npm rundev we can write npm run build

[02:54:04] which is going to show us all of the

[02:54:06] build errors inside of this folder and

[02:54:09] see it compile successfully running

[02:54:10] typescript this actually built

[02:54:12] successfully so I'm not sure what this

[02:54:14] error is so let's copy this paste it in

[02:54:17] and say when I try and host on this I

[02:54:22] get this

[02:54:24] why the front end installed MongoDB7 but

[02:54:27] pian install both vessel doesn't pass

[02:54:30] legacies by default to clean fixes are

[02:54:32] dark mango 6. Let's do what it

[02:54:34] recommends. There we go. It pushed the

[02:54:36] changes to GitHub. So let's try and

[02:54:39] deploy it again. We can just click

[02:54:41] deploy once more. And now we should not

[02:54:43] get this error again. You can see it

[02:54:44] passed the first 10 seconds which was

[02:54:46] where we had an error before. And there

[02:54:48] we go. Congratulations. You now deployed

[02:54:50] a project. Let's continue to dashboard.

[02:54:53] we get this domain right here that we

[02:54:55] can go to and you can see this will now

[02:54:57] be the login screen right here which

[02:54:58] will only allow us to sign in with a

[02:55:00] shiny domain. So I'm going to write my

[02:55:02] email, click send magic link and we see

[02:55:04] something went wrong which is because we

[02:55:05] don't have our environmental var

[02:55:08] variables yet. So go inside of here and

[02:55:10] then we need to fill out all of the

[02:55:12] environmental variables. Take add

[02:55:14] pasting in all of these. Click save and

[02:55:17] do a little redeploy like this. This

[02:55:19] will take another minute or so and then

[02:55:21] it will deploy it. Let's click right

[02:55:24] here view deployment. Look at logs right

[02:55:26] here. Build logs to see if it deploys

[02:55:29] like it should. There we go. Build

[02:55:31] completed. If we then give it a refresh

[02:55:33] and write in our email. Now it should

[02:55:36] work. I can see right here that it's

[02:55:37] going to localhost and that's because I

[02:55:40] forgot to change one of the

[02:55:41] environmental variables. So let's go to

[02:55:44] next or UL right here. You can see we're

[02:55:46] going to be using this UL instead.

[02:55:47] Pasted this in. Click save and click

[02:55:49] redeploy. Right now we're using this

[02:55:51] domain right here which is the domain

[02:55:53] that VIL gives us. But you probably want

[02:55:55] to add your own domain when you set up

[02:55:56] an AI system like this. The way you do

[02:55:58] that is that you just go inside of VIL

[02:56:00] again, go down to domains and then just

[02:56:02] click add existing and you can write in

[02:56:04] whatever domain you want and connect it

[02:56:06] using DNS. If you have any issues with

[02:56:08] that then you can just ask claude. After

[02:56:10] you've changed the domain, then it's

[02:56:12] very important that you go inside of the

[02:56:14] environmental variables just like I

[02:56:15] forgot to do now and change this next

[02:56:18] off URL to be the URL where your app is

[02:56:20] running. If you don't do that, you're

[02:56:22] going to run into the same issue that we

[02:56:23] just did before. Now it's deployed once

[02:56:25] again. Let's hit enter. Write in elot

[02:56:29] shiny.ai. Now it is sending the magic of

[02:56:32] URL. And now we are in and we have our

[02:56:34] AI support. We have our AI send invoice.

[02:56:37] We have our start onboarding. We have

[02:56:38] all of the processes that we built just

[02:56:40] before. There we go. Now you've deployed

[02:56:42] an AI system that could be for a client.

[02:56:45] Congratulations for reaching this far

[02:56:47] into the course. Great. Now you know how

[02:56:49] to build in these three levels. Websites

[02:56:51] as the easiest thing to build, then

[02:56:53] individual automations to automate stuff

[02:56:55] with trigger.dev and then how you can

[02:56:58] build a full app and dashboard around it

[02:57:00] that has multiple different automations

[02:57:02] in order to deliver a full project. What

[02:57:05] I'll show you next is my structure and

[02:57:07] my way of basically being able to build

[02:57:10] anything. Because now you know a couple

[02:57:12] of things you can build and deliver to a

[02:57:13] client. But if a client has some

[02:57:15] specific request and you want to build

[02:57:17] something that you don't have a tutorial

[02:57:19] on that doesn't exist on YouTube, how do

[02:57:21] you go about that? How do you build

[02:57:23] anything? I usually break it down into

[02:57:25] four different steps. The first thing

[02:57:28] before I even start building anything,

[02:57:29] we need to create a really, really good

[02:57:32] build plan. This includes things like

[02:57:35] the tick stack that we are using, how it

[02:57:37] will work, what database, what

[02:57:39] frameworks, what exactly is it that we

[02:57:42] are building. It also includes things

[02:57:44] like the design, how should it look,

[02:57:46] what should the feel of the thing we are

[02:57:48] building, what should that be like? And

[02:57:49] that is actually the most important step

[02:57:51] of building anything. And the reason for

[02:57:54] that is that there are so many things

[02:57:55] that goes into this step of figuring out

[02:57:58] okay what are we building and how do we

[02:58:00] build it specifically with the tech

[02:58:02] stack because if you're building

[02:58:04] something like an app or an automation

[02:58:06] the text stack is what determines what

[02:58:08] the cost of running it will be and how

[02:58:10] fast it's going to be as well and how

[02:58:12] well does it scale and if we at one

[02:58:15] point later on wants to add more things

[02:58:18] is that possible can we add more

[02:58:20] features can actually do the things that

[02:58:22] we want Can we give it the functionality

[02:58:24] that we need and want in our app or

[02:58:27] automation or whatever it is that we are

[02:58:29] building? This build plan, what most

[02:58:31] beginners do when they're building

[02:58:32] something is that they just write, I

[02:58:34] want to build this and this and this,

[02:58:35] hit enter, and then take it from there.

[02:58:37] But what happens most of the time is

[02:58:39] that they end up with a bad text stack.

[02:58:42] They don't go back and forth with the

[02:58:44] large language model in order to

[02:58:46] actually get the best text stack

[02:58:47] possible for what they're trying to do.

[02:58:49] and they end up launching something that

[02:58:51] is mediocre that makes them run into

[02:58:53] issues later down the line when it comes

[02:58:55] to scaling the cost, how fast there is

[02:58:57] functionality, etc. So before we start

[02:59:00] building anything, we want to have the

[02:59:02] text stack dialed, we want to keep going

[02:59:04] back and forth with Claude until we get

[02:59:06] the best tech stack possible. And then

[02:59:09] we also want to nail the design so we

[02:59:11] know exactly what the feel should be of

[02:59:13] the user experience before we even start

[02:59:15] building anything. It's much harder down

[02:59:17] the line to change something from bad to

[02:59:19] good rather than just building it from

[02:59:22] good in the start. After we have done

[02:59:23] that, after we have spent a lot of time,

[02:59:25] and I'm literally talking hours here, if

[02:59:27] you're building something complex,

[02:59:29] figuring out the text stack, figuring

[02:59:30] out the design, and making sure that all

[02:59:32] of these will be good, now it's time to

[02:59:34] build out the implementation plan. What

[02:59:36] people that are a bit more intermediate,

[02:59:38] but still beginners, what they do is

[02:59:40] that they maybe just make one plan of,

[02:59:42] okay, this is what I want to build, give

[02:59:43] it to Claude, and then have that build

[02:59:45] it off. And photo will actually do this

[02:59:46] next step itself which is building out

[02:59:48] an implementation plan. But you're going

[02:59:50] to get much better results if you

[02:59:52] actually verify that you have a good

[02:59:54] implementation plan. An implementation

[02:59:55] plan is basically a plan for how do you

[02:59:58] actually build the thing that you have

[03:00:00] written out in the build plan. And the

[03:00:02] way I like to do them is split them up

[03:00:03] in steps. And for each step I want a

[03:00:06] bunch of different checkboxes. So to

[03:00:09] give you an example, step one might be

[03:00:11] to set up file structure. And in the

[03:00:13] file structure, there's a bunch of

[03:00:14] different substeps in setting up the

[03:00:16] file structure and you want to write all

[03:00:18] of this out in your implementation plan.

