# Next-Gen Finance: The Evolution of Fintech | 2026 Cornerstone Conference

https://www.youtube.com/watch?v=nXm7mRcsYpg

[00:02] Okay, good morning everybody.
[00:06] So, my name is Margareta McGrath.
[00:09] I'm the advisory and strategy lead at Dell Technologies and I'm delighted to be moderating this session this morning.
[00:17] So, thank you all for joining.
[00:19] We're going to have a really engaging and interesting session.
[00:22] But before we begin, we'd like to invite you all to submit questions on the conference app.
[00:27] So I think you've all been probably doing that so far.
[00:29] But if you've got questions, this is a really engaging session on FinTech.
[00:33] We'd love you to put in your questions.
[00:35] Scroll down to the plenary session number six and type in your question.
[00:40] Okay.
[00:42] So today, fintech is no longer a category within financial services.
[00:47] It is financial services.
[00:51] From tapping your phone to pay to checking your bank account or paying a bill online to filing an insurance claim, technology is now embedded in every single financial transaction that we do.
[01:02] And now layered
[01:06] Onto all of this is AI, artificial intelligence.
[01:11] We're not just digitalizing the finance, we are fundamentally transforming it.
[01:17] The systems that move money around the world are in many cases a century old.
[01:22] But today they are being asked to support real time AI-driven highly personalized financial ecosystems.
[01:32] Trust in these ecosystems is been fully tested.
[01:38] Questions around data bias, cyber security are no longer theoretical.
[01:43] They are immediate and they are global.
[01:45] So today, this morning, I'm joined by four extraordinary leaders who are helping shaping this transformation and I call it actually a rupture rather than a transformation to use Mark Carney's language from every different perspective.
[02:00] Over the next 45 minutes, our panelists will discuss the opportunity to create more efficient, personalized, and inclusive financial
[02:06] Ecosystems, as well as addressing the risks that come with moving too fast or without the right guard rails in place.
[02:14] Our first panelist is Jess Calvin, global head of cyber security at city.
[02:22] In her role, she oversees threat detection, intelligence, and incident response to protect the firm's global infrastructure 247.
[02:31] Since joining city, Jess has driven a major transformation of cyber operations and vulnerability management.
[02:38] Prior to this, Jess served as the global head of vulnerability and chief finanformation information security officer for Europe and the Middle East and Africa at JP Morgan.
[02:50] Welcome, Jess.
[02:52] And thank you for getting up so early on the trains.
[02:54] Thank you so much.
[02:58] Also with us this morning is MV Dan Hukar, executive vice president and global head of operations at EXL.
[03:04] MVI oversees operations, finance, and
[03:07] Accounting across EXL strategic growth units.
[03:10] Her work focuses on reimagining operating models and integrating advanced AI capabilities to drive innovation.
[03:20] Over the course of her career, she has held senior leadership roles at Vodafone, Northern Trust, State Street, and Principal Financial.
[03:31] Huge welcome to you, MV.
[03:34] Delighted to have you here.
[03:37] Our next panelist is Kungju Hu, Korea representative to the G20 Empower Alliance and a global leader in financial services with more than 30 years of experience across the Asian financial sector.
[03:52] Her experience spans corporate strategy, digital innovation, and international relations.
[03:57] Kungju is recognized a woman leader of the year by the Asian Insurance Awards.
[04:02] She continues to shape global policy through her work with the G20 empower and APAC.
[04:06] She's also the founding
[04:09] President of IWS Korea.
[04:13] Huge welcome to you Kungju.
[04:18] And our last panelist, we're joined by Priyanka Raj Kumar who's senior vice president of payment services and AI transformation and strategy at EXL.
[04:29] Priyanka has been instrumental in driving AIEL transformation, delivering nearly $3 billion in annual savings for major US healthcare payers.
[04:41] With 25 years of global experience across analytics, product engineering and strategy, Priyanka oversees AI innovation, product development, and market growth, helping organizations scale digital and AI transformations.
[04:58] Huge welcome to you pri.
[05:03] Okay, so please do ask your questions on the app everybody.
[05:07] We really want to have an engaging panel conversation here.
[05:09] And we're going to get into some of these questions already.
[05:13] So Jess, if I can start with you, let's talk about the underlying systems that are supporting much of our financial transactions today.
[05:19] Much of them are built on legacy and old infrastructure.
[05:24] I know this too well from my role at Dell and there many of them are over a century old.
[05:29] So we're laying on AI, real time payments and increasingly complex digital ecosystems at rapid speed.
[05:37] So what do you see are the biggest gaps that are out there right now between legacy systems and the next gen of tech?
[05:44] Thank you so much for the question and it's great to be here.
[05:49] Um, I think what we're trying to do is we're layering 21st century technology on top of 20th century infrastructure.
[05:56] Um, and there's a few major challenges with that.
[06:00] The first one is, you know, purely operational.
[06:03] We're from financial services, we're very used to traditional operations like batch processing which happens overnight over weekends.
