From Pilots to Production: Making AI Real in Financial Services

Skyler Mattson:

AI is in news headlines, boardrooms, capital plans, earning calls. AI is at my dining room table right now, probably doing seventh grade math. And honestly, I'm not mad. I'm not mad about the help with the math. But beneath all the noise is a much harder question. How do leaders make AI real inside complex organizations? And by real, we mean responsible, we mean at scale and in a way that delivers real enterprise value. So tonight's conversation is about moving past the hype into substance. I've been told there will be no hype, only substance, so let's stick with the plan. I'm joined by two leaders who have sort of different but complimentary takes on this topic. So quick introductions and then we'll dive in. Dennis Gada is the executive vice president, Global Head of Banking and Financial Services at Infosys, where he leads our largest business unit and serves on the executive committee. Dennis, thanks for bringing us together tonight. So yes, we're going to be talking about technology a lot, but first I want to ground us in a leadership reality. Dennis, you've said the biggest barrier to AI isn't the technology, it's the organizational readiness. And I'm sure many of us either are experiencing this or hearing this. Now, Dennis, you sit down with CEOs a lot today. What's the first uncomfortable truth that they need to confront before AI can unlock any real value?

Dennis Gada:

Yeah, so I think when I meet some of the leaders, the first thing I ask them is, what are they doing with AI? And everybody says, we've done some of these POCs. They have a strategy on PowerPoint. Or maybe they have done some vibing, and they can show some prototypes. But I think the uncomfortable question or the important question to ask is, ok but finally, what have you implemented enterprise grade into production? On the business side, on the technology side, and how are you measuring value out of that? And I think that's where it gets a little tricky. And I think the other conversation that happens is, OK, if that's not happening at speed and scale, why is it not happening? What is the problem? Is the foundation a problem, or data, the process? Because if you're going to just apply AI to the existing mess, it's going to amplify that. So I think those are the two things which we really have a lot of discussions. First, get over the hype. Get over the POCs and pilots and vibing. And everybody thinks, oh we did vibing, so we know AI. That's not AI, right? What are you really using AI for in terms of actual production implementations? And then how are you fixing the foundation before you are ready to scale with AI?

Skyler Mattson:

I see a t-shirt. Vibing is not AI, bruh. OK, Rob, from your vantage point as an investor, do you see that same leadership gap showing up when you're assessing where to invest?

Rob Heyvaert:

I think there are two different things. There's incremental change that companies need to get better at. And then there is this big existential question about what AI really would do or will do to an industry. We're only one industry, called financial services. Do we have to at the same time take a step back and say guys this is so huge. This is about rewiring the whole industry and the industry is not broken the industry works well, but the industry is based on product on the reducing, the reduction of risk, the lack of transparency. So I think AI to use a buzzword is going to rewire financial service holistically and only for one reason only that that technology is so powerful that it ultimately will bring a better product to the consumer. I think it's massive. It's a big opportunity. I think companies like yourself are doing a great job in the forward deployment. In other words, making AI real is a much bigger job than people think.

Skyler Mattson:

What's making AI real? So you've talked about the market moving from a gold rush to something more grounded. What tells you AI is creating real value versus an impressive demo, by being a flashy prototype?

Rob Heyvaert:

The power of the technology. So if you think about it, and Jensen said it so well yesterday, he said, we didn't wake up engineers or smart people didn't wake up one day say I want to be behind a computer typing code and I want to be a typer. I thought I loved how he did it. It was very interesting. I didn't wake up to be a typer. I woke up to create products and capabilities. So I think that power is amazing. I don't think we have to be. So I'm a bullish guy, not a bearish guy. We don't have to be... I do believe for what it's worth that AI will deliver on the promise. Efficiency-wise will deliver on the promise of better products. It's the unknown that is an issue and the forward deployment is a very important place to go. But I'm personally very bullish that... I would say the successes will be there. And it's much more than productivity because at the end of the day, I think we will have economic growth. We talked about it this afternoon. And at the same time, we'll have an opportunity to reinvent ourselves and be more relevant for our clients, do it at a cheaper cost point, but be in a better ecosystem. So I think we're in the first innings of a deployment, but we will need services companies to do it well. And we also will need to rethink our business model. And I think many people talk about it. But at the end of the day, when the industrial revolution happened and they built these wonderful machines and they put them on the shop floor, somebody once said, oh my God, those machines are there, but the shop floor is not right. Instead of just having the machine dropped into a shop floor, what we really need is to rethink the whole shop floor. So that is something I think that's not enough happening. So it's at the one time, let's be incremental. Let's make sure AI pays off. On the other hand, we have to rethink holistically how we think about serving our clients, how we clear, how we settle, how we take risk, how we get transparency. So I think it's a big re-architectural job in my humble opinion.