[03:00:20] And if you're building something very

[03:00:21] complex, you might have something like

[03:00:24] 40 steps in your implementation plan in

[03:00:26] order to build the thing out that you

[03:00:28] want. And before you even start building

[03:00:29] anything, you want to make sure that you

[03:00:31] have this implementation plan dialed. In

[03:00:33] step two, the next thing we might want

[03:00:35] to build out is maybe the database

[03:00:37] structure and then maybe authentication.

[03:00:39] Claude knows the best sequence of how to

[03:00:42] build out anything, but you want to

[03:00:44] force it to make a good implementation

[03:00:46] plan. You want to force the AI to really

[03:00:48] make a decision and make a good decision

[03:00:50] for how to build out the thing that you

[03:00:52] want. And then after you've made the

[03:00:53] build plan and you've made a good

[03:00:54] implementation plan, that is when you

[03:00:57] start building. And the way you do that

[03:00:59] now is that you give the large language

[03:01:00] model this implementation plan of

[03:01:02] course. And then you tell it, please

[03:01:04] start with step one. And the large

[03:01:06] language model is now going to start

[03:01:07] with this substep, then do this substep,

[03:01:10] and then do this substep and come back

[03:01:11] to you and say, well, step one has now

[03:01:14] been complete. Do you want me to start

[03:01:15] with step two? And then it's going to

[03:01:17] start step two. It's going to do the

[03:01:19] substeps inside of step two, and then

[03:01:21] it's going to finish. This is much

[03:01:23] better than just giving it an entire

[03:01:25] plan or just giving a build plan and

[03:01:27] letting it control everything itself. By

[03:01:29] this, you force it to do it in the best

[03:01:31] sequence possible. And you have planned

[03:01:33] everything out before you even start

[03:01:34] building anything. I hope you start to

[03:01:36] see the power of building out both a

[03:01:37] build plan and an implementation plan.

[03:01:39] And actually, when it comes to like the

[03:01:41] time that you're building, step one and

[03:01:44] two will be where you spend 80% of your

[03:01:47] time. Step three when building something

[03:01:49] out. And actually building it will only

[03:01:51] take around 20% of the time. A lot of

[03:01:54] people in the AI space get this wrong

[03:01:56] and they spend 80% of the time just

[03:01:57] letting the LLM build and only 20% of

[03:02:00] the time planning it out, if they even

[03:02:01] do plan it out at all. So when you made

[03:02:03] a build plan, you made an implementation

[03:02:05] plan and the AI has gone through every

[03:02:06] single step all the way down to step 40

[03:02:09] or however long the implementation plan

[03:02:11] is and the AI is not done building it,

[03:02:13] then it's finished and you'll have to

[03:02:14] test and refine it a lot. You'll

[03:02:17] definitely run into bugs. It almost

[03:02:18] always happens and you basically just go

[03:02:20] through them, paste the bug to plot and

[03:02:22] have it finish everything. That is

[03:02:24] basically how you go about building

[03:02:26] something. In all of this, you want plot

[03:02:27] to be almost a consultant that can

[03:02:29] explain to you what is the best text

[03:02:31] stack. What should I use for building

[03:02:32] this? What gives me the best results?

[03:02:34] And a bonus tip that gives you even

[03:02:36] better results is if you use two large

[03:02:38] language models. It could either be two

[03:02:41] claw code sessions, could also be one

[03:02:43] claw and then one codeex. But if you

[03:02:47] have one AI as kind of like the

[03:02:49] consultant that decides what to use and

[03:02:51] then you take that decision and you give

[03:02:54] it to another AI model and you say, "Is

[03:02:56] this true?" They're then going to bounce

[03:02:57] ideas and make sure you get the best

[03:02:59] result possible. And by doing that, you

[03:03:01] can start with absolutely no technical

[03:03:02] ability and you can figure everything

[03:03:04] out on the fly. Awesome. Let me actually

[03:03:06] demonstrate exactly how you do this by

[03:03:08] building something that I have no idea

[03:03:10] how to build. and I'll show you exactly

[03:03:12] how you use this four-step framework

[03:03:14] right here to do the exact thing that we

[03:03:16] just talked about. Let's get into it. To

[03:03:18] show you that this actually works and to

[03:03:20] show you how to actually do it, let's

[03:03:22] build something that I haven't built

[03:03:24] before just to show you that it actually

[03:03:25] works and that you can build something

[03:03:27] really, really good without knowing how

[03:03:28] it works. First, I want to build a lead

[03:03:31] enricher. What that basically means is

[03:03:33] that I want to be able to insert a name

[03:03:36] or an email or a LinkedIn URL and from

[03:03:40] that I want to search up a database of

[03:03:42] some kind of sort and from that I want a

[03:03:45] software that can go out and get all

[03:03:47] information on that person. So if I

[03:03:48] wanted to find someone's email, I could

[03:03:51] just paste in their LinkedIn URL and

[03:03:53] that would give them their email. I have

[03:03:54] a slight idea of how to build this. I

[03:03:56] can probably use some API from like

[03:03:58] Apollo or something, but I wanted to

[03:04:00] show you how you can know none of that

[03:04:03] and get all the answers from Claude in

[03:04:05] order to build it out. So, I'm not going

[03:04:06] to help it and tell it what to do.

[03:04:08] Claude is going to make the decisions

[03:04:09] for us. I have a blank project right

[03:04:12] here. I'm just going to head under the

[03:04:13] terminal and click new terminal and then

[03:04:15] open up a Claude session right here and

[03:04:17] click yes, I trust the author. Then

[03:04:20] going to drag it on top right here. And

[03:04:22] I'm going to write I want to build a

[03:04:25] lead in richer.

[03:04:28] First I want to build a

[03:04:32] build plan. Ask me some questions about

[03:04:35] how it should work. So we can create a

[03:04:41] build plan as amp file. We just set it

[03:04:45] to auto mode and hit enter. And now Lord

[03:04:48] will ask us a couple of questions to

[03:04:50] basically build the thing that we want

[03:04:52] to build. And you can actually see it's

[03:04:53] even using the superpower skill which

[03:04:55] kind of forces this entire process. This

[03:04:57] superpower skill first creates the build

[03:05:00] plan of what we want to build which is

[03:05:01] basically what it calls a design spec

[03:05:03] and then it actually also builds out an

[03:05:04] implementation plan. This is what's

[03:05:06] going on under the hook for the

[03:05:08] superpower scope. You can see it says

[03:05:09] what's the input to the lead enricher?

[03:05:11] What do you start with? I want to be

[03:05:15] able to start with anything

[03:05:18] and from that.

[03:05:21] So either email

[03:05:25] company domain

[03:05:28] name goes company

[03:05:31] linked in URL and from that get all of

[03:05:36] the information. So, I should

[03:05:40] be able to

[03:05:42] insert what I have. Hit enter. What

[03:05:46] output fields do you want the enricher

[03:05:47] to return for each lead?

