# How Mach 1 Stores is Using Taiga & AI to Make Real Decisions From Raw Data

https://www.youtube.com/watch?v=VUh_0EPoFII

[00:06] Hello, it looks like we are live.
[00:08] So, we're just going to wait a few more minutes for a few more people to join and then we'll get going on the webinar.
[00:14] So, again, this is this is Jenny Hill with Taiga and we're going to talk to Mach 1 here in a few minutes.
[01:47] Hi, everyone.
[01:49] We're going to wait about 1 more minute.
[01:51] We've got people joining here quickly and then we'll get going on the webinar.
[02:40] All right, let's go ahead and get started.
[02:43] Welcome everyone.
[02:43] Um I am Jenny Hill.
[02:46] I am the customer success lead here at
[02:49] Taiga.
[02:50] And today we're going to be talking about how you can take raw data and make it turn it into real decisions with AI.
[02:56] Um we have a guest speaker who's one of our clients and um as well as another member of the team that I will go through and introduce.
[03:06] And then we'll get right into the um the meat of the presentation.
[03:12] So uh while we're speaking, there is a chat.
[03:14] So if you have any questions that come up along the way, uh you are welcome to throw those in the chat and we'll try to get those answered at the end.
[03:22] And for anything that is not answered, we will follow up with you um after after the call.
[03:29] So and it looks like John just posted um he's going to be helping to monitor that as well.
[03:35] Um all right, so here we go.
[03:37] Uh let's talk a little bit about AI, which is I know it's been the hot topic in in uh almost every conference and presentation you probably have been involved in in the last at least for sure the last month or so.
[03:53] The speakers today are myself, Jenny.
[03:56] I'm going to be monitoring or moderating, I should say, and trying to keep Alan and John on track here with the presentation.
[04:03] to be monitoring us, too.
[04:05] I'm going to be monitoring also.
[04:07] Yeah, good luck with that, Jenny.
[04:10] Yes, I'm yeah.
[04:11] So, if I get cut you guys off, you know, don't take offense.
[04:15] First of all, we have John Oakley, who is the CIO and co- one of the co-founders of Taiga.
[04:20] He's going to be engaging this conversation with Alan Meyer, who is the CEO of Meyer Oil, which does business as Mach 1 Stores in the Midwest.
[04:32] So, welcome, gentlemen, and we're looking forward to to getting going and hearing what you guys have to share.
[04:39] What we're going to cover today is we're going to talk a little bit about and and Alan's going to give his perspective on what life before Taiga was like.
[04:44] So, Alan is one of our very first clients, and so he's going to talk about how things were before Taiga, and then what
[04:54] He's doing now.
[04:55] He's going to talk about how how he chose us and how he has evolved Taiga into the AI into an AI-powered workflow.
[05:03] So, then we're going to get into what's next, and we actually have we're we're teeing up a webinar part two with Mach 1, and we'll talk a little bit about that, and then we have a Q&A.
[05:17] So, get those questions in.
[05:19] All right, so I'm going to turn it over to Alan.
[05:21] Why don't you give us a little bit of background about your stores?
[05:25] Yeah, thanks, Jenny.
[05:25] So, 25 stores as you see there, we've got 18 gas stations, seven truck stops, built our last one last November.
[05:39] Um the branded Phillips 66 and Marathon, so and continuing to grow.
[05:43] So, do you want me to get into kind of how I met you guys and all that now or?
[05:51] Sure, go for it.
[05:53] You can Yeah, go for it.
[05:54] So, as you had on that last slide, 2021,
[05:59] I got a cold email from Bill Ivers at Etega.
[06:03] Never met him or didn't know anything about him and basically said, "We'll take your data and put it on a website for you."
[06:09] And we had tried that with a couple other companies and and never even got to where we could see our data.
[06:16] So, replied to the email, said, "Yeah, I'd I'd definitely do a webinar."
[06:21] And long story short, Bill said, "Give me a chance and I'll have your data on our portal and by the end of the week."
[06:31] And mainly just because I didn't believe him, I said, "Let's try it."
[06:32] And what I mean, 48 hours later, we were looking at our transacting level data through your portal.
[06:39] And I mean, it was taking weeks or months with other companies and it never worked.
[06:46] So, became a pretty big believer pretty quick there and over, you know, the one of the biggest things is, you know, you guys are a
[06:54] relatively small software company at the start and kind of watched and assisted with seeing this thing grow, which is kind of how I'm a bigger fan of it than working with large companies that aren't very flexible or agile.
[07:07] So, you guys kind of fit the culture and the profile I was looking for, too.
[07:15] So, and over the years, I've always go from being frustrated because I couldn't get access to my data and over the last couple years basically saying, "Now I've got so much data, I don't even know what to do with it."
[07:31] And um I think AI has a long way towards fixing that problem.
[07:35] Awesome. And we'll get into exactly what you're doing here in a minute.
[07:39] Um so, thank you thank you, Alan, for that.
[07:42] So, let Oop, let me go to the next slide.
[07:45] So, for those of you who are not familiar with Taiga, John's going to talk a little bit about how it works and then, you know, how do you how do you get this real how do you get this raw data and put it
[07:56] into actionable insights.
[07:58] So, John, would you mind going through a little bit about how Taiga is set up?
[08:02] Yeah, sure.
[08:04] So, Taiga platform you know, we're a data platform first.
[08:08] So, you can think of us as you know, think of like a an iceberg.
[08:14] You've got, you know, the big part of the iceberg that's underwater.
[08:17] That's our data platform and then on top we have our mobile friendly display layer, everything from executive dashboards, things focused for category managers, workforce optimization, all sorts of different tools for running convenience retail and then getting great answers for anybody in your team.
[08:34] Um, beneath the water we have all of those integrations to pretty much every system you already work with and all that comes into one consolidated model within Taiga and gives you nice clean apples-to-apples answers across all your stores.
[08:50] And so, that's kind of, you know, the short sweet answer about how you can understand Taiga.
[08:53] We're a we're a data platform first.
[08:55] We want to make it so
[08:57] that any answer you want to answer about your business, you've got the data to back that up at the time you need it where it's ready and able for you to get to it.
[09:09] And then here's just a little bit more if you would do you want to go into this this slide as well?
[09:14] It's a follow-up to the previous.
[09:16] Sure.
[09:18] So, I guess a big problem in our industry of convenience retail is this chimera of retail.
[09:23] We've got all kinds of different technology that was, you know, originally started in grocery or or boutique retail or all these other places, gas stations and stuff like that.
[09:34] And you all have done a, you know, amazing job of, you know, finding and picking the best technology to to, you know, help, you know, really run your businesses.
[09:43] But, when it comes to understanding what those things are all doing and then seeing the big picture across all those things, they're all locked up in all these different silos of data.
[09:51] And, um you know, that that is what we call the data fragmentation problem.
[09:54] That's, you know, I've got all
[09:57] This information over in my point of sale systems.
[09:58] I've got some of this information in my back office system.
[10:01] I've got some of this information in my loyalty programs.
[10:04] I've got some of this information in my payroll and HR systems, etc.
[10:08] Uh you know, food prep, all the other things that you guys are doing like to stay uh stay on top of the game, to stay relevant, to, you know, drive your business forward.
[10:17] All these require various different technology systems, and they don't always work together very well from a um you know, understanding what's going on perspective.
[10:27] We We try to help you to to get that picture put back together.
[10:33] Thank you, John.
[10:36] All right.
[10:38] So, moving on to back you know, back to you, Alan, and um with some commentary with John as well, but this is the way that Mock 1's technology environment is is set up.