[06:09] But now we're in a world where
[06:12] Our customers and our industry wants real time access to systems 24 by 7.
[06:19] And the architecture and the data, it's just not architected that way.
[06:23] The other aspect of this is resilience.
[06:25] You know, traditional um systems, financial systems, they're built for stability.
[06:30] We want our systems to be up.
[06:31] We'll have maintenance windows.
[06:33] We'll bring them down at a certain time, patch them, reboot them, and good to go.
[06:38] Tie that back in with 24 by 7 requirements for operations.
[06:43] It needs to be up all of the time, real time.
[06:45] So that resilience again goes back to a rearchitecture of how we need to have the systems operate.
[06:51] Then we come to, you know, stability.
[06:54] Um, the way that we develop our our systems and our applications traditionally have been maybe we'll do a quarterly release with new features, with new capabilities, things our customers want to see.
[07:06] Now we need to be releasing in some cases multiple times a day, and that's a fundamental change to how we would.
[07:12] Develop those systems and how we would work those systems.
[07:16] The final piece I'll touch on is identity.
[07:19] Traditionally humans are accessing systems.
[07:22] Now we've got agentic AI accessing systems.
[07:25] It's no longer just a humanled identity.
[07:27] The access methods are no longer human to system.
[07:32] We're looking at APIs which fundamentally changes the attack surface that we have.
[07:37] So all of those combined with, you know, acceleration requirement for 24 by 7 continuous uptime availability data means that organizations are faced with a complete transformation of how we architect, produce, and support, and that brings risk.
[07:53] It brings data risk.
[07:55] It brings stability risk, and it brings operational risk, you know, and those are the transformations that financial services have been on for quite some time now, you know, a couple of years, but that's really been evolution.
[08:04] Yeah, I would say it's not even just financial services; I think it's any legacy organization right now with underlying, you know, old school technology and infrastructure.
[08:10] It's a...
[08:14] Huge journey to get this right.
[08:15] That's right, and I'd be remiss if I didn't say the cyber threat obviously changes with legacy systems as well, but I'll touch on that in a moment.
[08:22] Brilliant.
[08:22] Kungju, we can touch on you.
[08:25] So, Korea has one of the most advanced financial systems in the world.
[08:30] How is fintech evolving differently over there?
[08:33] And what can we learn and others learn around the world from Korea?
[08:37] First off, I wish to congratulate Idol UK for this most successful conference.
[08:43] When we talk about Korea on the world stage, the conversation usually starts with K-pop, K beauty, K drama.
[08:52] So, um, I want to thank you for bringing K finance to attention.
[08:57] So Korea's digital finance is driven by world-class ICT infrastructure.
[09:00] We have one of the fastest speeds in the world and 95% penetration of smartphones.
[09:08] Deregulations and competition between local tech companies were also big drivers of growth.
[09:14] I know that a lot of
[09:16] People in this room today are on either Google or Apple, but Korea has its own platform, its own ecommerce giants and infrastructure as well.
[09:27] What we did was we connected them into an integrated ecosystems.
[09:33] So the result is what I called super app.
[09:36] You can book a taxi, send gifts to your friends, do financial transactions in one app in a few minutes.
[09:41] You do not have to switch apps and you log in and out of different apps.
[09:49] So, Korea has built a super app where finance is seamlessly embedded into everyday lives of the people.
[09:55] Another example I want to mention is a project that I launched and I call it um alternative credit scoring model.
[10:03] I used customer purchase data of a major bookstore and integrated this with a digital bank to provide lower interest rates to the bank customers.
[10:12] But of course, I had to go through regulatory approval and a process that was not
[10:16] Neither quick nor simple.
[10:19] But here's what I think is the most important insight.
[10:24] The real secret is not the technology.
[10:27] It's the infrastructure underneath.
[10:30] Korea made early investments in digital identity.
[10:33] It built bridges between fintech and legacy financial institutions.
[10:39] And perhaps most importantly, it created regulatory sandboxes.
[10:43] So what is the lesson learned?
[10:47] Digital finance works best when it is part of an ecosystem and not just a standalone app.
[10:52] That's incredible just to hear the ease in Korea.
[10:55] Yeah.
[10:58] But really quickly on this one, how do you find coming back over to the UK and other places dealing with legacy apps and stuff like that when Korea you've got it all integrated?
[11:09] Is it much easier to engage?
[11:12] It must be a big step backwards, right?
[11:17] Well, you know, different experiences all has advantages, pros and cons, but nice answer what we're doing right now.
[11:26] Priyanka, as AI becomes more embedded in financial systems, one of the biggest questions is, can we trust it?
[11:34] Can we trust it managing our mortgages?
[11:36] Can we trust it making investments?
[11:38] How should we think about building systems for trust, particularly in highly regulated industries?
[11:45] Okay, so trust is a big word.
[11:48] So, let's just unpack it, right?
[11:48] Trust in humans is very different from trust in AI.