[03:05:50] Let's do person basics and contact info

[03:05:53] and company info. And then, yeah, let's

[03:05:56] get socials as well. Submit.

[03:05:58] Hit enter. Which enrichment data source

[03:06:01] do you want to use? This is where we

[03:06:03] want to make sure that we actually make

[03:06:04] a good decision. Choosing an enricher is

[03:06:06] the biggest cost of running this. And

[03:06:08] right here, it's giving an option to

[03:06:10] have multiple and use each as fallback.

[03:06:12] I like that idea. So, let's do that.

[03:06:14] What do you want to interact with in

[03:06:16] Richer? Let's do a small web app like

[03:06:18] this. How should leads be input into the

[03:06:20] web app? Both single bulk one form from

[03:06:23] a single lead. Let's just do single lead

[03:06:27] only for now. Do you want to store

[03:06:29] enriched leads or is it lookup only?

[03:06:31] Store history. And if you reach a

[03:06:33] question where you don't know the

[03:06:34] answer, you can literally just scroll

[03:06:36] down. If you're using the superpowers

[03:06:37] plugin, you can also just go out of it

[03:06:39] and then discuss it with Claude. But we

[03:06:41] can write chat about this and then say,

[03:06:44] what would you

[03:06:47] recommend? And then it says, I'd

[03:06:49] recommend store history in a database.

[03:06:51] Reasons saving on cost blah blah blah.

[03:06:54] The only reason to skip storage is if

[03:06:55] you have a strict privacy policy

[03:06:56] compliance reason. Let me just say for

[03:06:58] now that we don't need to store it. Now

[03:07:00] you can see it says got it. Look up

[03:07:01] only. How should the provider fall back

[03:07:04] chain work when enriching a lead

[03:07:06] sequential with merge? Try Apollo first

[03:07:08] then Hunter. Let's try and just say

[03:07:10] let's just try Hunter first.

[03:07:15] And I want it to stop when it finds the

[03:07:18] email

[03:07:20] because that is the main thing. If

[03:07:22] there's something that you don't

[03:07:23] understand, something that you don't

[03:07:25] have an answer to that you think is

[03:07:27] best, you should just ask claw what it

[03:07:29] would recommend. Here you can see it

[03:07:31] says what order should the providers be

[03:07:34] tried in? Gives us a couple of options.

[03:07:36] I'm just going to say hunter only. But

[03:07:38] you can l just go back and forth with

[03:07:39] claude like this. What's the text stack?

[03:07:42] Should I match the shiny automations

[03:07:44] template? It knows this from memory. Um,

[03:07:47] let's do next year only and no

[03:07:50] trigger.dev for now. I have enough to

[03:07:51] draft the design. Here's the proposal,

[03:07:53] architecture, blah blah blah. That's all

[03:07:56] of this. And now it's creating what's

[03:07:57] called the design spec. And the design

[03:07:59] spec is basically what I called the

[03:08:01] build plan in my previous explanation of

[03:08:04] how to build anything. You can see it

[03:08:05] has two questions, two things I like to

[03:08:07] decide before I write the spec. How

[03:08:09] should I handle the LinkedIn in URL

[03:08:10] case? Let's for now skip

[03:08:14] URL and let's do a drop down. The skill

[03:08:17] I'm using right here, which is also very

[03:08:19] valuable for you to know about, is

[03:08:21] called superpowers. And what superpowers

[03:08:23] basically does is that it forces Claw to

[03:08:25] think like an engineer. All you have to

[03:08:26] do is literally just search GitHub

[03:08:28] superpowers. You're going to find this

[03:08:30] GitHub repo right here. It's called Ora

[03:08:32] Superpowers. Click on this and you can

[03:08:33] see it has 28,000

[03:08:36] stars. All you do is that you literally

[03:08:38] just copy this right here and you go to

[03:08:41] claw and you say, "Please install these

[03:08:45] skills. search up the documentation for

[03:08:48] installing global claw skills before you

[03:08:52] do it. It should just send this like

[03:08:55] this and it's going to install the

[03:08:56] skills for you. After you've done that,

[03:08:58] you might have to restart claw just like

[03:09:00] this before it's ready. But now you can

[03:09:03] see spec written and saved to ducks

[03:09:05] superpowers and then it has the spec

[03:09:07] right here and that is sign. There we

[03:09:09] go. So if we click this, you can now see

[03:09:11] it built out a full plan for us. And we

[03:09:14] want to read this carefully to make sure

[03:09:15] that aligns with exactly what we want to

[03:09:17] build. So you can see it says lead

[03:09:18] enrich the sign spec. It shows the date.

[03:09:20] It shows the status which is draft a

[03:09:22] goal a personal web tool for enriching a

[03:09:24] single lead at a time. The user enters

[03:09:26] whatever they have email company name or

[03:09:27] domain name and domain name and company.

[03:09:30] And the app returns the lead's contact

[03:09:32] information company info and social

[03:09:34] presence. There we go. Non goals. We

[03:09:36] don't need a CSV enrichment. This is not

[03:09:38] what we want. It shows the stack right

[03:09:40] here. It shows the user flow, shows the

[03:09:43] architecture, shows the component

[03:09:45] responsibilities, what each page should

[03:09:47] do. You can basically see we have a full

[03:09:49] plan already for exactly what it is we

[03:09:51] want to build. If we find something

[03:09:52] inside of here that we don't like, we

[03:09:55] can solo claw to fix it in the

[03:09:56] implementation plan. It even shows how

[03:09:58] the enrich lead should be structured. It

[03:10:01] shows error handling. It shows file

[03:10:03] layout and it shows out of the scope

[03:10:05] maybe for the future like this. Nice.

[03:10:07] Now we have our full build plan and

[03:10:09] right now we're building a fairly simple

[03:10:11] tool. It's just like a personal tool. If

[03:10:13] you were to build out a full software

[03:10:16] for example, then this would definitely

[03:10:17] be an extremely long build plan. You

[03:10:19] would have to include pricing and

[03:10:20] authentication and databases and maybe

[03:10:23] you're running some virtual machine or

[03:10:26] like hosting the back end. All of this

[03:10:27] would be inside of your bill plan or

[03:10:29] what superpowers call it the design

[03:10:31] spec. You can see it says spec written

[03:10:33] and saved. I'll say I have read it now.

[03:10:37] Please continue. Now you can see it's

[03:10:39] using a different skill also coming from

[03:10:41] superpowers that's called writing plans.

[03:10:43] And this is where we go to the next step

[03:10:45] which is building out the implementation

[03:10:47] plan with these substeps. And building

[03:10:49] out this implementation plan is actually

[03:10:51] the thing that usually takes the

[03:10:52] longest. I'll let it do that and then

[03:10:54] I'll come right back. And there we go.