[10:47] Do you guys want to talk a little bit about um Alan, what you've got going on from a technology standpoint?
[10:54] so yeah, so we have NCR, their RPOS version
[10:59] of software.
[11:01] So, we've got that at all stores.
[11:05] And so, we've got that where without Tega, I'd be using what they call remote service manager, which is kind of it is it's in a document data per location that I can look at.
[11:14] I can pull up any transactions by day or whatever, but it's all kind of within the ecosystem of that store.
[11:24] Uh I Rely as our back office software.
[11:25] Um that we were using them for department level sales, accounting, so on and so forth, but train we were they don't even pull transaction level data as far as I know.
[11:37] Uh Ignite is our loyalty program, always has been, so we kind of same thing.
[11:42] They have a list of the transaction data that's done within the loyalty program.
[11:48] Uh the food service, we recently moved all of our food service production uh over to Day Cô Genie.
[11:56] So, everything that gets managed from
[11:59] met food service labeling, production data, all that's with Day Cô Genie today.
[12:07] Awesome. Thank you.
[12:12] All right. So, now let's get into the to the meat of the presentation and talk about AI.
[12:18] Um it's an exciting, but also overwhelming topic in all industries.
[12:25] And so, there's a lot of there's a lot of encouragement to experiment with it, see how it works for you, but there's also, you know, some apprehension about any you know, pitfalls that might that might occur.
[12:36] Can you tell us a little bit about your AI journey so far, Alan, and and what you've found?
[12:42] Yeah. Uh so, I was at a conference in Florida in November, and uh don't think I even knew it at the time, but John Oakley was there with Tega.
[12:50] And one of the presenters was a former Google executive that worked in the AI division.
[12:57] And sat right next to John and
[13:00] listened to this former Google employee.
[13:02] spend basically an hour explaining to me that, you know, everybody says AI is coming.
[13:07] Uh I'm going to spend the next hour showing you that AI is not coming.
[13:12] AI is here today.
[13:14] And before that, I was kind of using AI as pretty much everybody in the world, as far as I can tell, is mostly using it.
[13:19] They're using it kind of a glorified Google search as one-off questions.
[13:23] You know, you have a question that you used to Google, now you're using AI.
[13:28] Um the Google guy in that presentation I mean, it was nothing short of absolutely blowing me away with what he was able to accomplish.
[13:37] I had no idea that AI could do that today.
[13:39] So, I had a dinner scheduled that night after the meeting.
[13:43] I was I had my eyes uh wide open explaining to John, "Holy crap, this is amazing."
[13:48] John was already fully aware and probably didn't learn anything in that meeting.
[13:53] Uh so, he was explaining to me kind of some more stuff and kind of getting into the weeds.
[14:01] So, I quickly canceled that dinner that I had that night, much to some people's chagrin, and went to dinner with John.
[14:07] And we I sat there for 4 hours as John explained to me where AI is at and where it's coming.
[14:15] And I still I think that was one of the most powerful dinners I've ever had in my entire life.
[14:18] So, came back from Florida.
[14:21] So, I okay, this is we absolutely have to learn what this is going to be.
[14:27] So, signed up for a couple online classes and uh signed up for a couple's subscription accounts and watch YouTube videos.
[14:34] And just kind of dove deep into it and started screwing around with it to see what it can do.
[14:41] And it was equally frustrating as it was productive.
[14:48] Uh trying to figure out exactly what No, I knew this was something, but I wasn't getting what I was wanting to get and I really didn't know how to get from point A to point B.
[15:00] So, I don't know how much more you want
[15:01] to me to expand on that right now, uh Jenny, but that's kind of my initial uh experiences with AI.
[15:09] Um well, we can definitely jump into it.
[15:11] So, so the one of the um the things that I've heard at at several conferences,
[15:16] some people have even asked about Tiger specifically, you know, what is Can't I just go into one of these AI platforms and ask it to do all the things that maybe Tyga's doing or some you know to something like that.
[15:29] And so Can you talk a little bit about that?
[15:30] I tried doing. That's literally what I that was my first attempt at AI.
[15:35] Uh so, you know, kind of what I was doing and what that kind of reminds me of is calling, you know, basically just putting a data dump into an AI project or an AI agent and thinking that you can give it you know, mountains of data, you know, let's just say 3 months of every transaction in your company and then you can just start asking it questions like it should know what you did at this store on a Tuesday at 2:00 6 weeks ago.
[16:02] That's how I tried doing it.
[16:07] Uh one, if you're just creating just a data dump and you expect AI to know what you're trying to accomplish and you have a massive amount of data set that includes a bunch of information that isn't pertinent pertinent to what you're trying to achieve your AI is going to get confused really quickly and it's going to start giving you answer either answer you're not looking for or wrong answers or you're going to hit your data limits as I regularly do and you're going to have to take a step back and kind of get the key is getting your stuff structured as as we'll kind of get into.
[16:51] Yeah, that that's a very very real and common realization as you start to drive drive into these tools as it you have to think about them a little a little differently like than that.
[17:01] You
[17:04] know, the there are a lot of people out there that are you know, pushing that mantra that you know, AI is just going to figure out all things for you.
[17:13] That you know, is you know, as you can think of is very problematic and inefficient.
[17:19] Um you know, one AI is trying to please you.
[17:21] It's just trying to come up with an answer that you're going to want.
[17:22] Uh you give it too much stuff that's unrelated to what you're trying to talk about, it's going to get confused.
[17:29] You have to think of it more like um um the way uh from a training perspective.
[17:33] So, if you are training an employee uh coming on to your organization, you know, you would provide them with the right tools to help them to be successful.
[17:40] You have to do that with AI, too.
[17:42] And so, one of the fundamental things it needs is clear and certified data sources.
[17:49] And so, like, you know, if you're going to ask it about how am I doing in my salty snacks across my stores, tell me any insights about those right now,
[17:58] what it needs to understand, you know, how convenience retail organization operates, what what salty snacks means,
[18:04] where where those go, where they're positioned within the store, how those have performed for you over time, and then all the stuff about the actual data what's going on with those in your stores in a clean way.
[18:16] Um you know, that that's the that's the part that that approach to it misses on is is it you you have you have to set your AI up for success.
[18:25] Um and that, you know, involves, you know, being able to to guide it and give it the right information.
[18:31] You can't just say, "My data is all over here in this giant bucket of all this stuff that's happened for forever, go figure it out."
[18:36] It's going to it's going to get confused.
[18:40] It's going to make stuff up.
[18:42] It's going to find wrong answers.
[18:44] And ultimately, the result is you make bad decisions or uh you know, it it it worse, you know, you make, you know, something happen in your organization where you lose money.
[18:54] Or you know, make a decision that that causes harm to your business.
[18:58] Yeah, at one point, John, my AI was telling me how to improve my Domino's pizzas uh service.
[19:04] And I don't own any Domino's
[19:06] Pizzas.
[19:08] Well, it probably assumed that like you had had something you know, Italian food or pizza related.
[19:14] And so it it it had some, you know, world experience with pizza and it came from.
[19:17] That's good to know how to improve my Domino's if I ever choose to go that route.
[19:20] Yeah, if you ever go down that route, it'll probably be able to help you.
[19:23] That. Got that knowledge in there now.
[19:26] Exactly.
[19:29] All right, let's get into the Excuse me, some of the actual workflows that that you went through, Alan.
[19:35] Um Can you walk us through what that looks like in actual practice?
[19:40] Yeah, so I mean, there is I mean, I think we've got 25 various projects that we're kind of actively using AI for today.