[11:53] But humans, it's more a feeling.
[11:55] It's a belief.
[11:55] I trust you.
[11:56] Do you trust me?
[11:56] There's an assumption going on there.
[11:59] Um, you kind of like trust a model.
[12:01] When it comes to a machine, it's about a system.
[12:03] It's about an architecture.
[12:06] How do you design a system that has controls in it, that has transparency, traceability into it?
[12:11] And you design a system from day one that is transparent, that has controls and is
[12:17] Regulatory ready from day one.
[12:21] It's not an afterthought.
[12:24] So that's what trust means in a system, especially in a financial system.
[12:28] And what goes into building trust in these systems is some key components.
[12:33] The first one is data.
[12:35] Any algorithms that you use, they fundamentally use data to make a decision.
[12:40] So is your data right?
[12:40] Where is it coming from?
[12:43] Was it accurately represented to the model or the algorithm that is making a decision?
[12:47] So these are questions that we have to look at, answer, and make sure that the systems are handling them.
[12:55] The second one is you can't just let a model or algorithm make a decision.
[13:00] There has to be governance around it.
[13:03] What level of decision making are we allowing?
[13:05] Are there again controls and thresholds in place?
[13:08] Are there barriers?
[13:10] Are there guardrails between where decisions are being made and how they're being made?
[13:14] And the final point that I would like to make is accountability.
[13:18] The accountability of a decision lies with us the humans.
[13:20] You can't just outsource that to a machine.
[13:23] So end of the day the question is about how do we build a system with AI that has controls in it that has accountability and transparency to it.
[13:37] When we put all these layers in into a system I think we have a system that can be trustworthy.
[13:42] Yeah.
[13:42] Great.
[13:43] Great insights and thank you for that.
[13:46] Kungju, from your experience in financial services and insurance, where have you seen AI be most helpful so far and where is it still falling short particularly when it comes to trust in real world applications?
[14:00] Well, I would say as of today, AI has been effective where there is high volume data and repetitive decision making and insurance is a perfect example.
[14:10] Underwriting claims, fraud detection, product recommendation.
[14:15] AI definitely has been delivering value in
[14:19] Improving speed, consistency, and efficiency.
[14:21] It has been enabling a shift from reactive to predictive service.
[14:24] At an industry level, we're seeing the rise of personalized financial guidance.
[14:27] So, people expect financial services to be built around their data, their lives, and their goals.
[14:32] So, people like yourself, um, you've grown up with Netflix recommendations; bring those expectations into financial services.
[14:39] However, there's a gap, like you said, that technology has not yet closed, and that gap is trust.
[14:47] Trust on security, predictions, and ethics.
[14:52] From my experience, financial decisions involve people's savings, retirement, health.
[14:58] And if you look at insurance, it's about life and death.
[15:03] And these are deeply sensitive and emotional areas.
[15:05] People just don't want an answer from AI.
[15:07] They want to understand why.
[15:10] So, you just cannot give them a black box.
[15:13] What if AI makes a wrong recommendation and the...
[15:20] Wrong investments?
[15:21] Who is responsible?
[15:23] When I was in the corporate pension business, many of my clients like CEOs managed investment on their own online.
[15:29] But when the market crashed, I was the one who had to take all their complaints.
[15:35] So while AI is excellent at supporting decisions, human touch, human judgment remains essential.
[15:41] The real opportunity is building AI that's not just powerful, but transparent, fair, and accountable.
[15:50] To fully utilize its potential, we need to invest not just in technology but in the skills, ethics and responsible policies.
[15:59] And trust will be, I think, the key to future innovation.
[16:04] So the choices we make today will shape how AI impacts our future.
[16:10] Great points.
[16:11] Come you love the focus on skills and literacy and trust and governance as well with Priyanka's points.
[16:19] Mativa, how is personalization and segmentation changing the way FS and
[16:21] Financial insure institutions designing products that serve customers globally?
[16:26] How do you see that being unpacked?
[16:28] Sure.
[16:29] Um, so good morning everyone and good to be here.
[16:34] Um, just to your question Margareta, financial institutions are undergoing a fundamental shift.
[16:42] You know, we used to have this one-size-fits-all sort of approach.
[16:45] That's moving towards deeply personalized experiences and it's reshaping the industry.
[16:52] Um, you know, Kungju just referred to it as well.
[16:54] Uh, and the shift is happening both in the front end as well as the back end and I'll explain that.
[17:03] So, traditionally, um, models were, you know, focused on demographic data, right?
[17:11] So, age income brackets, geographies.
[17:14] Today, institutions use behavioral, psychographic, transactional data to build dynamic micro segments, right?
[17:21] Uh, and sometimes
[17:24] That does come down to the individual level and makes it very personalized.
[17:28] So you go from a broad demographic segment to actually a segment of one, right?
[17:35] And that's possible because of, you know, the technology we have.
[17:39] Two is product design by itself is becoming modular and adaptive.
[17:42] Rather than fixed product catalogs, financial institutions are building configurable product architectures, right?