[03:10:56] Now you can see the plan is complete and

[03:10:58] saved inside of plans. And this is what

[03:11:01] the implementation plan is. Right? If we

[03:11:04] click on plans and I'll click on this,

[03:11:06] you can see we have the entire file

[03:11:08] structure. We have task one, do this and

[03:11:10] this and this and this. Set all of these

[03:11:12] things. Step five, step six. So it's

[03:11:16] basically split up in tasks and then

[03:11:18] steps inside of these tasks exactly like

[03:11:22] this. So now we have two options. to be

[03:11:24] one sub aent driven where each thing are

[03:11:26] built where a fresh sub agent is doing

[03:11:29] each task. This speeds it up a lot or do

[03:11:31] we want to do everything in line. I'm

[03:11:34] going to do one sub agents start the

[03:11:37] build and only now after we've made a

[03:11:39] spec and after we've made a plan is when

[03:11:42] we start building it out. So now it's

[03:11:44] going to build out this little mini app

[03:11:45] for us which is a lead envir. And there

[03:11:48] we have it. Now Claude has built out

[03:11:50] this entire thing by following the

[03:11:52] implementation plan and this took 13

[03:11:55] minutes. The next thing we want to do is

[03:11:57] testing and refining and we are probably

[03:11:59] going to run into some bugs. You can see

[03:12:01] we have our environmental values right

[03:12:03] here. Let me copy that, paste it and

[03:12:05] then rename it to just local. And let's

[03:12:08] then update this Hunter API key. Go to

[03:12:11] hun.io and go to API. Copy this API key.

[03:12:15] Go back. Paste it in right here. Close

[03:12:18] it down. And let's try and run the

[03:12:20] server. So we can now write npm rundev.

[03:12:22] That's going to give us localhost 3000.

[03:12:24] Paste it in. And let's see if our app

[03:12:26] works. So it says lead enricher right

[03:12:28] here. What do you have? Let's say we

[03:12:30] have a name and company. And the name we

[03:12:34] can write this. Let's test it on myself

[03:12:37] first. Company name shiny.ai. Let's see

[03:12:40] if we can find it using this. Click

[03:12:43] enrich. Now we run into this issue right

[03:12:45] here. So I'm going to copy this. This

[03:12:47] was as expected. Paste it in and say we

[03:12:50] run into this issue. Hit enter and then

[03:12:52] we should be able to fix it. It was

[03:12:54] something to do with how it's passing

[03:12:55] the information. So let's see if it

[03:12:57] works now. So you can see it's making a

[03:12:59] bunch of updates now in the code. Now it

[03:13:02] wants to test it. We can just stop it

[03:13:03] here. Let's try and restart the server.

[03:13:05] Write name and company email again.

[03:13:08] Click enrich. And there we go. Now we

[03:13:10] have a bunch of information. And it

[03:13:12] could even just from my name and the

[03:13:13] company name, it could find my email.

[03:13:15] Let's see if I just write shiny.ai and

[03:13:17] click enrich. See what data it then

[03:13:19] finds. Then you can see it actually

[03:13:20] finds Oliver, my co-founder, and his

[03:13:23] email. And there we go. The lead

[03:13:24] enricher works. It's not a very

[03:13:25] impressive app because it literally just

[03:13:27] uses the Hunter API. But it was also

[03:13:29] show you this flow right here of

[03:13:31] building something that you've never

[03:13:32] built before by using these four steps.

[03:13:35] Creating the build plan, creating the

[03:13:36] implementation plan, and then having the

[03:13:38] LLM build for you, but spending a lot of

[03:13:40] time on the build plan and the

[03:13:41] implementation plan. And then when you

[03:13:43] have all of that when you have sharpened

[03:13:45] your axe, that's when you start building

[03:13:47] it. And I guess we can do the next step

[03:13:49] which is testing it and refining it. We

[03:13:51] have already done a bit of bit of

[03:13:52] testing, but let's refine it as well. I

[03:13:54] can give it the front end skill now and

[03:13:56] say make it look clean and modern like a

[03:14:01] modern SAS.

[03:14:04] Hit enter. And now it's basically just

[03:14:05] going to give our entire software a full

[03:14:07] makeover of how it looks. That's a part

[03:14:09] of the refining step, actually making it

[03:14:12] look good. There we go. Now we have

[03:14:14] given it a bit of a makeover. Now, find

[03:14:17] anyone's work email. And now we can

[03:14:18] choose either email, domain, name,

[03:14:20] company. It works exactly like it did

[03:14:21] before, but now it just looks a little

[03:14:24] better. Awesome. The reason I wanted to

[03:14:26] build this out is just to showcase how

[03:14:28] you basically build anything. So, when a

[03:14:30] client comes to you and tells you, I

[03:14:32] want to build out this and this and

[03:14:33] this. I want to try and automate this.

[03:14:35] Is this possible? You can literally just

[03:14:37] ask Claude about how would I build this?

[03:14:39] How would I go about this? Use this

[03:14:41] framework right here for first creating

[03:14:43] a build plan, creating the

[03:14:44] implementation R and then having it

[03:14:46] built out. And that basically allows you

[03:14:47] to build anything you really want or

[03:14:49] would never need. If you're already

[03:14:51] technical, then you might be able to

[03:14:52] skip some of these stages. Like you

[03:14:54] might already say that I want to use

[03:14:55] this and this and this text stack

[03:14:56] because you already know that this text

[03:14:58] stack would be good for the task. But if

[03:15:00] you're a beginner and you're just

[03:15:01] starting out and you don't have that

[03:15:03] much technical expertise, don't skip the

[03:15:05] step of just chatting with Claude and

[03:15:07] finding out what the best tech stack

[03:15:09] would be for things like scaling, cost,

[03:15:11] how fast it is, the functionality, etc.

[03:15:14] Go back and forth with Claude a lot

[03:15:16] about that and you're going to get the

[03:15:17] best results. In the next part of the

[03:15:18] course, we're going to talk about

[03:15:19] pricing because now you know how to

[03:15:21] create a bunch of stuff. But how do you

[03:15:23] actually price it? How much can you

[03:15:24] charge? What pricing model should you

[03:15:26] use? That is what we're going to cover

[03:15:27] now. Now you know how to build a bunch

[03:15:30] of different things and you also know

[03:15:31] the framework for basically being able

[03:15:33] to build anything using clawed code. But

[03:15:36] how do you actually price your services?

[03:15:39] How much should you be charging? This is

[03:15:42] a question I get a lot. And I always use

[03:15:44] the golden rule that when it comes to

[03:15:47] developing an AI project, you always

[03:15:50] want to give your clients a 5x return on

[03:15:54] investment. So, let's say that you save

[03:15:57] your client $5,000 in payroll or in

[03:16:00] whatever solution that you implemented.

[03:16:02] You would be able to charge a $1,000

[03:16:06] from that for your service. If you keep

[03:16:08] this 5x return on investment, your

[03:16:11] client will always be happy. And a happy

[03:16:13] client means low churn, which basically

[03:16:16] means that they stay with you for a very

[03:16:18] long time. If they're paying you

[03:16:19] monthly, you of course want them to stay

[03:16:21] as long as possible. And it increases

[03:16:23] referrals. If you did a good job for

[03:16:25] them, then they're going to refer you to

[03:16:27] their friends that are also business

[03:16:29] owners that might also have the issue

[03:16:30] that you're solving. And this is where

[03:16:32] you potentially can achieve negative

[03:16:34] churn, which basically means that your

[03:16:36] agency grows without you doing anything.

[03:16:39] Just from serving your clients right now

[03:16:42] without paying anything else in

[03:16:43] marketing, your agency would still

[03:16:46] increase because the churn is so low and

[03:16:48] clients are referring you to other

[03:16:50] clients in a higher rate than the churn

[03:16:53] you have. And that all starts by giving

[03:16:54] your clients a good return on

[03:16:56] investment. That's the entire point of

[03:16:58] this business model. And in any service

[03:17:00] business really, you need to go in and

[03:17:02] do a hell of a job. Give them a return

[03:17:04] on investment in order to actually

[03:17:05] deserve the $1,000. But how do you

[03:17:08] actually calculate the return on

[03:17:09] investment for your services? It is

[03:17:12] basic math. Let's just use an example

[03:17:14] that you are installing voice agents and

[03:17:17] right now you're working with a client

[03:17:19] that right now has 10 people in customer

[03:17:22] support and they are taking calls all

[03:17:23] day. They're answering the same

[03:17:24] questions over and over again. So now

[03:17:27] you installed your voice agent system

[03:17:29] and this now means that they go from 10

[03:17:31] customer support reps to let's say five

[03:17:34] customer support reps. This means that

[03:17:35] they've cut five customer support reps.