[19:52] And what the majority of them basically is is take this data set from this source of data, take this data set from this separate portion data, and then tell AI how to combine Look at these two data sets, make sure the AI understands what the
[20:08] Data sets are, and then make sure it understands what you're trying to achieve.
[20:14] And then, you know, for years we'd look at this spreadsheet over here and this spreadsheet over here.
[20:17] They're both telling me something important, but I got to combine those two to kind of create a metric that means something to me.
[20:24] That is where today I think AI AI excels at.
[20:30] So, uh one use case where we really started diving in as a use case uh today when we're first getting into AI was food service.
[20:38] So, you know, when you think of we are entirely a grab and go food service program, one is our proprietary label, Casey's Kitchen.
[20:46] Second is our uh Little Caesars.
[20:49] We do have four Little Caesars programs, but all of this grab and go.
[20:53] So, you know, if you think of our how we go to market with our food service, there's really three uh metrics that really matter.
[21:02] How much food are you making?
[21:04] How much of that food that you made are you going to sell?
[21:06] So, now you know your revenue, you know your waste,
[21:09] and how much labor did it cost to
[21:11] produce that? So, you know, you have
[21:13] these three data sets. You got your
[21:14] labor chart, you got your sales chart,
[21:16] you got your production chart.
[21:18] When we first started trying to do this,
[21:20] we were basically just creating a data
[21:22] dump of our transaction journals to try
[21:25] to figure out and then trying to combine
[21:28] that. And ultimately, it led to me
[21:29] calling John Oakley saying, "AI is uh
[21:33] it's broke it literally I literally
[21:34] broke the AI. The AI told me that
[21:37] um I refused to work on this project is
[21:40] ultimately what it did. And I tried
[21:42] reloading the project, I tried doing it
[21:43] on I mean, I tried everything every way
[21:45] from Sunday to try to figure out how to
[21:47] fix it. And it kept telling me, "No, you
[21:49] keep asking me to do this for you and
[21:51] I'm not going to do it." And that's when
[21:53] John kind of showed me how to better
[21:56] structure a project. So, now you know,
[22:00] we've got it to where we filter the data
[22:03] with Tega and it has a time stamp. Every
[22:07] item sold in our company has a time
[22:10] stamp, has the item sold, the revenue on
[22:13] it, and which location it was. And with
[22:16] Tega, I can go back 1 week, 1 month, 1
[22:19] year, 2 years, whatever it is. That's
[22:20] always sitting in their environment for
[22:22] me to pull. And I can pull 1 week or I
[22:25] can pull a whole year's worth of data.
[22:27] Um
[22:29] I grabbed that, I grabbed the production
[22:31] report, and I grabbed these timesheets.
[22:33] And so, through a series of trial and
[22:36] error and kind of getting the
[22:37] instructions for the AI and the data in
[22:41] the format that I want it to be where I
[22:43] can easily scale it out as new data
[22:46] comes in. That's how I built out that
[22:48] project. So, today every way, John.
[22:54] This is the same thing you
[22:55] Yeah, here. I'll I'll pull that up and
[22:57] then that way that way you won't have to
[22:58] talk to it as much as as uh
[23:01] Yep.
[23:02] So, this is something that we created
[23:04] for me and the food service director
[23:06] kind of built this out in our AI
[23:08] project. So, me and him share this AI
[23:10] project. We are uploading the data
[23:13] uh me or him, and then
[23:16] from there we can build out a series of
[23:18] tentacles of where we want to go with
[23:20] it, but this is kind of an overview page
[23:22] that we send out to our ops team. So,
[23:24] our COO, our DMs, our store managers,
[23:27] this is kind of the the 10,000 foot view
[23:31] of our food service performance. So, if
[23:33] you look at the series of tabs that uh
[23:36] is on the screen there,
[23:38] one on the location overview tab, it
[23:41] kind of gives the those three metrics I
[23:43] talked about by location. It talks about
[23:45] our sales, our cost of goods sold, our
[23:48] waste percentage, our by category,
[23:52] little cedar waste percentage, and our
[23:53] labor cost, and then shows our profit.
[23:56] So, you got your 10,000 foot view by
[23:58] location, then you have your sales
[24:00] breakout by in a 13 period calendar.
[24:05] Then then it shows your sales versus
[24:07] waste by category by location. Then it
[24:10] shows the sales versus waste by every
[24:12] item in the in the company.
[24:15] Then it goes on through
[24:18] um labor efficiency. So, every Monday at
[24:22] a specific store, it says, "Hey, your
[24:24] labor cost at this store on Mondays is
[24:26] X. Your labor cost at this store on
[24:28] Tuesday is Y." So,
[24:32] it took us a series of hours to kind of
[24:35] build out how we want to structure this
[24:38] and kind of fine-tune what we want. Now,
[24:42] to get a new set of data, it takes 5
[24:45] minutes. So, we get uh this data up to
[24:51] date mo um basically every day or every
[24:55] other day. It then gets sent out for
[24:59] them to have access to and then it can
[25:00] do a couple things. One, as an AI,
[25:03] you can we the DMs or the food service
[25:07] person or myself can look at this as an
[25:09] overview page, but now the AI has
[25:11] absorbed and processed that data. So
[25:14] then if I wanted to just ask it about
[25:16] our cheeseburgers, you know, you can
[25:18] start going down the rabbit hole of any
[25:20] one schedule, any one
[25:24] uh menu item. You know, then you can
[25:26] start going down to the exciting stuff
[25:28] of AI because now you've kind of you
[25:30] made sure that
[25:32] the AI
[25:34] absorbed the data in the way that you
[25:36] wanted it to be absorbed and you put the
[25:38] instructions in there explaining exactly
[25:41] what the data is in kind of what you're
[25:43] hoping to achieve. So, you know, for
[25:47] when we're talking about changing build
[25:49] twos or changing the labor schedule,
[25:53] um or adding or removing menu items. I
[25:57] mean, that was done
[26:00] on a quarterly or bi-annual basis just
[26:02] because it took took hours to kind of
[26:04] process through the data to even come up
[26:06] with, you know, I'm a very
[26:07] numbers-driven guy as John's John knows.
[26:10] And so to me, the numbers will tell the
[26:12] story if you just get the numbers. But
[26:14] the problem is sometimes it takes a
[26:16] while to accumulate those numbers and
[26:18] and whatever takes a long time to do,
[26:20] you can't do very often cuz we're all
[26:21] very busy people.
[26:23] When you get it to where I mean, we're
[26:25] checking our labor schedules, that can
[26:28] get updated and when I say check a labor
[26:30] schedule, it will we have it it's not on
[26:33] the sheet that John just had up there,
[26:35] but on a separate pri- separate chat
[26:38] within that same project,
[26:40] at any given point in time, I can say
[26:43] run me the last four weeks of the labor
[26:45] efficiency. And so it will take, for
[26:48] instance, Friday at 4:00 p.m.
[26:51] and it'll take every Friday at 4:00 p.m.
[26:54] in the data set that I pre that I told
[26:56] it to select. So, the last four Fridays,
[26:58] the last eight Fridays, whatever I
[26:59] decide is important to me,
[27:02] it will take all those sales in that in
[27:04] that 4:00 p.m. hour of that day of the
[27:06] week. It will look at how much labor I
[27:08] had there, and then it will tell me my
[27:10] labor efficiency as a percentage of
[27:13] revenue. So, my labor cost percentage of
[27:15] revenue is 25%, 28%. So, then our food
[27:19] service me and our food service director
[27:21] has set these guidelines that okay,
[27:23] here's the threshold that we need to
[27:25] increase labor, here's the threshold we
[27:27] need to decrease labor, and gave him his
[27:30] structure on how to
[27:32] what our target range is for a labor
[27:34] schedule. And what used to get changed
[27:36] every three, six, 12 months, we're
[27:40] changing on a monthly basis. We had a
[27:42] store
[27:43] um
[27:44] we just added a 25 hours of labor at a
[27:47] store a couple days ago.