[17:51] Um, and data and AI are the core engines making this happen.
[17:55] They do that by predicting needs, by personalizing communication, by reducing churn.
[18:05] Yes.
[18:06] Um, and then in the back end there is segmentation in risk and compliance as well that shapes underwriting and credit decisioning and, you know, other risk elements.
[18:18] So how does that happen?
[18:18] Alternative data allows institutions to serve customers that were previously underserved or entirely excluded.
[18:27] Dynamic credit decisioning updates risk profiles continuously rather than doing an annual review.
[18:32] And then KYC and fraud models right are calibrated therefore reducing false positives.
[18:41] So there is a fundamental shift both in how you go to customers as well as how you manage your back-end systems to make that happen.
[18:51] But having said that there are key tensions and challenges.
[18:54] Um I'd say privacy and consent is one.
[18:57] Customers are wary of hyper surveillance.
[19:01] Regulatory frameworks like GDPR constrain you know data use.
[19:05] Explainability you know we touched on it earlier.
[19:09] Personalized credit decisions must be explainable.
[19:12] Um three is potentially trust asymmetry.
[19:15] And what I mean by that is personalization can tip into manipulation if not governed carefully.
[19:24] Right? So you could for example target customers, vulnerable customers with high fee.
[19:28] Products.
[19:29] Yeah.
[19:29] And so that's clearly a challenge and a tension that we have to work through.
[19:33] And then um Jess mentioned it large financial institutions have legacy infrastructure, right?
[19:40] Um and modernization is expensive and and slow.
[19:44] I think the institutions winning this shift towards personalization are those that treat personalization not as a marketing tactic uh but as a co-product philosophy right um so in summary I'd say personalization has shifted from a campaign tactic to a realtime capability that guides decisions across the customer life cycle.
[20:10] Context is important.
[20:13] The right data foundation is important.
[20:15] And then I think mostly important is responsible AI controls, right?
[20:21] That can help institutions provide highly relevant, personal experiences while staying transparent, fair, ethical and
[20:30] compliant.
[20:31] Those are great insights. Thank you,
[20:33] Mativa, for sharing those.
[20:36] Jess, in April, Anthropic announced
[20:38] Project Last Wing um using a frontier AI
[20:42] model reportedly capable of finding
[20:44] thousands of zeroday vulnerabilities
[20:47] that traditional tools have missed or
[20:49] could miss. What does this level, this
[20:53] capability level mean for cyber security
[20:55] in financial services? And how can you
[20:58] use AI to defend a system that AI can be
[21:01] used to discover and exploit its weakest
[21:04] points at scale?
[21:06] Thank you for the question. This is my
[21:07] favorite topic right now.
[21:08] You would like that one.
[21:11] So, let me talk about the models first
[21:13] of all, then I'll talk about what I
[21:15] think it means. Um, and then we can talk
[21:18] about how we can use AI to defend
[21:20] against AI. So the project glasswing and
[21:23] and the and the models itself these
[21:25] models are very very powerful and the
[21:27] conversation over the past few weeks has
[21:29] moved away from one specific model which
[21:31] um glasswing was part of into the
[21:34] frontier AI models that are following
[21:36] very fast behind it. Um what we have
[21:38] discovered is these models are extremely
[21:40] good at finding vulnerabilities in
[21:42] systems in systems that have been
[21:44] rigorously tested by um subject matter
[21:47] experts and cyber experts. They're
[21:50] they're exceptional at scanning across
[21:53] significant batches of code and
[21:56] stringing together vulnerabilities to
[21:58] create a viable attack path which to
[22:00] some degree kind of blows away the
[22:03] previous model of critical high, medium,
[22:05] and low vulnerabilities because you have
[22:07] that chaining effect. So, you know,
[22:08] they're they're very legitimate and
[22:10] they're, you know, they're very exciting
[22:11] from a from a cyber security
[22:13] perspective.
[22:15] What this really means in my opinion is
[22:17] we're moving away from a very
[22:19] traditional model of
[22:22] patch release. The exploitation time for
[22:25] vulnerabilities I think several years
[22:27] ago was maybe you know 19 days from the
[22:30] point of vulnerability was released to
[22:32] the point you could expect it to be
[22:33] exploited. We're seeing that get
[22:35] narrower and narrower and narrower and
[22:37] you know we would expect that to be now
[22:39] in hours and not days. So your
[22:41] traditional I have 30 days to patch, 60
[22:44] days to patch, 90 days to patch is kind
[22:46] of those times are are really, you know,
[22:49] in the past or will shortly be in the
[22:51] past. So we're now forced to think
[22:53] differently about how we secure
[22:55] environments within financial services
[22:57] and how we can really shore up the
[23:00] ecosystem. What it also means it and
[23:02] what we're seeing is it lowers the
[23:04] barrier to entry. Previously, um, people
[23:08] that were capable of executing a kind of
[23:10] vulnerability exploit such as the ones
[23:12] we're seeing would be highly skilled,
[23:14] few and far between. These tools really
[23:16] lower that barrier, which means you're
[23:18] likely to see kind of more of that
[23:20] coming at you. That's really in terms of
[23:23] what we think this is going to mean for
[23:25] the industry moving forward. It's a true
[23:27] paradigm shift for us in cyber security.