[03:17:38] Each rep might be paid something like

[03:17:40] $2,000 a month. So when you have cut

[03:17:43] from 10 down to five, that is five reps,

[03:17:46] that basically means that you're now

[03:17:47] saving this company $10,000 a month. So

[03:17:53] using the 5x return on investment rule,

[03:17:56] that mean that from that, in order to

[03:17:58] keep the client very happy, you'll be

[03:18:00] able to charge $2,000

[03:18:02] a month for that. And if you do it like

[03:18:04] this, then they have no reason to churn.

[03:18:06] You're literally saving them still

[03:18:08] $8,000 every month just from using your

[03:18:10] system. Of course, you need to do some

[03:18:12] work. You need to maintain it. You need

[03:18:13] to make sure that voice agents actually

[03:18:14] do a good job, but that is also what

[03:18:17] you're paid to do. That is how you can

[03:18:18] calculate a return on investment. You

[03:18:20] might also have a system that directly

[03:18:23] generates revenue. Let's say you have

[03:18:25] like a cold email AI system that you

[03:18:29] install for clients and that system is

[03:18:32] generating a business, let's say

[03:18:35] $100,000 a month in new revenue. When

[03:18:37] you do this, you need to calculate in

[03:18:39] gross profit because this is a revenue.

[03:18:42] Gross profit is what you have left after

[03:18:44] you remove the fixed cost. So things

[03:18:45] like payroll for taking care of the

[03:18:47] client. How much is the company actually

[03:18:49] profiting from that $10,000 worth of

[03:18:52] revenue? That might be, let's say, if

[03:18:53] they have good margins, like $40,000. So

[03:18:56] that is in theory what you're making

[03:18:58] them every single month. You can charge

[03:19:00] one of that, which means if you're

[03:19:01] running this system right here, you'll

[03:19:04] be paid $8,000 a month. So, you always

[03:19:07] want to give them at least a 5x return

[03:19:09] of their money. And of course, this

[03:19:12] $100,000 a month in revenue that you're

[03:19:14] giving them needs to be new revenue,

[03:19:16] right? Not their existing revenue

[03:19:18] already. So, if you can by installing

[03:19:20] your system, give them a 5x return on

[03:19:22] investment, then they're going to be

[03:19:23] very happy and they're going to stay

[03:19:24] with you forever. But what pricing

[03:19:26] structure do you then use? If you want

[03:19:29] to make it really easy and if you want

[03:19:31] to follow the the golden rule when it's

[03:19:34] available, when result based, when it

[03:19:36] makes sense, it's a very very good

[03:19:38] offer. If you literally just say to them

[03:19:40] upfront that you're going to be taking

[03:19:43] 20% of the gross profit that you make a

[03:19:46] company, that is such a no-brainer

[03:19:49] because there's no risk in it for them.

[03:19:51] You remove all of the risk. Now there's

[03:19:53] only an upside and it also shows that

[03:19:55] you're very confident in what you're

[03:19:57] doing. The negatives are of course is

[03:19:59] that you need to be able to track all of

[03:20:01] this. So this usually works with larger

[03:20:04] clients where you can actually go in and

[03:20:06] control and see okay how much money am I

[03:20:08] making them that you can follow each

[03:20:10] lead in the process that they have a

[03:20:12] good CRM etc. That's one of the really

[03:20:14] good offers resultbased and when you are

[03:20:16] good at what you do resultbased will

[03:20:18] also usually pay you more than something

[03:20:21] like a regular retainer. Let me give you

[03:20:23] an example. Shiny.ai AI, which is our AI

[03:20:25] agency, does lead reactivations, and we

[03:20:28] always use result-based. We had one

[03:20:31] client where we probably could have

[03:20:32] charged maybe like $1,000 a month on a

[03:20:35] retainer, but because we did it as

[03:20:38] resultbased because we went after the

[03:20:40] golden rule. That client in total

[03:20:42] actually made us around $240,000

[03:20:46] in LTV. So, we probably at least 10x the

[03:20:50] lifetime value just from doing it result

[03:20:52] based. Resultbased is one of those

[03:20:54] offers that you would rather have a

[03:20:55] retainer if you're bad at what you do.

[03:20:57] But if you're really really good at what

[03:20:59] you do, you would always rather have

[03:21:00] resultbased because you're paid for

[03:21:02] every result that you bring. So if you

[03:21:04] can scale that up, your result also

[03:21:05] scales infinitely. Well, that kind of

[03:21:07] makes sense. But this offer right here

[03:21:09] is not always the best one to use

[03:21:11] because sometimes it could be hard to

[03:21:13] calculate how much something is worth.

[03:21:15] Let's say that you are implementing

[03:21:17] customer support widgets on websites,

[03:21:19] right? to this little website widget in

[03:21:21] the bottom right corner. It's very hard

[03:21:23] to say how much a conversation is worth.

[03:21:26] You could still do result based. It will

[03:21:28] probably not be of gross profit.

[03:21:29] Instead, it might be per conversation,

[03:21:32] for example, that you're charging them a

[03:21:36] little something. Maybe you're charging

[03:21:37] them something like $1 per conversation.

[03:21:39] So, this still definitely still add up

[03:21:41] if they have like 200, 300 conversations

[03:21:43] a month. But that's one way that you can

[03:21:44] do these kind of like support agents as

[03:21:47] result based. But you cannot really like

[03:21:49] calculate gross profit directly from

[03:21:51] like a support agent. So sometimes

[03:21:54] resultbased really doesn't make sense.

[03:21:56] Instead, what I recommend at that point

[03:21:58] is the upfront offer plus recurring

[03:22:01] because sometimes it's simply like

[03:22:02] impossible to track and it also becomes

[03:22:05] a headache if you have like 50 clients

[03:22:07] to track the performance every single

[03:22:09] month. So charging something up front

[03:22:10] and then a recurring afterwards is

[03:22:12] sometimes the best option. But if you do

[03:22:14] that, I'll always recommend that you use

[03:22:16] the satisfaction guarantee. This has

[03:22:18] worked so well for us. Let's say that we

[03:22:20] want to charge a client $2,000 upfront

[03:22:23] and then let's say $500 a month. If we

[03:22:26] just give them this offer right here,

[03:22:28] there's a bunch of risk, right? The

[03:22:29] client is taking a risk. Okay, does this

[03:22:30] even work? And they're paying a pretty

[03:22:33] big amount right up front without even

[03:22:35] seeing if it's going to work yet.

[03:22:37] Instead, what I want you to do is to

[03:22:39] give a satisfaction guarantee. So, you

[03:22:42] still charge $2,000 up front. You say,

[03:22:44] "I'm going to build this out. It's going

[03:22:46] to cost $2,000 as a setup fee, but I'm

[03:22:49] so confident in my own ability that I

[03:22:52] can build this really well out for you."

[03:22:54] So, after I've built out the entire

[03:22:56] system, so if you're building a

[03:22:57] dashboard, you will build out

[03:22:58] everything. If you're if you're building

[03:23:00] a website, you'll build out everything.