[27:49] And it was
[27:50] we never would have been that big of a
[27:52] food service increase cuz I mean that
[27:53] was 20 somewhat thousand dollars a year
[27:56] in budgeting your food service. But, you
[27:59] know, we have a target labor cost, and
[28:02] the AI clearly showed that your labor
[28:05] cost is your labor percentage is here,
[28:07] and if you want to get to your range,
[28:09] you need to add this much labor. So, it
[28:11] turned what would have been a one-day
[28:13] project into about a 10-minute project.
[28:16] Yeah. It really It really helps you to
[28:18] like with those ambiguous strategic
[28:21] problems. You know, once once you design
[28:23] playbooks like Alan has done here for
[28:26] for you know, like focus to a business
[28:29] strategy within your organization, you
[28:31] really can like you know, iteratively
[28:33] learn through that very quickly and test
[28:36] different hypothe- hypotheses. And uh
[28:39] it's really good at those, you know,
[28:41] helping you to answer those ambiguous
[28:43] questions. Yeah, so.
[28:45] Yeah.
[28:46] This is what it's uh
[28:47] A good example of mashing up a lot of
[28:49] different things there to to be able to
[28:52] to focus in on a problem.
[28:55] What
[28:56] I mean the biggest key for me and what
[28:58] I, you know, it's it's funny about AI. I
[29:01] think it went from being severely
[29:02] underrated to being completely overrated
[29:05] where people just basically were
[29:07] thinking it was like a Google search.
[29:08] Now they think it can just magically do
[29:10] everything.
[29:11] And
[29:13] um
[29:14] through talking to peers in our industry
[29:18] and you know, just other business owners
[29:19] and other industries. And I mean the
[29:21] biggest thing right now the biggest
[29:23] thing I always try and push on people is
[29:25] just
[29:26] you got to have your data structured. It
[29:28] it you can't just dump it into things.
[29:30] You know, I think you had it on one of
[29:31] these slides, you know, talking to Oh,
[29:33] right there. Treat it like an AI like a
[29:34] new employee.
[29:36] Yeah, that's perfect.
[29:36] what to do.
[29:38] Like start you know, start it starting
[29:39] for those on the on the call that are
[29:41] like okay, now what what do I do? So
[29:43] you're getting into it already.
[29:45] Yeah, sorry for jumping the gun there.
[29:47] But you know, you need a new employee
[29:48] you just
[29:50] a link to the all the transaction files,
[29:54] you know, to our stores and said go come
[29:56] up with the food service thing. You
[29:57] know, you're you're going to have
[29:58] mistakes. And so
[30:02] being very concise, figure out what
[30:04] you're wanting to achieve and make the
[30:06] data, make sure you're only giving it
[30:08] the set of data that is pertinent to
[30:10] what you're trying to achieve. Uh that
[30:12] way it doesn't get confused because you
[30:14] know, I've got my cooler sales in and
[30:17] I'm trying to ask it food service
[30:19] questions, you know, because it will
[30:20] start pulling the data sets from other
[30:22] things that doesn't matter today.
[30:25] Um and the other part about it would be
[30:28] instructions, you know,
[30:30] treating it like a new employee I think
[30:31] is a great way of saying it because
[30:34] AI is a
[30:35] uh is best done with a two-way
[30:37] conversation.
[30:39] You know, so give it the right data, put
[30:41] the instructions in the best way you
[30:43] know how to tell it what the data is and
[30:46] what you're trying to achieve, but as
[30:48] it's
[30:48] you're having these communications, talk
[30:51] to it like it's a person. And if you
[30:52] want something if you want it in a
[30:54] different format or ask it why it did
[30:56] this.
[30:57] You know, a lot of times, well, hey, I
[30:59] should add that I always double-check at
[31:02] least a small sample size to make sure
[31:04] the output matches what the input is.
[31:07] So, specific to the food service
[31:09] file
[31:11] and it tells me I sold, you know, 400
[31:15] cheeseburgers this month. I'll grab a
[31:17] specific location for a specific day and
[31:20] make sure that what
[31:22] the AI tells me I sold that day is what
[31:26] matches what Tegan says it'll match that
[31:29] day. And when those two numbers are
[31:32] wrong, I'll just ask AI, "Well, I've I
[31:35] see that I sold 12 cheeseburgers at that
[31:37] location that day and you show me eight.
[31:39] Why are we wrong?" And it will run
[31:41] through its analysis
[31:43] and
[31:44] sometimes it's wrong and sometimes I'm
[31:46] wrong. Sometimes it will say, "Well, you
[31:49] miscounted." And then I apologize to the
[31:52] AI and say, "You were correct. I I had
[31:54] that wrong." Other times,
[31:57] uh
[31:58] it will tell me how it screwed up and
[32:00] then I will say, "Well, okay, how did
[32:03] you screw up?" It will tell me why it
[32:05] screwed up. Then I'll say, "Well, put
[32:06] into your memory what you did, what how
[32:10] you screwed that up and to not do that
[32:11] again so you never screw up that way
[32:13] again. So, there's a lot of ways like
[32:15] that that you're creating this two-way
[32:17] communication. That was hard pressed by
[32:20] me by John Oakley because John always
[32:22] said, "Hey, you got to
[32:23] you got to be careful and you got to
[32:25] double-check because it will be wrong.
[32:28] There's it's not a matter of if, it's a
[32:30] matter of when. So, prevent it, minimize
[32:34] it as much as you can and you got to
[32:35] kind of cross double-check this stuff.
[32:38] Yep. And you really like like Alan
[32:41] saying here, like you should have
[32:42] regular performance meetings with it.
[32:44] Like, okay, tell me what's working, tell
[32:46] me what's not not working. If we were to
[32:48] bring on another agent or you know,
[32:51] train some train a person to do this,
[32:53] would they have all the right
[32:54] information to do what you're already
[32:56] doing or have we not documented this
[32:58] stuff out for that? You know, yeah,
[33:00] because
[33:01] uh
[33:02] um one one thing that's difficult about
[33:05] uh these tools to see at the surface is
[33:08] every time you start a new session with
[33:09] it, like it's it is a a new session. It
[33:12] it's it's reading through whatever you
[33:15] provided in the past and it it's making
[33:18] assumptions based off of that. It's not
[33:19] like you're talking to the same one you
[33:20] had the last session with. And so, you
[33:23] have to make sure that you're having
[33:25] these regular meetings to plan for the
[33:27] long term with it.
[33:28] You know, can we make a long-term
[33:30] playbook of this? How would you teach
[33:31] another agent how to do this? Have we
[33:33] done that already or should we do that
[33:35] now? Do you know everything you need to
[33:37] know to be able to accomplish this task?
[33:40] Please send that back to me in a list of
[33:43] bullets here so I can review it ahead of
[33:44] time because I want to make sure you're
[33:46] doing your job well. Um it's all about
[33:49] designing your intent when you're
[33:51] working with these tools. Uh we are very
[33:54] much in, you know, what I would consider
[33:55] the Wild West period of working with AI.