[23:30] But all is not lost. you know,
[23:32] ultimately the financial services and
[23:33] and the tech industry is really coming
[23:35] together around this. And what we're
[23:37] seeing is the the aggregation of skill
[23:40] sets to help defend against these
[23:43] threats and specifically embedding AI
[23:46] into core cyber security processes. So
[23:48] to give a few examples, embedding AI
[23:51] into your vulnerability management
[23:52] processes means you can you can use the
[23:54] tools to detect where you've got
[23:55] weaknesses before someone else does.
[23:58] There's also some really good thinking
[24:00] around self-healing of vulnerabilities
[24:02] and using AI to help do that. So you
[24:04] don't necessarily need humans in the
[24:06] loop. It really also helps us understand
[24:09] what's our true attack surface. I
[24:11] mentioned earlier on how modern
[24:12] technologies are changing our attack
[24:14] surface. So identity is is the new
[24:16] frontier. You can use AI to help you
[24:19] understand your attack surface, but then
[24:21] also to protect against it. So embedding
[24:23] AI into core processes into your
[24:25] development processes. So you know when
[24:27] your teams are developing code, it's
[24:29] secure from the very beginning and will
[24:31] and will continue to to be that way. I
[24:34] shared this perspective um recently at
[24:36] an industry event and I I I genuinely
[24:38] believe that these AI frontier models
[24:40] will ultimately make us safer in the
[24:42] long run. Using them dayto-day will make
[24:46] our ecosystem safer. It's going to take
[24:48] us time to get there. This isn't an
[24:50] overnight thing. Um but fundamentally I
[24:52] think they're a very powerful force for
[24:54] for financial services.
[24:56] Great to hear. Okay Priyanka if we can
[24:59] come to you just building on Jess's
[25:01] points around human in the loop when you
[25:03] need them at different times this the
[25:06] relationship between AI agentic AI and
[25:09] humans.
[25:10] Do we always need a human in the loop?
[25:12] Um are there areas that we can safely
[25:14] allow machines and agentic AI to to
[25:17] drive and govern?
[25:18] Got it. This is a question that I face
[25:21] every day both at work and at home.
[25:24] Luckily at home it's easier. The
[25:26] question is do I let my son outsource
[25:28] his homework to AI and get it done
[25:33] completely outsource the accountability
[25:35] to AI? So the answer is no.
[25:40] So in a system uh before we make that
[25:44] decision whether human is needed or not
[25:46] we have to understand how AI makes
[25:49] decisions. So fundamentally AI is an
[25:52] algorithm and they work on patterns of
[25:55] data. They understand patterns and they
[25:58] look at that patterns use statistics
[26:00] probability to make a decision saying
[26:02] this is the most probable outcome. AI
[26:05] does not have a moral understanding. It
[26:08] does not understand empathy. It does not
[26:10] understand fairness. So these are the
[26:13] areas where human is needed. So when you
[26:16] really think about should an AI make a
[26:19] decision or should a human make a
[26:20] decision, you really have to look at the
[26:23] value of human judgment in that
[26:26] particular decisioning transaction. So
[26:29] if it's a lowrisk transaction, if a
[26:32] decision can be reversed very easily,
[26:34] then AI would be much better at it. For
[26:37] example, if you have to look at a
[26:39] thousandpage contract, look for clauses
[26:42] and then run data against it, create a
[26:44] report, create a recommendation and send
[26:46] it to the business, AI is probably much
[26:49] better at it because they are efficient.
[26:51] They don't get tired and they reduce
[26:53] operational error. But if you look at
[26:56] the decision about if should you deny a
[26:59] claim an insurance claim a financial
[27:01] claim a loan then this has a much bigger
[27:03] impact and this is where the human needs
[27:06] to be made accountable. They are in the
[27:08] decision making process. So um overall I
[27:12] think that we don't need a human in the
[27:14] loop in all transactions but the human
[27:16] should control the system.
[27:19] So
[27:19] yeah thank you for that. come to you.
[27:22] Would you like to build on that and what
[27:23] are your thoughts around this?
[27:25] Well, it's a question that I think
[27:26] resonates with many people today. Um, is
[27:29] it going to be human versus machine or
[27:32] human plus machine? Will this lead to
[27:34] utopia, dystopia? But I think it's worth
[27:37] to step back and realize that we're
[27:39] living in a period of coexistence. And I
[27:42] think that that coexistence itself
[27:44] requires careful design. It requires
[27:47] people to understand what the machine is
[27:49] doing and what why is it doing it. So
[27:51] the fundamental question I want to ask
[27:53] here is are they using transparent data.