[03:23:02] You're basically going to say that I'm

[03:23:03] going to build out everything so you can

[03:23:04] see exactly how it looks. then we're

[03:23:06] going to hop on a meeting and if you

[03:23:08] don't like the result at that moment

[03:23:11] then you can get a full refund. So then

[03:23:13] we have the satisfaction guarantee. The

[03:23:15] reason why this is so strong is because

[03:23:18] now all of a sudden there's no risk in

[03:23:20] it for them. If they don't like what you

[03:23:21] have built, then they can just get their

[03:23:22] money refunded right afterwards. But if

[03:23:24] you're actually good at what you're

[03:23:26] doing, which you hopefully are, then

[03:23:28] people are not going to refund because

[03:23:29] they're going to see a website that they

[03:23:30] actually like or they're going to see a

[03:23:32] dashboard that actually automates the

[03:23:33] things that they want automated. It's

[03:23:35] very important sometimes for some

[03:23:36] projects that you get some money up

[03:23:38] front. The reason for that is that it

[03:23:40] has happened countless of times for us

[03:23:43] that we have started the project without

[03:23:46] taking any money up front and then the

[03:23:48] client simply like stopped the

[03:23:49] answering. Maybe he got cold feet, maybe

[03:23:53] something else came up, maybe he was

[03:23:55] just too busy. But the reason they could

[03:23:57] just like stop answering and don't care

[03:23:58] anymore is because they were not

[03:24:00] invested. You want people to be invested

[03:24:02] right from the start when you work with

[03:24:03] them. So, this is like the perfect

[03:24:05] middle ground. You still get them

[03:24:06] invested, you still get money up front,

[03:24:08] but you remove the risk by giving them a

[03:24:10] satisfaction guarantee. And if you're

[03:24:11] good at what you're actually doing, we

[03:24:13] have given this offer a bunch of times,

[03:24:15] and we have never had to give a refund.

[03:24:17] Because every single time we have shown

[03:24:19] someone a website, for example, or shown

[03:24:21] someone a system that we have built out

[03:24:22] for them, they have said, "This looks

[03:24:24] good. I'm ready to try it. Awesome.

[03:24:26] Let's get started." And for a lot of the

[03:24:28] time, they actually forget about the

[03:24:29] satisfaction guarantee when they see

[03:24:31] something that's really good. And then

[03:24:32] I'm a big fan of recurring businesses. I

[03:24:34] don't like to work on a project basis. I

[03:24:36] like having recurring revenue coming in

[03:24:38] every single month. So when you charge

[03:24:41] an upfront fee and then charge a lower

[03:24:43] recurring fee, this lower recurring fee

[03:24:45] is usually not a problem. And this

[03:24:47] really helps when you start to stack

[03:24:48] like 10 plus clients, all of a sudden

[03:24:51] this recurring revenue is just stacking

[03:24:52] and compounding. What you basically

[03:24:54] justify is that this is to keep up

[03:24:56] support. It's to keep the system active

[03:24:58] to fix anything if any issues come up.

[03:25:00] That is what this fee basically like a

[03:25:02] service fee covers, right? Which is a

[03:25:04] monthly recurring fee. Another thing

[03:25:05] that's very important to talk about is

[03:25:07] that you always of course want to make

[03:25:08] sure that you profit. You don't want to

[03:25:10] build out a full system for them and

[03:25:12] then cover, let's say, the software cost

[03:25:14] yourself and then realize that you that

[03:25:16] you're not going to profit. A way to

[03:25:18] overcome that is to use tiers just like

[03:25:21] you would with the software where you

[03:25:23] have like tier one, tier two, tier three

[03:25:25] that becomes more and more expensive.

[03:25:27] You can do that as well with your AI

[03:25:30] services. The key here being that you

[03:25:32] always want to make sure that you

[03:25:33] profit. Let's say that you're building

[03:25:35] out the voice agents to use the example

[03:25:36] from before. Then you could have, let's

[03:25:39] say, a 100 calls on plan one. Let's say

[03:25:42] a,000 calls on plan two and let's say

[03:25:45] 5,000 plans on plan three and then have

[03:25:48] maybe like a fourth plan that's like

[03:25:50] 5,000 plus, which is maybe per

[03:25:52] conversation or something that you can

[03:25:54] charge for. By doing that, by having

[03:25:56] different tiers, you always make sure

[03:25:58] that you are in the green. If you're

[03:26:00] just charging like 500 bucks a month and

[03:26:03] all of a sudden you realize that this

[03:26:04] company is massive and they're going to

[03:26:05] have like 10,000 calls a month, you

[03:26:08] might get ruined. So, you want to make

[03:26:10] sure that you're dialed in on this and

[03:26:11] that you are always in the profit. A

[03:26:14] question I get a lot as well is, should

[03:26:16] I include software in my price? Like,

[03:26:19] let's say that you're building out the

[03:26:20] voice voice agents as an example. it

[03:26:22] costs maybe like $100 a month for you to

[03:26:24] run it. Should you tell that to the

[03:26:26] client and get them to pay $100 extra or

[03:26:29] should you include it in your price? So,

[03:26:30] if before if you're charging 500 bucks,

[03:26:32] then it would just charge 600 bucks

[03:26:34] without telling them. And from my

[03:26:36] experience, what I've seen is that a

[03:26:38] confused person doesn't buy. If someone

[03:26:41] is confused about what they're going to

[03:26:43] pay, what they're going to get, they're

[03:26:44] not going to purchase anything. So, you

[03:26:46] want to make your offer as simple as

[03:26:48] possible. Instead of saying, well, it's

[03:26:50] like $500 a month for the service and

[03:26:53] then $100 a month for the actual

[03:26:55] software, we almost always include it

[03:26:57] all in our price and just have a very

[03:26:59] simple offer that is easy to understand.

[03:27:02] There's one scenario, however, where we

[03:27:03] don't do this, but where we actually do

[03:27:05] this, and this is if the client wants to

[03:27:07] like own the system completely or like

[03:27:10] run it on their own servers. This is

[03:27:12] usually like bigger clients that wants

[03:27:14] to run everything on their own servers.

[03:27:15] The price tag for that is of course

[03:27:16] going to be much higher as we now need

[03:27:18] to install it on their servers. And when

[03:27:20] we're doing that, they are of course

[03:27:21] also paying for like the hosting and the

[03:27:23] software cost. So I'd say when you're

[03:27:25] just starting out like 99% of the time

[03:27:27] you would just include the cost inside

[03:27:29] of your price. When you start going up

[03:27:31] market and working with larger clients,

[03:27:33] that's sometimes when you have to sell

[03:27:36] them. Well, it's going to cost this and

[03:27:37] this on your own servers and you

[03:27:39] basically take the price out of your

[03:27:41] service fee. I mentioned it a bit

[03:27:43] earlier, but please, this is just like a

[03:27:46] reminder. You never say the price of the

[03:27:50] service before you have shown the value.