[33:58] It's very similar to me being an older
[34:00] guy uh to um back when, you know, the
[34:04] first few
[34:05] uh iterations through uh the the early
[34:08] internet days where lots of opportunity
[34:10] but there's also lots of pitfalls. Um
[34:13] you should be using these tools but you
[34:15] should be using them in a very safe way
[34:17] and boxing it into a place where you
[34:19] know that you can control it. Um and and
[34:23] make sure that you can be successful
[34:24] with it but uh it is very it is very um
[34:28] um powerful tool already that you can
[34:31] work with and there's lots and lots of
[34:33] variety out there to
[34:35] to excel with it. That being said, along
[34:37] with that, there's lots of new companies
[34:39] coming into our industry and to other
[34:41] industries. They're doing a great job
[34:42] with their models that we're working
[34:44] with and partnering with to help them to
[34:47] be successful as they take us to the
[34:49] next stage and the stage after that when
[34:50] it comes to these tools.
[34:53] I think that
[34:55] a couple weeks ago John and I were at um
[34:57] we're at Chicago at SOI and he was
[34:59] showing me even more things that we can
[35:01] do
[35:03] AI and Taiga and everything and what I
[35:04] think I was most surprised about was
[35:07] like when we would go into a session. So
[35:09] we'd have our breakouts but then when we
[35:10] would go back to a session, John would
[35:12] actually tell it to stop working. Like
[35:15] you need to stop working. I'm going into
[35:17] this session because it would just keep
[35:20] going and I think that was a
[35:22] light for me that in that you really
[35:24] have to tell it what to do and what not
[35:26] to do. And when we got back he'd say,
[35:28] "Okay, we're working again now." You
[35:30] know, and then and take it from there.
[35:33] responsibility.
[35:34] Right. It it is I don't you know, it's
[35:36] not a person but you if you can treat it
[35:38] again like you guys are saying as you
[35:40] would a a new employee who doesn't know
[35:44] much background of what's going on then
[35:46] then that really does help as well.
[35:48] So so if you were starting
[35:49] Yeah, and that you know, and that point
[35:51] too Jenny, the
[35:53] you know, that food service thing is
[35:54] just kind of one very specific use case
[35:56] but as you know, you have your agent and
[35:59] your agent in your account and you start
[36:02] talking about like you know, I've got a
[36:04] couple different AI subscriptions and it
[36:06] knows it's Alan Meyer, it knows it's CEO
[36:08] of Mock One Stores.
[36:10] One thing when you're ever you're
[36:11] talking to
[36:13] an AI and you can you're want to talk to
[36:17] about a take like if I want to ask
[36:19] AI about Taiga, I can describe Taiga or
[36:23] I can use Taiga and then put in the
[36:24] website. It will read the website and
[36:26] then it'll know everything about taking
[36:27] it just by that website. So, think about
[36:30] ways to kind of make sure a your AI I
[36:33] use Claude and Gemini, but whatever one
[36:36] you use, making sure it understands who
[36:39] you are,
[36:41] what your company is, you know, what is
[36:44] your goals just as a company not as an
[36:46] AI.
[36:47] And then
[36:48] I mean, it's it is crazy. It used to be
[36:52] where, you know, you start thinking of
[36:54] like you'd have these questions. Now my
[36:56] default is always, "Okay, how can AI fix
[36:58] that?" So, you know, whether it be
[37:01] market analysis,
[37:03] real estate,
[37:05] you know, for growth, for M&A, for
[37:08] auditing procedures. I mean, if it is
[37:11] data that you can then process to spit
[37:15] out an outcome.
[37:16] Um
[37:18] you should either I mean you are either
[37:21] putting that on AI to today, you should
[37:23] have that on a roadmap to put that on AI
[37:26] today, or you're going to be falling
[37:29] behind your competition because
[37:32] I think AI is
[37:34] um
[37:35] I think AI is two things. One, I think
[37:37] it's the great equalizer of our
[37:39] industry. I don't need to be a 2,000
[37:43] seats for company to afford AI. I need
[37:46] $20 a month. So, it's amazing. I think
[37:49] this thing can be a great leveler that I
[37:51] don't need to scale to have it.
[37:54] But two,
[37:55] you know, the the people in our industry
[37:59] that are the hyper consolidators, the
[38:01] large players, I guarantee you they've
[38:03] got teams of people working on this
[38:05] right now. And
[38:07] if you are at least trying to develop
[38:10] your own AI
[38:12] roadmap, how your how it fits with your
[38:15] company,
[38:16] um you are falling behind every day
[38:19] because I guarantee you they are all
[38:22] better at it today than they were
[38:23] yesterday and they'll be better at it
[38:24] tomorrow than they are today.
[38:29] So so for those of uh the individuals
[38:31] who are part of the webinar right now
[38:32] who are currently using Taiga. If you
[38:35] would give them um you know they they
[38:38] end the session or everyone's very
[38:39] excited about it. What's the first thing
[38:42] that they could do? Um is there a
[38:45] specific area in Taiga you would suggest
[38:48] that they look at first
[38:50] um to kind of get a feel for how to use
[38:52] the data at a Taiga
[38:54] with AI?
[38:56] Yeah uh for me uh depending on how labor
[38:59] optimization one labor I think is one of
[39:01] the biggest keys to how profitable your
[39:04] locations are. So we gauge our labor
[39:08] optimization by transaction counts. So
[39:11] to me whether you're buying a candy bar
[39:14] or buying a $80 bottle of bourbon um
[39:18] that to me takes the same amount of time
[39:20] so that requires the same amount of
[39:22] labor. That's how we do it. A lot of
[39:24] people do it labor per labor cost per
[39:27] sales dollar. A lot of people do labor
[39:30] dollar per profit dollar.
[39:34] All of that could be done within Taiga.
[39:36] So labor one you know you're only
[39:39] looking at a couple of very small data
[39:41] sets. One what's your schedule what's
[39:43] your labor schedule. Two
[39:46] what is your metric for labor
[39:49] efficiency? Is it transaction counts? Is
[39:51] it labor cost? Whatever it may be.
[39:55] You're going to marry those two data
[39:56] sets with an AI. What Taiga will give to
[39:59] you from the sales
[40:02] and whatever you how you determine what
[40:04] your labor cost is
[40:06] and then you start giving metrics of
[40:08] what your goals are. Where do you think
[40:10] is too staffed and where do you think it
[40:12] is understaffed?
[40:15] And it will start spitting out outcomes.
[40:18] The one thing I would say is you build
[40:19] that out
[40:21] do it with the mindset that I'm willing
[40:23] to learn it today, but I want to reuse
[40:26] this project a month from now when I
[40:28] have next month's data. So, think about
[40:30] how you're going to piggyback additional
[40:32] months of the data, you know, the the
[40:34] updated labor schedule, the updated
[40:36] sales, whatever it may be. You know,
[40:39] always think about how you're scaling
[40:40] that out as time goes on.
[40:44] Awesome. John, do you have any from your
[40:46] perspective, you have any
[40:47] recommendations on what they could take
[40:49] a look at first?
[40:50] Yeah, I I
[40:51] I I think it's I think it's incredibly
[40:52] useful for any sort of
[40:55] problem that is semi-ambiguous. And so,
[40:58] I I get approached a lot of times where
[41:00] our customers have, you know, like we
[41:01] want to set targets for this or we want
[41:03] we we want to better understand that.
[41:05] It is a great way to be able to help you
[41:09] go through the ideation process.