[27:57] Um gender bias is one of the clearest
[28:00] and most persistent problems in AI data
[28:02] today. It's learning from historical
[28:05] data and that data often reflects
[28:08] longstanding
[28:10] social inequalities. We need to audit
[28:13] the data sets build fairness checks into
[28:15] the models. And the next question is who
[28:17] will do the work? Do we have enough
[28:20] women representation in the STEM field
[28:23] and STEM education? Unless AI is
[28:26] carefully designed, monitored and
[28:28] supported by greater representation of
[28:32] women, existing inequalities will be
[28:35] reinforced rather than reduced. So to go
[28:38] back to your question I'd like to say
[28:40] that as of today there are areas where
[28:42] human outperform machines and equally
[28:44] areas where machine can outperform
[28:46] humans but in any case human should take
[28:49] on the role as orchestrators and
[28:52] validators and in both cases I think the
[28:55] fundamental question is the same who is
[28:57] in control and who is responsible for
[29:00] the outcome going forward we must
[29:03] prepare for the future of AGI and
[29:05] eventually ASI as Well, but I think the
[29:09] other challenge that we face here is
[29:10] that the speed of the technology is much
[29:13] faster than the regulation and the way
[29:16] we respond to these changes. But through
[29:19] all of this, I want to come back to one
[29:21] fundamental point that the goal should
[29:23] never be human versus machine. It should
[29:27] be human empowered by machine.
[29:30] Thank you, Kungju. Great to hear that
[29:36] Mativa as financial services
[29:39] increasingly rely on AIdriven
[29:41] decision-making bias be is becoming more
[29:43] and more of a critical concern. How do
[29:46] you see financial institutions fintexs
[29:48] ensuring that AI models are not just
[29:51] reinforcing these old age biases
[29:54] particularly with credit decisions
[29:56] insurance pricing and actuarial
[29:58] modeling? Sure.
[30:02] Um I think that's a very pertinent
[30:03] question and probably one of the most
[30:06] consequential you know challenges and
[30:09] applied artificial intelligence today.
[30:12] AI systems traditionally learn from
[30:16] historical data which often reflects
[30:19] social and economic inequality and when
[30:23] deployed these biases can be exported at
[30:25] scale. Right. Um there are several
[30:29] measures that instit institutions are
[30:31] taking um to mitigate bias. Uh excl you
[30:35] know since we are a data and AI company
[30:38] um some of the measures and you know
[30:40] that we're taking to make sure that
[30:41] these biases don't creep into our
[30:44] systems are we treat bias as a design
[30:48] risk.
[30:50] Define fairness objectives, protected
[30:52] classes, uh, no-go attributes before
[30:55] modeling begins.
[30:58] Then build on AI ready data, right? Um,
[31:02] strong data lineage becomes very
[31:04] important. Documenting feature rationals
[31:07] becomes important. Um, and that helps,
[31:10] you know, reduce the hidden proxies that
[31:12] might lie in that data. operationalizing
[31:15] responsible AI by establishing very
[31:18] clear accountability model inventory
[31:21] approval gates you know audit ready
[31:24] documentation
[31:26] I think running pre-eployment fairness
[31:28] testing including sensitivity uh stress
[31:31] tests to validate outcomes and not just
[31:33] the accuracy of the model um you know we
[31:37] talked about it ensuring there's a human
[31:38] in the loop
[31:40] right um high impact decisions like
[31:42] credit decisions insurance pricing
[31:44] actural modeling with clear overriding
[31:48] exception handling and appeal mechanisms
[31:52] making decisions explainable and
[31:53] traceable I think is a responsibility to
[31:56] enable jur justification of outcomes to
[31:59] customers regulators audit I think
[32:02] continuously monitoring performance
[32:03] becomes important as well um you know
[32:07] the the AI could drift um so making sure
[32:10] that those KPIs are you focused on you
[32:14] know making sure those KPIs actually
[32:15] don't drift um and then setting
[32:18] automation boundaries you know deciding
[32:21] what the human can do what the AI can do
[32:24] uh what requires a second line review
[32:26] etc. Um, one of the other, you know,
[32:30] areas that's been used very widely is
[32:32] using synthetic data.
[32:34] So, you actually don't use actual data,
[32:36] you use synthetic data and train the AI
[32:38] model allowing for the creation of
[32:41] balanced data sets that represents all
[32:43] groups fairly and minimizes bias. So I'd
[32:46] say from an EXL perspective, we embed
[32:49] fairness, governance, risk, compliance,
[32:53] human oversight right up front into the
[32:56] operating model such that the outcomes
[32:58] are faster, better, demonstrabably fair.
[33:01] Uh, and AI models can expand with
[33:03] inclusive access instead of scaling in
[33:07] equity.