[03:27:52] It's what we talked about, right? You

[03:27:53] don't, for example, if you're typing

[03:27:54] with someone in the DMs over email and

[03:27:57] they're going to ask how much is it

[03:27:58] going to cost. You never say the price

[03:28:00] before you show what you're actually

[03:28:02] delivering. You also want to say on the

[03:28:04] call that you're trying to give them a

[03:28:05] 5x return on investment. They need to

[03:28:07] see the value that they are going to get

[03:28:10] to compare it to the price that they're

[03:28:11] going to pay. You want to be judged on

[03:28:14] the value, not on the price. And the

[03:28:16] only way that you can do that is that

[03:28:17] you hop on a call with them. You talk

[03:28:18] about the issues that they currently

[03:28:20] have and then show them how your

[03:28:22] solution could be a solution to their

[03:28:23] problem. If you [ __ ] this up, then

[03:28:25] you're never going to get them on a call

[03:28:26] because they're just going to see, well,

[03:28:27] it's $2,000. That's way too much in

[03:28:30] debt. But if they realize that it's

[03:28:32] actually going to save them $10,000,

[03:28:34] then paying that $2,000 is a steal. And

[03:28:36] the other thing is that a lot of people

[03:28:38] when they start selling in the AI space,

[03:28:41] they are selling AI. They're f focusing

[03:28:43] on the features like we are using the

[03:28:45] newest models. We're using the newest

[03:28:46] technology. We are running it smooth so

[03:28:48] it's as fast as possible. The reality is

[03:28:51] that most companies that you're going to

[03:28:53] work with doesn't really care that it's

[03:28:55] AI at all. All they care about is the

[03:28:58] outcome that they're going to get. Will

[03:29:00] they be able to save $10,000 a month?

[03:29:02] Yes or no. They don't care about if it's

[03:29:04] AI or what it is doing it. They just

[03:29:06] care about the outcome. So when you are

[03:29:08] on calls and you're explaining the value

[03:29:10] that you focus on the pain that they

[03:29:12] have right now, the amount that they're

[03:29:13] paying extra right now, which they don't

[03:29:15] have to, and then from that, show them

[03:29:17] the outcome. Show them what you can help

[03:29:19] with. Don't focus on the features. Don't

[03:29:21] focus on it being AI. I hope that

[03:29:23] clarified some things in terms of

[03:29:25] pricing. Now you have all of the

[03:29:27] information that you really need in

[03:29:28] order to go out and actually start

[03:29:29] selling AI services. But in what

[03:29:32] sequence do you actually do it and what

[03:29:34] does your road map look like? Now that

[03:29:36] is what we're going to talk about in the

[03:29:37] next part of the course.

[03:29:40] Now you have literally everything you

[03:29:42] need in order to go out and land your

[03:29:44] first clients. Don't let this just be

[03:29:46] another YouTube video that you watch,

[03:29:48] forget about, and where you never

[03:29:49] actually take action. Go out and

[03:29:52] actually use this information. To make

[03:29:54] it extremely easy for you, let me give

[03:29:55] you a road map that you can do right

[03:29:57] now. The first thing is what you've

[03:29:59] actually already completed. it is to

[03:30:01] learn a skill. Just from watching this

[03:30:03] video alone and understanding how to

[03:30:05] build with tools like cloud code, you

[03:30:07] already now know a valuable skill that

[03:30:09] you can go out and sell. You should of

[03:30:11] course still keep learning as much as

[03:30:14] humanly possible. The AI space is always

[03:30:16] moving. So you need to stay up to date,

[03:30:18] but don't get in the trap of overarning

[03:30:21] and never actually going out and

[03:30:22] utilizing this information that you're

[03:30:24] learning to anything. So, while you're

[03:30:26] still learning and while you're still

[03:30:28] keeping up to date with the AI space,

[03:30:30] now it's time to go out and land

[03:30:32] clients. And these clients are not going

[03:30:34] to be the one that pays you the most.

[03:30:37] Start by working for free. Reach out to

[03:30:40] people on cold email like I've shown you

[03:30:42] how to do this in course and offer to

[03:30:44] design a website for them for free.

[03:30:46] become their internet partner or get

[03:30:48] started on Upwork like you also know how

[03:30:51] to do now and charge something low like

[03:30:53] $15 an hour just to get your first

[03:30:55] project done and to build that Upwork

[03:30:57] profile that can become a very very

[03:30:59] valuable asset. One of the biggest

[03:31:01] mistakes beginners make in this space is

[03:31:03] that they think well I'm starting an AI

[03:31:05] agency so I should spend most of my time

[03:31:07] actually learning AI but that is

[03:31:09] actually not the truth. The main thing

[03:31:10] you should be learning is how do you

[03:31:12] market it? So, how do you do marketing

[03:31:14] for your services and how do you sell?

[03:31:16] Those are the first things that you need

[03:31:18] to figure out and those are usually the

[03:31:20] things that are the hardest for most

[03:31:22] people in this space. People massively

[03:31:24] underestimate how much time they should

[03:31:27] actually be spending on marketing and

[03:31:29] selling. If you're just starting a

[03:31:31] business, try and spend 4 hours every

[03:31:34] single day doing marketing and doing

[03:31:36] selling. So spend four hours outreaching

[03:31:39] to everyone you know, sending out Upwork

[03:31:41] proposals, building code email

[03:31:43] campaigns. Do everything you can in

[03:31:45] order to land that first client. And

[03:31:47] again, the first step is not going to be

[03:31:49] to make a bunch of money. You're still

[03:31:51] building the skill, which is the actual

[03:31:52] valuable part. The next step after

[03:31:54] landing a couple of projects and landing

[03:31:56] a couple of clients is literally just to

[03:31:59] sustain yourself. And when you talk

[03:32:01] about making it a business, right, the

[03:32:02] success criteria of making it in

[03:32:04] business is to stay in business. This is

[03:32:07] the point that you need to get to

[03:32:09] because if you can just sustain

[03:32:11] yourself, if you can just pay yourself a

[03:32:13] small salary every single month that

[03:32:15] covers your rent so you don't have to

[03:32:16] worry about it, that literally means

[03:32:18] that you can go all in and that you will

[03:32:20] never give up. People say that nine out

[03:32:21] of 10 businesses fail, which is true,

[03:32:23] but the majority fail right here at step

[03:32:25] one and two. And that's because they

[03:32:27] never reach the point so at one point

[03:32:29] they say, "Fuck it. Let me try something

[03:32:31] else." And they give up. or they might

[03:32:33] just be forced to, well, this month I

[03:32:34] can't pay in. I need to go out and get a

[03:32:36] job, so I'm going to quit with this or

[03:32:37] hold off for a bit. And they never get

[03:32:39] started again. See if you can get to

[03:32:41] this point right here. This should be

[03:32:42] your main goal right now. Can you just

[03:32:44] sustain yourself at two to $3,000 a

[03:32:47] month, get to that point because then

[03:32:49] you never quit again. And then after

[03:32:51] that, you reach step four, which is that

[03:32:53] you start charging more. At this point,

[03:32:55] you should have a consistent stream of

[03:32:57] clients coming in either through Upwork,

[03:32:59] either through cold email. And when you

[03:33:00] have that, that's when you charge more.