[41:12] You can pull a data set from Kaggle, you
[41:15] can pull data sets from other places,
[41:17] and then put them together using an
[41:19] agent and start asking it questions and
[41:21] help it to ask you questions back
[41:24] "Well, how how how do you see this? How
[41:27] how would you see it better? What do you
[41:28] see that's exceptions in here? What do
[41:30] you see works well? What doesn't work
[41:32] well? Are my stores very similar? Are
[41:34] they very different from this
[41:36] perspective? Give Give me different ways
[41:38] to think about this." And then you would
[41:40] just kind of continually build on that.
[41:42] You can do that process in a matter of
[41:44] minutes for a lot of different use cases
[41:47] to come up with a way to you know, plan
[41:51] out a strategy. And then from there you
[41:53] say, "Okay,
[41:55] we need to do this and figure out a way
[41:57] to make this something that we can turn
[42:00] into a playbook, something we can do
[42:01] over and over again." And that might be
[42:03] something that you're asking Jenny and
[42:05] her team at Kaggle to build you new
[42:06] dashboards that are focused on this area
[42:08] that you're going to deploy out to store
[42:10] managers to tell them, "Okay, your
[42:11] target is this and then it should look
[42:13] like this, and there's this this perfect
[42:14] chart that does this for you, and then
[42:16] we're going to, you know, watch that
[42:18] with this AI agent to make sure that
[42:20] everybody's keeping track with it, and
[42:21] it tells Alan about it." There's things
[42:23] you can do in that regard. Um the the
[42:26] other the other thing that it's really
[42:28] good for is those throwaway projects. We
[42:31] are dealing with this challenge right
[42:33] now. Maybe um you know, one thing that
[42:36] uh a lot of people approached me with
[42:38] like turn of last year was a lot of
[42:40] states were making changes to things
[42:42] like EBT and stuff like that. They
[42:43] wanted to better understand what was,
[42:46] you know, state by state, region by
[42:47] region they had that was um
[42:49] uh
[42:50] you know, uh
[42:51] able to be purchased with EBT, and then
[42:53] how people were actually purchasing.
[42:55] Ans- Answering those kind of convoluted
[42:57] deep data questions
[42:59] um are other really great ways to
[43:01] leverage these tools to kind of surface
[43:02] that up to very simple answers where,
[43:05] "Okay, I'm only going to need to do this
[43:07] once, but it's something that would take
[43:10] me me and a team hours or days or weeks
[43:13] to do."
[43:14] Um with these tools it's incredibly
[43:16] easy. You can go and look at uh you
[43:19] know, your item sales summary on Taiga.
[43:21] Pull the export of everything that's in
[43:23] these categories for this month across
[43:25] all your stores, and then plug that in
[43:27] there, and then, you know, run from
[43:28] there. Or maybe it's uh you know, um you
[43:32] know, "Tell me tell me, you know, some
[43:34] insights about, you know, whether or not
[43:36] I'm running the right type of
[43:37] promotion." So, you can put in a handful
[43:39] of promotions across uh
[43:42] uh from your promotion tracking page,
[43:44] and you know, pull the scan data for
[43:45] that. Say, "Okay, which vendors am I
[43:47] actually needing to spend the most time
[43:49] with to uh get better deals with? Or,
[43:52] you know, what deals should I approach
[43:54] them with to try to work with my more
[43:56] local vendors to see if I can't get
[43:58] better local deals." Stuff like that.
[43:59] Yeah, those ambiguous problems that Put
[44:01] is really, really good at doing today.
[44:05] Um now, tomorrow and and out in the
[44:09] future as we look at like later in 2026
[44:11] and then 2027, there are a lot of lot of
[44:14] new uh um um companies coming into our
[44:18] industry and other industries that have
[44:20] models that are already trained to
[44:21] handle these things. They just need
[44:22] access to really good data sources like
[44:24] Taiga. Uh and we're collaborating with
[44:27] those companies and then helping them
[44:29] and helping them to understand how you
[44:31] guys do business to bring you those next
[44:33] steps and the ones after that. So, uh
[44:35] like I said, today we're kind of in the
[44:36] wild west of of what's going on, but
[44:40] this stuff is rolling so quickly and
[44:42] changing so quickly
[44:44] um that um there's going to be constant
[44:47] new opportunities here. And so, you
[44:48] really have to be paying attention to
[44:50] it.
[44:51] Awesome. Thank you. So, we talked a lot
[44:55] about um AI today and I want to make
[44:57] sure that we have time for questions.
[44:59] So, again, if anybody has any questions,
[45:01] please feel free to throw those into the
[45:02] chat.
[45:04] Um Alan is very passionate about what he
[45:06] does at his organization and then also
[45:08] how he uses Taiga across various areas
[45:12] within the organization. And so, I just
[45:14] wanted to see where all this is also a
[45:16] test for me to see if I can make the
[45:18] poll
[45:20] feature work here, but um
[45:23] get the poll to work. If you all would
[45:25] be interested in doing another webinar
[45:28] um
[45:29] with um a member of Alan's team to talk
[45:31] about some of the items that I have
[45:33] listed there. Alan touched on labor
[45:35] optimization a little bit, um but we can
[45:37] also talk about how Mohawk One uses
[45:40] Taiga for loss prevention, vendor
[45:42] negotiation, waste reduction, and really
[45:44] gives um
[45:46] um
[45:47] like meaty use cases. Um if you're
[45:49] interested in that, if you could I think
[45:51] everyone should have a little plus
[45:53] button at the bottom of your chat. And
[45:55] if you want to click on that, if that is
[45:57] a topic of interest or a series of
[45:59] topics of interest, then we could we can
[46:02] plan for that. So,
[46:04] please feel free.
[46:06] Looks like we are Nice. There we go.
[46:11] answers in.
[46:12] Yeah, I was worried that they'd be like
[46:13] my wife, say no.
[46:15] Yeah.
[46:17] No, we don't want to see Alan again. No,
[46:19] no, no, no.
[46:20] We're open. He goes golfing tonight so
[46:22] that we don't have to hang out with him.
[46:27] Uh I heard I I there is a what? There is
[46:29] a no.
[46:30] I bet I bet that's Frank Horton.
[46:32] I bet I bet that's Frank Horton. I saw
[46:34] him on the list. I'll be I bet I get a
[46:36] text.
[46:39] All right. Well, as
[46:41] um it looks like the
[46:43] overwhelming answer is yes. So, we will
[46:46] get we will get Alan and his team back
[46:48] on for another session. So, we really
[46:51] appreciate you helping us out today,
[46:52] Alan. And again, if anyone has any
[46:54] questions, please drop those in.
[46:58] And um we'll give just a few minutes. I
[47:01] can turn my turn my camera off for a
[47:04] minute if anyone
[47:05] Yeah.
[47:05] Alan
[47:06] Well, regarding those questions, Janie,
[47:07] the one the one thing I would tell
[47:09] people is I mean, I think it can get
[47:12] very intimidating and I think that some
[47:16] of the cautionary tales that we were
[47:18] kind of talking about earlier in this is
[47:20] something to be aware of,
[47:22] but
[47:24] the the good side of this, like
[47:26] I mean, it's it it never it blows my
[47:29] mind almost every day. I it is just
[47:31] crazy what we're doing with this today.
[47:34] And I mean, we are literally today only
[47:37] limited by
[47:39] AR creative
[47:41] creativity on how we can utilize it and
[47:45] two, my limited skill set and trying to
[47:49] use AI. So,
[47:51] you know, it really is a thing where you
[47:53] just kind of dive deep into it and kind
[47:55] of mess with it and you're kind of you
[47:58] know, you're learning how to use AI as
[48:01] AI as your AI is learning who you are.
[48:05] And
[48:08] you know, I'd get a lot of I get this
[48:10] I've got my wife and all my kids on pro
[48:13] pro accounts. I'm making them do it.