[33:08] It's a it's a lot, right? when you
[33:10] brought us through that just hearing
[33:11] about AI ready data fairness testing
[33:14] there's a lot for organizations in terms
[33:16] of getting their core foundational data
[33:19] right and make sure that bias is
[33:21] eliminated but it's constantly a journey
[33:23] right
[33:24] oh absolutely and you have to keep
[33:26] retraining the models
[33:27] there's data there's more data that's
[33:30] been produced in the last two years than
[33:32] in our lifetimes prior to that so
[33:35] there's just so much data and therefore
[33:37] data is the foundation it becomes very
[33:39] important. You know, the thing with
[33:41] technology is it does not remove bias.
[33:44] It operationalizes it unless it's
[33:47] continuously addressed.
[33:48] That's right.
[33:49] And and that's what institutions need to
[33:51] keep doing. That's what um you know,
[33:54] companies like EXL keep working on.
[33:56] Yeah. Thank you for that. So, we're
[33:59] we're quickly running up on time. We do
[34:02] want to move to some questions and I
[34:04] know we've got some questions coming in.
[34:06] So, we've got the first question here
[34:07] and it's a brilliant question from Nina
[34:09] from our IWS Finland team. So, welcome
[34:13] Nina. How do you see blockchain and
[34:15] cryptocurrencies changing the FS sector?
[34:18] Jess, can I hand this one to you? Yeah,
[34:22] it's a great question. the it's it's
[34:25] another opportunity for the financial
[34:27] services sector to expand their
[34:29] offerings to to customers provided that
[34:32] there's the appropriate governance
[34:34] oversight and obviously security um
[34:36] around it. It's certainly something um
[34:39] from my perspective that provides us
[34:41] with an interesting security lens on how
[34:44] to secure crypto and and secure um
[34:48] cryptocurrency
[34:49] um on the on the on the city angle. Um
[34:52] but you know ultimately I think it it
[34:54] provides another opportunity for us to
[34:56] to offer those modernized services that
[34:58] we've been talking about today.
[35:01] Any any thoughts from the rest of you on
[35:04] blockchain?
[35:05] Want to build on that?
[35:06] Uh I think that technology is right
[35:10] right now. We are seeing that
[35:12] governments are now embracing this. We
[35:14] probably we will also see a move to
[35:17] digital currency. We're moving away from
[35:19] the fiat currency. We were earlier in
[35:21] the commodity gold and now we've moved
[35:23] to fiat currency which is what we have
[35:25] right now and eventually we'll start
[35:27] moving more towards digital currencies.
[35:29] So that's right that's how I see the
[35:31] shift happening. What a time, right?
[35:33] Crypto and AI together. Cool.
[35:35] Blockchain.
[35:36] Very powerful.
[35:38] Very powerful time to Yeah. Okay. We've
[35:40] got a question coming in from one of our
[35:42] fellow alumni, Anna. Wonderful to have
[35:44] you here, Anna. How does a super app
[35:46] work in financial transactions across
[35:48] different providers? I think this one
[35:50] was deliberately made for you, Kungju.
[35:52] So, over to you. Well, honestly, I'm not
[35:55] a regulatoratory person, so I can't
[35:57] really speak from a regulatory point of
[35:59] view, but um Korea has adopted my data
[36:02] and open banking, but it took um a very
[36:05] very long time to get all of this
[36:07] prepared. So um any customer can open up
[36:11] their app if they approve of every
[36:14] single um approval process by different
[36:18] financial institutions on your app. You
[36:20] can um do transactions with all of the
[36:23] other banks that uh you have accounts
[36:26] with. At the same time you can also uh
[36:29] through my data evaluate not just bank
[36:31] information but securities and insurance
[36:34] as well. So in one financial app um you
[36:38] can have one account in one maybe bank
[36:40] or insurance or securities and you can
[36:42] check all of the other um or even do
[36:45] transactions within that one app. But
[36:47] like I said that was um a very rigorous
[36:51] uh procedure and um it was also
[36:54] application. Not all financial
[36:56] institutions can do it. you need to
[36:58] apply with the government and then they
[37:00] come and do um all the due diligence, go
[37:03] through entire regulatory process, get
[37:05] approval and then it's um activated.
[37:08] Incredible for customers, for users to
[37:11] be able to use that app like ease.
[37:14] Thank you for that, Kungju. Um we have
[37:15] another question coming in from Maya at
[37:17] IWFBC, so British Columbia. What will we
[37:21] be discussing at IWS Cornerstone 2029
[37:24] with regard to FinTech? This is a great
[37:27] question. What is the hopeful story of
[37:29] where AI and FinTech can take us? And
[37:32] actually, I'd love to ask each of you
[37:34] for your thoughts on this as we're
[37:35] beginning to ramp down on time. So,
[37:38] Jess, if we could start with you.
[37:39] I love this question. Thank you so much.