[03:33:03] You choose the clients that can pay you

[03:33:04] more and you say no to the irritating

[03:33:06] clients that are broke and that can't

[03:33:08] pay. So you work with fewer clients that

[03:33:10] pay you more. Step number five. Now you

[03:33:13] need to find bottlenecks. Scaling a

[03:33:16] business is basically just the process

[03:33:18] of finding out the bottleneck, solving

[03:33:20] that bottleneck, and then going to the

[03:33:22] next bottleneck. Let's say that this is

[03:33:24] a funnel that looks like this. Let's say

[03:33:26] that we're getting a bunch of leads from

[03:33:28] cold email, right? They all come through

[03:33:30] here. But then we have a sales process

[03:33:32] right now where it's only us taking

[03:33:34] sales calls and this only means that we

[03:33:35] have time for some calls and our

[03:33:37] calendar is completely booked up. The

[03:33:38] bottleneck of this business right here

[03:33:40] is to hire more salespeople or by

[03:33:43] figuring out how can we only get the

[03:33:44] most qualified through this funnel. So

[03:33:46] maybe you open this up and all of a

[03:33:48] sudden you get way more calls and way

[03:33:50] more closes. Now the bottleneck might be

[03:33:51] that you only have yourself as a tech

[03:33:53] person. So then you need to hire more

[03:33:55] tech people or you need to automate more

[03:33:57] with cloud in order to make the

[03:33:59] onboarding process more smooth, better,

[03:34:01] etc. But that is the process of scaling

[03:34:02] a business and that's step five. And

[03:34:04] honestly, you're going to be at step

[03:34:05] five forever. That is the process. Now

[03:34:08] you're just finding bottlenecks, solving

[03:34:09] those bottlenecks and this basically

[03:34:11] just keeps going on forever as your

[03:34:13] business grows. This course is about

[03:34:15] building a one person business with

[03:34:17] clawed code. But this doesn't mean that

[03:34:19] you should limit yourself to just being

[03:34:21] one person. You start out by being one

[03:34:23] person in the business and Claude code

[03:34:26] can definitely carry a lot of that

[03:34:27] weight and doing it and do a lot of work

[03:34:29] for you and you can probably get to 10

[03:34:30] 15 clients yourself where Claude does

[03:34:33] most of the work. However, don't get

[03:34:34] stuck in this mindset of okay, I

[03:34:36] shouldn't hire anyone because sometimes

[03:34:38] the best solution is not AI and that's

[03:34:40] even coming from me. Sometimes the best

[03:34:42] solution is to just find a good

[03:34:44] salesperson that can convert more calls,

[03:34:46] you can close more deals. Or sometimes

[03:34:48] the best solution is literally just to

[03:34:50] find another tech person that can

[03:34:52] onboard more clients and make you more

[03:34:53] money. A mistake that I see a lot of

[03:34:55] people make is that they try to overmate

[03:34:57] where the easy solution that they could

[03:34:59] have done in a couple of days is usually

[03:35:00] just hiring someone and instead they

[03:35:02] spend weeks building out and automating

[03:35:04] a solution that then turns out mediocre.

[03:35:06] But that is the endgame. The reason why

[03:35:08] this course has built a oneperson

[03:35:10] business is because that's how you start

[03:35:12] and that's also how you can get to make

[03:35:13] a really good living for yourself just

[03:35:15] working by yourself with cloud code. I

[03:35:17] hope that this road map is valuable for

[03:35:19] you. Write it down and literally just

[03:35:21] start outreaching. Start setting up your

[03:35:23] cold email campaigns. Start setting up

[03:35:25] work. Buy connections. Send proposal

[03:35:27] requests. Actually spend a bunch of time

[03:35:29] doing these Loom videos and just keep

[03:35:31] going. When I started my business, I was

[03:35:34] considering myself smart. But it still

[03:35:36] took me 4 months of just straight grind

[03:35:39] outreaching every single day before I

[03:35:40] landed my first client that paid me 400

[03:35:43] bucks. These expectations of being able

[03:35:45] to make tens of thousands of dollars

[03:35:46] your first month is just completely

[03:35:48] unrealistic because you haven't learned

[03:35:50] that skill that makes you worth that

[03:35:52] much. In order to do that, you actually

[03:35:53] need to build something up that can take

[03:35:55] months if not years. And the best thing

[03:35:57] that you can do right now is to set your

[03:35:59] expectations right from the start. Say

[03:36:00] to yourself that it's okay if I put my

[03:36:02] heart and soul into this and that it

[03:36:04] takes one year. Don't expect it to take

[03:36:06] a month. And when it comes to growth in

[03:36:08] a business, it's usually exponential.

[03:36:10] You're going to have a pain in the

[03:36:11] start. You're going to be hovering

[03:36:12] around a couple hundred bucks a month

[03:36:14] probably when you're just starting out.

[03:36:15] But then all of a sudden, something

[03:36:16] clicks and that's when the graph goes

[03:36:18] like this. And just like the stock

[03:36:20] market, it doesn't go in a nice curve.

[03:36:22] It goes like this, right? It goes up and

[03:36:24] down, up and down. And then you're going

[03:36:26] to have good months. And then you're

[03:36:27] going to have terrible months. And

[03:36:28] you're going to have even better months.

[03:36:29] and then you're going to have even

[03:36:30] bigger falls and then it goes like this,

[03:36:32] right? That is how it always goes. Take

[03:36:34] it from me, keep grinding and it's going

[03:36:36] to be so worth it when it actually

[03:36:38] clicks. The last thing I want to say is

[03:36:39] that if you're already in a job or

[03:36:41] already in a position right now, don't

[03:36:43] let that stop you. It's a myth that you

[03:36:46] have to quit everything that you're

[03:36:48] doing right now in order to go all in on

[03:36:49] something because for a lot of people

[03:36:51] that actually keeps them from starting

[03:36:52] in the first place because what if you

[03:36:54] cannot afford to quit your job? That is

[03:36:56] a big risk if you don't make this work

[03:36:58] the first couple of months. Instead,

[03:36:59] what you do, and this is by no means

[03:37:01] easy, is that you build your business

[03:37:02] from 5 to 9. Then you are at work from,

[03:37:06] let's say, 9:00 to 5:00, and then you

[03:37:08] might be able to get a couple of hours

[03:37:09] in at the evening before you then go to

[03:37:11] sleep, right? That still gives you 4

[03:37:13] hours right here and 2 hours right here.

[03:37:15] And usually you're the most aware right

[03:37:17] when you wake up. So spend this time

[03:37:18] working on your business. And then

[03:37:20] probably in weekends, you probably have

[03:37:21] 10 hours that you can spend on your

[03:37:23] business. So this you have to go all in

[03:37:25] when you're just starting out. It's a

[03:37:26] myth. You can start getting your first

[03:37:28] couple of clients, making your first

[03:37:30] money, setting up your Upwork profile,

[03:37:32] and do all of that while you're still

[03:37:33] working a job. Don't quit your job.

[03:37:35] Don't think that you need to quit your

[03:37:36] job. Do it on the side until you can

[03:37:38] actually afford to quit. I really

[03:37:40] appreciate that you watch this full free

[03:37:41] course, and I really hope that you found

[03:37:43] it valuable. If you want my help, then

[03:37:45] you should check out our community, the

[03:37:47] 1% in AI. Inside of here, we have the

[03:37:49] 30-day challenge and the 90-day

[03:37:51] challenge. And the great thing about

[03:37:52] these challenges is that when you

[03:37:54] complete them, then you actually get

[03:37:55] your first month completely refunded.

[03:37:57] The way it works is that you get this

[03:37:58] tracker right here that shows you what

[03:38:00] to do every single day, what modules to

[03:38:02] watch inside of the 1%, what task to

[03:38:04] complete, and it basically holds your

[03:38:06] hand throughout the entire process of

[03:38:08] learning the skills that you need to

[03:38:09] learn and setting up your AI agency. So,

[03:38:10] if that sounds interesting, then I'll

[03:38:12] leave it as the top link right below

[03:38:13] this video. Thank you guys so much for

[03:38:14] watching. I hope you have a wonderful

[03:38:16] rest of your day. I hope that you're

[03:38:17] actually going to use this for

[03:38:19] something. Don't let this be another

[03:38:21] YouTube video that you've just watched

[03:38:22] and never actually took action. I am

[03:38:24] rooting for you.