[48:15] They are so tired of me talking about
[48:17] AI, I assure you.
[48:18] But, same thing like it's a skill set
[48:21] that you develop. And
[48:24] the
[48:25] you know, whether it be them or other
[48:27] people, and they'll say, "Well, AI was
[48:28] wrong." And yeah, AI is wrong, but it's
[48:32] not perfect 100% of the time. But,
[48:35] think about why did they get it wrong?
[48:37] Was it where did you just ask the wrong
[48:40] question? Did you give it the wrong
[48:42] data? And if it was wrong, don't just
[48:44] assume that you know, don't throw your
[48:45] hands up in the air and say, "Well, this
[48:46] isn't working." Figure out why it broke,
[48:49] and then say, "Okay, here's how I'm
[48:50] going to make sure he doesn't break like
[48:52] that again."
[48:53] Those concerns and those spot checks on
[48:56] what we're doing with various projects,
[48:59] I'm guessing I would find something
[49:02] the output of what we were looking at
[49:04] was wrong, I bet 15-20% of the time when
[49:07] I
[49:08] 2 months ago, 3 months ago as we were
[49:10] kind of getting into it.
[49:12] It's probably been a month since I saw
[49:13] anything on the reporting side of this
[49:16] come wrong. And it's because, you know,
[49:18] we kind of found
[49:20] you know, it would confuse Let's just
[49:22] take this food service one. It would
[49:24] confuse our cheeseburgers with our
[49:26] double cheeseburgers. It confused our
[49:28] cheeseburgers with the sliders. It would
[49:29] confuse this location with this
[49:31] location. You know, there were multiple
[49:33] places it screwed up. And each time we
[49:36] would correct it, and then make sure you
[49:38] understood what it is, and don't do that
[49:40] again.
[49:41] To the point now that I mean, it the AI
[49:44] knows more about our food service
[49:45] programs and what we are, what we sell,
[49:48] what we're trying to accomplish better
[49:49] than I do because it's it's constantly
[49:52] getting pumped in.
[49:53] Here's what we're wanting to do. Here's
[49:55] what we're wanting to do. Similar to
[49:56] like what John was saying earlier, where
[49:58] it's like, okay, stop working. I'll be
[49:59] back in an hour type of thing.
[50:06] Awesome. Thank Thank you, Alan. Um we
[50:08] have had a couple questions come in. So,
[50:10] the first one is more of a general Tiger
[50:13] question. Um
[50:15] somebody has asked, what is the What is
[50:18] the first thing that you check in Tiger
[50:19] each day, Alan? When you
[50:21] So, I've got The one thing I really like
[50:24] about um
[50:26] Tiger is you got all these different
[50:28] sets uh different places you got the
[50:30] information. And you got your And way of
[50:34] doing it, but then I can arrange the and
[50:37] filter the columns and the dates to what
[50:39] I like and then save that view. So,
[50:42] every morning I look at two things. I've
[50:44] got the sales report that shows
[50:47] the total inside sales and total
[50:51] transactions, total loyalty percentage,
[50:53] and gallons sold by company for that
[50:56] day. How did that compare for that same
[51:00] day previous week? And where is that one
[51:02] trending?
[51:04] That is That gets pulled up almost every
[51:06] day. And then the other one is how did
[51:09] my food service perform? Where is the
[51:11] sales at that?
[51:12] that I figured that was one of your top
[51:13] morning ones with the food.
[51:14] Yes.
[51:15] Those Those two I do every day.
[51:18] You're like, I'm not getting into food
[51:20] service right now. It's uh
[51:21] Oh god, I know. Well, you can thank
[51:24] Hey, hold on.
[51:26] That's I blame Frank Orton on that,
[51:28] actually. If he's still on here, he's
[51:29] He's the one who told me I need to get
[51:31] into food service.
[51:32] But
[51:33] You guys are doing a great job at it.
[51:35] Yeah, thank you. The
[51:38] uh then third I don't do on a daily
[51:39] basis, but tracking uh So, we are big
[51:43] big into self-checkout. So, how What is
[51:47] the percentage of my total transactions
[51:49] on self-checkout by location on a daily,
[51:51] weekly, monthly basis? And is that
[51:53] trending up or up or down?
[51:55] Uh for us, a path to labor efficiency is
[52:00] self-checkout usage.
[52:04] All right. Very nice. Awesome. Um
[52:08] we also have a question. So,
[52:12] you're looking at all this information
[52:13] at Tyga. What is Tyga doing to make sure
[52:16] that the data is actually is accurate
[52:18] that that's being um
[52:20] you know, that's being pulled in? Maybe
[52:22] John, I think that one for you.
[52:23] Yeah, I'll handle that and then then
[52:24] then I'll answer Travis' question from
[52:26] Phillips, too. Um
[52:28] So, um so so Ty- Tyga's doing a lot of
[52:31] things and some of those are are AI
[52:33] native things and some of those things
[52:34] are are um just, you know,
[52:37] basic things built into our pipeline
[52:39] from how we do stuff. Um
[52:41] so, you know, everyone's familiar with
[52:42] the connector all your sources, your
[52:44] point of sale, which is your back
[52:45] office, and those sorts of things in
[52:46] near real time, pulling transactions
[52:48] that are flowing in, pulling information
[52:50] from your back office, your labor
[52:51] system, your loyalty programs, and those
[52:53] sorts of things, and then bringing it
[52:54] all together. Um but, you know, that's
[52:57] just the most basic level of okay, we
[52:59] get that information. There are layers
[53:01] and layers and layers of different
[53:02] systems in place that are checking and
[53:05] making sure that that is all gelled
[53:07] together. And so, we have all sorts of
[53:09] tools that are doing things everything's
[53:11] from like the most basic things like,
[53:12] you know, automatically category logging
[53:14] everything that comes in to your to the
[53:16] next standard. So, if you go and you're
[53:18] talking to your study group partner or
[53:19] somebody in
[53:20] uh at looking at an Excel file or
[53:22] something like that, you can see apples
[53:23] to apples, and that sort of stuff. But,
[53:24] also cleaning up discrepancies between
[53:27] you know, the the this um
[53:29] um uh this salad that Alan's selling at
[53:32] at this store is on this UPC, but it's
[53:34] the exact same item over here, but it's
[53:35] selling on a slightly different one
[53:36] because it's on a different POS, or it's
[53:38] coming from
[53:39] uh you you an older price book record or
[53:41] this or that and the other. Making all
[53:42] that stuff make sense all the way up to
[53:45] in understanding those real
[53:47] discrepancies and actual data that we
[53:49] bring to you and say, "Oh, okay, like we
[53:51] actually have this issue here that, you
[53:53] know, we're going to have to help you
[53:54] correct in your pricebook or in your
[53:56] fuel costing or this or that and the
[53:57] other." So, lots of systems underneath
[54:00] the layers of systems under there that
[54:02] are constantly checking and rechecking
[54:04] your data to try to help you to be to
[54:07] have clean data.
[54:08] Then, you have all of our dashboards and
[54:11] reports and stuff like that. And those
[54:12] are those what we call certified data
[54:13] sets. And so, you go to the product
[54:15] scorecard, the the you know, the list of
[54:17] transactions there, that is a true set
[54:19] of what your transactions are or
[54:21] whatever you're whatever you're looking
[54:22] at this store or that store or this
[54:24] group or that group.
[54:25] Same thing with item summary or or sales
[54:28] per labor hour, all those metrics are
[54:30] already trued up by the time you get to
[54:32] them.