[37:42] Um, look, like I like I said earlier on,
[37:45] I I think where the future lies is we've
[37:49] got such a tremendous opportunity to
[37:51] harness AI to use it both for good in a
[37:55] secure and trusting way. And my fellow
[37:59] panelists talked a lot today about
[38:00] trust. I think that's going to be a
[38:02] consistent dialogue. With my bias to
[38:05] security, I fundamentally think that the
[38:08] AI future models will ultimately give us
[38:12] an ecosystem within fintech that is
[38:15] truly modernized, truly 24 by7, stable,
[38:19] resilient, trusted, and secure. And that
[38:22] feels like a pretty powerful
[38:24] opportunity. And if if I'm here in 2029,
[38:27] remind me. And hopefully that is the
[38:28] conversation we have.
[38:30] Do that. Good answer, Mativa.
[38:34] Um, I think it's a great question. Um,
[38:39] so, you know, we none of us have a
[38:40] crystal ball,
[38:42] but if we were to make um some sort of
[38:44] predictions and look into the future,
[38:47] um, I'd say we'd be working
[38:51] far more efficiently because we have AI
[38:53] to support us. we'd probably be doing
[38:56] more higher value work uh because the AI
[38:59] is actually doing the grunt work, right?
[39:02] We'd be taking more judgment calls
[39:06] um working in a more responsible
[39:08] environment because we've got trust and
[39:10] security built into our systems. Um I'd
[39:15] also say that uh you know as we sort of
[39:18] think about what's happening within
[39:21] financial services today I mean clearly
[39:23] I expect the regulatory regime to catch
[39:26] up you know uh pretty pretty
[39:29] significantly uh we are seeing that uh a
[39:32] little behind from where technology is
[39:34] and you know probably that um and then
[39:37] lastly I'd say that from a you know just
[39:40] from a from each of us as individuals
[39:43] I think our ability to include customers
[39:47] and and and folks that have been
[39:50] excluded earlier because we now have
[39:52] more responsible AI, we bring more
[39:54] people into the fold. Um, and you know,
[39:57] I mean, just the Javon's paradox, right?
[39:59] I mean, as we as we see more and more
[40:02] technology and AI coming in, we'll
[40:04] probably see markets expanding. we, you
[40:07] know, we'll see more addressable markets
[40:09] that companies can go after, more, you
[40:12] know, more of the planet being able to
[40:14] take advantage of the the services and
[40:17] and products that companies have to
[40:19] offer. So, we'll see definitely see a
[40:21] much more expanding market which is all
[40:23] good for the planet.
[40:24] Absolutely. And I think that point on
[40:25] inclusivity, right, touching into
[40:27] people, giving people access to bank
[40:29] accounts that may never have had access
[40:31] to a bank account before, digital
[40:32] literacy, AI literacy is huge. So that
[40:35] inclusivity point is something to be
[40:37] really hopeful about. Kungju thoughts
[40:40] from you on this one.
[40:42] Um I think going forward there's going
[40:46] to be a lot of developments. I think we
[40:48] need to look out for what I call AI to
[40:50] AI economy. A2A maybe um the agents
[40:53] handling the agents. We need to look out
[40:55] for that. But I think um I think we
[40:58] should think about what happened uh in
[41:00] the history when the Roman Empire
[41:02] collapsed. It was because their force
[41:05] was replaced by outside forces. So I
[41:09] think we should not forget that and that
[41:11] human beings should really be the one
[41:13] holding the light. Yeah.
[41:14] Thank you.
[41:20] Thank you Priyanka. Your thoughts.
[41:23] I think we're seeing the change already.
[41:25] Finance used to be like institutions
[41:27] like banks and insurance agents. Now if
[41:30] you look at it, finance has become more
[41:32] of a capability. You see it in your
[41:34] shopping app. You you see it in your
[41:36] Door Dash when you're trying to buy
[41:38] something expensive in Amazon. Pop up.
[41:41] Yeah.
[41:42] You know, buy now, pay later. That's
[41:44] what it is. That's actually a loan. But
[41:47] that's not how it is. So the experience
[41:50] is changing. So I think probably banks
[41:53] might be obsolete if they know transform
[41:56] themselves.
[41:57] Whoa.
[41:58] Right. So digital cap everything,
[42:00] finance is going to be in your
[42:01] fingertips. But what I really long for
[42:04] is I look at my Robo Mop and longingly
[42:07] look at it and say, "When will you do my
[42:09] laundry and dishes?" That's what I'm
[42:12] waiting for by 2029. I really need that.
[42:23] Wonderful. And I I love the idea of the
[42:24] future of banks, right? It's going to be
[42:26] dramatically different in the next
[42:27] couple of years, right? How we bank, how
[42:29] we interact with our banking
[42:30] institutions will be different. So I
[42:32] know we're nearly up on time. Jessica,
[42:35] Mativa, Kungju, Priyanka, thank you for
[42:38] such thoughtful and engaging insights.
[42:40] Honestly, this has been a really
[42:41] stimulating discussion and wonderful to
[42:43] hear all of your thoughts. Um, please
[42:45] stay seated for our next session. We'll
[42:47] begin shortly. And thank you everybody
[42:50] for joining in this session today. Thank
[42:51] you. Thank you.
[42:52] Thank you.