[54:33] And then, when somebody like you or Alan
[54:36] goes and says, "Okay, I want to take
[54:37] that and I want to load that into an
[54:38] agent." You can be assured that that is
[54:40] 99.95
[54:42] whatever percent accurate,
[54:44] that that's what your real data is up to
[54:46] that moment. So, we put lots and lots
[54:49] into that to make sure that you've
[54:50] always got really good data.
[54:54] I guess moving on to Travis's question
[54:56] here. Um
[54:57] Yes, please.
[54:58] So, Travis is at Phillips. He's
[55:01] um
[55:02] working in AI over there and you know,
[55:04] he asked uh
[55:06] um you know, working through the basic
[55:08] foundations of prompting AI with the
[55:09] four Ws and how.
[55:12] Do we leverage it for outstanding tasks
[55:14] prior days, prior weeks and stuff like
[55:15] that?
[55:16] Alan, I'll let you answer this from your
[55:18] perspective, too, but from my
[55:19] perspective, yes, we we do for that, but
[55:22] kind of talk
[55:23] to specific how why of that. Obviously,
[55:26] for anybody that doesn't know the four
[55:27] Ws, they're the the who, what, where,
[55:29] why and then how.
[55:31] So, basically, what Travis is talking
[55:34] about there is the whole process of
[55:35] designing intent. Uh so you're talking
[55:38] to um your AI agent, you're trying to
[55:40] teach it, you know, how you want it to
[55:42] do something, uh what, where, when, and
[55:45] why, okay? Um and like Alan I've say it
[55:48] said this whole time, you know, trying
[55:49] to, you know, talk to it in the same way
[55:51] that you would train or coach new
[55:52] employee or an existing employee that
[55:54] you're trying to teach skills to and
[55:55] making sure that you're preparing them
[55:57] for the long term.
[55:59] Um that part is kind of the fundamental
[56:01] part of what everybody's doing with
[56:03] working with AI, trying to train it,
[56:05] teach it to do different things. Um this
[56:07] is the second part of what Travis is
[56:09] saying there is
[56:11] what about out uh finishing up,
[56:13] buttoning up, leveraging it to do tasks
[56:17] from prior days that needs to clean up.
[56:19] I do do a lot of that, um but I only do
[56:23] that in areas where I feel confident
[56:26] that the who, what, when, where, why,
[56:27] and how is clearly defined out with that
[56:31] agent where it already knows that ahead
[56:32] of time. And so uh with my agents I work
[56:36] run through uh
[56:37] at the beginning of every session I do a
[56:40] um typical uh
[56:42] uh scrum style stand-up with them to
[56:44] make sure that everything is primed in
[56:46] the way that I want it and then run
[56:47] through a series of tests to make sure
[56:49] that it is prepared to either do any new
[56:51] task with me or button up anything that
[56:53] I had from the prior day.
[56:55] Um and so that way I can make and ensure
[56:57] that I know that I'm confident that it
[56:59] can move on to those tasks. So like what
[57:01] Jenny was saying about like when we're
[57:02] at the show a couple weeks ago and
[57:05] um
[57:06] I was working on a few things and I told
[57:08] it, "Well, I'm going to be, you know,
[57:09] off in a in a session now, so I don't
[57:11] want you to work on this, that, and the
[57:13] other thing, but it's okay for you to do
[57:14] this and this and this." Um that's
[57:16] that's where I'm making those managerial
[57:19] decisions with it. And so that's why uh
[57:22] earlier I was saying, you know, treat
[57:24] this as a management relationship. Um
[57:27] and you'll be successful with it. Hope
[57:30] that helps.
[57:32] Yeah,
[57:33] just kind of going back to that original
[57:34] project where we pulled the screen of
[57:36] John for the food service that who the
[57:37] four Ws.
[57:39] So like for that project specifically,
[57:42] you know, every well, Claude Gemini
[57:44] basically work the same and I suspect
[57:46] they all do. You know, you have your
[57:48] your your cloud drive to store your data
[57:52] and then you got the instructions that
[57:54] you're supposed to get it. So almost
[57:56] always at the very at those
[57:58] instructions, I start with "Hey, this is
[58:01] Well, for that one, I want to track the
[58:03] performance for each menu item in our
[58:05] Kate's Kitchen and Little Caesars
[58:07] category for mock one. So now it knows
[58:09] I'm trying to track for Kate's Kitchen,
[58:11] Little Caesars, and it has the mock one
[58:13] and my website. So now it knows exactly
[58:18] who is who's looking for it. Then I take
[58:21] the next section to describe exactly
[58:23] what files I'm going to give it and what
[58:26] those files mean. So there's a series of
[58:28] lines basically explaining the data
[58:30] sets.
[58:31] Then
[58:33] I have a third file that basically
[58:37] has the name of the menu item that Tega
[58:40] uses, the name of the menu item that Day
[58:42] Code Genie uses cuz those names don't
[58:44] necessarily match up. So let's say it
[58:47] may be cheeseburger in Tega, it may be
[58:50] something cheese some version of
[58:52] cheeseburger in Day Code Genie. I say,
[58:55] "Okay, our cheeseburger's named this in
[58:56] Tega, named this in our production file.
[59:00] Here is my standard retail, and here is
[59:04] our menu cost, and here is that shelf
[59:07] life." So it now knows what Tega calls
[59:11] it, it knows what my production file
[59:12] calls it, it knows what the regular
[59:15] price is. It can show it when we're
[59:17] doing various employee or other types of
[59:19] discounts. It then knows what the shelf
[59:22] life is for this. Once you kind of get
[59:25] all that structured into that, now you
[59:27] got the who, you got the what, the uh
[59:31] I mean, we've gotten to the point now
[59:32] where automating our build tastes. And
[59:34] what it is is each one of those menu
[59:36] items, we have the minimum we want to
[59:39] put out there and the maximum. So, let's
[59:41] just say for cheeseburger example, and I
[59:43] want to have at minimum two out there at
[59:44] all times that we're going to sell it.
[59:47] At most, I want to have 10 cuz maybe
[59:48] that's the most I can fit. Whatever
[59:50] those two numbers are. Anywhere in
[59:52] between that, I want to target a 15%
[59:54] waste.
[59:55] You get my build to where I'm going to
[59:59] have enough out there that will give me
[01:00:00] a 15% waste or 10% waste. You know,
[01:00:02] we've actually played with those target
[01:00:04] waste numbers. But, as long as it knows
[01:00:08] what you're willing to go low and
[01:00:10] maximum and and hey, let's try to get
[01:00:13] this to this waste percentage,
[01:00:16] you've kind of given us guidelines.
[01:00:23] Thank you. Thank you, Alan, and thank
[01:00:25] you, John. Um we are 1 minute from the
[01:00:27] top of the hour, so I want to be
[01:00:29] cognizant of everybody's time. Um if you
[01:00:32] guys have follow-up questions, again,
[01:00:34] you're welcome to put them in the chat
[01:00:35] or you can I believe here we've got some
[01:00:38] information,
[01:00:39] um a QR code that you can use or you can
[01:00:42] call, text me, send me an email, um
[01:00:45] myself or John, and we will we'll get
[01:00:48] back to you. And again, thank you so
[01:00:51] much, John, and thank you so much, Alan,
[01:00:53] and we look forward to um
[01:00:55] continuing the conversation here on the
[01:00:57] next webinar.
[01:00:59] Thanks, Alan. Thanks, Danny. Thanks,
[01:01:01] John.
[01:01:02] Thanks, everybody.
[01:01:03] Take care, everybody. Have a great day.