Skyler Mattson:

Okay, since you're talking about rethinking all of it, I want to jump to this question, Dennis, in large, regulated institutions, how do leaders genuinely focus on rethinking workflows? It's not just about automating what exists, which is a lot of what we're seeing out there, but where do you start when you have to rethink the whole system?

Dennis Gada:

Yeah, see again in large or mid-sized financial services firms, the most important thing people talk about is complexity, regulations, bureaucracy and so on, right? And to Rob's point, I think the power of AI is to really see how some of those challenges can be addressed. And I’ll make it real with maybe some examples. So one of the large financial services firm that we work with, some of their platforms have been 40, 50 years old, mainframe, cobalt-based platforms, to modernize something like that before AI would have costed anywhere between $500 million to a billion dollars, and could have taken anywhere between 5 to 10 years. And then there was still no guarantee of success. Today, and we are working on a couple of live programs like that where there is very high confidence that it can be done within maybe a couple of hundred million dollars, a couple of years. So it's still not it's free or it's like at the press of a button. There's still a lot of work to be done, but it is getting real that many of these complex things which were... too expensive or almost impossible to do within a reasonable time frame are now getting possible. And when you start working on things like that, you realize that for this specific client, as an example, wow, this is working. Now let's do two more of these, right? Because there is just so much legacy to clean up. And I think that's the whole reducing complexity is a very big element of how AI can help. Second, on the process, I mean, we can, you know, even after all the digitization and everything that has happened in many large banks, especially on commercial banks, I know there are some leaders from industry here opening an account, KYC, et cetera, still takes like 20 to 30 days after all the digitization. And then by automating it, you will go from maybe 20 days to 15 days. That's not what you're looking for, right? How can you fundamentally reimagine the process where even for a large complex corporate, you can do KYC within hours, if not one or two days. I think those are the fundamental shifts like modernization, complete reimagination of processes, customer service, we've seen, we've implemented for some institutions where I can genuinely tell you because we have done it and we tested, you know, AI agents doing customer service actually get better outcomes, get better ratings than humans. Because it's really making it much more faster and convenient for customers to interact with AI agents. And they don't even know whether it's an agent or maybe sometimes they do. So I think those are big areas of shift that we see.

Skyler Mattson:

Where do leaders still underestimate that infrastructure transformation that has to happen to make AI stick?

Rob Heyvaert:

I would argue that innovation is the ability to move in both, to be able to incrementally help change, use phenomenal companies like Infosys to do so, and at the same time, think big and the think big is not the awesome front-end app. I mean I believe in a super app ultimately there will be such a thing where somebody will ultimately understand who you are as individual and will be able to cater the financial services needs around you as an individual but we are so far removed from that. The models are there. The system, we could build it together. If the plumbing in the back is not doing what it's doing and there's nothing there, I would make a bold prediction. We need to spend between four to eight trillion dollars in financial services to even be anywhere near the power of the LLM models that exist today.

Skyler Mattson:

Good news, your stock’s going up.

Dennis Gada:

In fact, talking about that, we had a investor day, I think in February, and very similar, chairman whom you know very well, Rob. Nandan, he presented a slide which said, the technology is here and the actual adoption is here, right? And people are worried about how technology is going to get even better. There'll be new models and all of that. Good, but there is such a huge gap. And in spite of all the acceleration and so on, it will still take a lot of time for the adoption to get it. I've had CIOs and CXOs of leading financial services institutions say this on stage, right? That even if the innovation were to stop today. No more innovation. There is enough to be done for the next five to seven years, if not more, to adopt what's already there.

Rob Heyvaert:

It's kind of fascinating because people say, oh, it's going so fast, it’s going so fast, it's going, it's very slow. In financial service, it's very slow. So, and it's interesting to see the contrast is growing and what that actually means. You know, my case for investing and in your case for helping your clients I was on a panel this morning about prediction markets. If I would be in the prediction business, I would say any modern transformation for financial services is going to be seven and 10 years, not one year, two years. And that's very good news for guys like you. And it's also very good news for those who don't have legacy. So we haven't talked about this, but I do believe that, and somebody said, no, no, 80% of the incumbents are going to adapt with AI. I think there is this risk in financial services, and you're seeing it popping up with the Robinhoods and the Revoluts of this world, that they were once feared, then dismissed, now feared again. So it's been fascinating to see how AI will also break through with better service models with less legacy. And time will tell. I mean, I think Jamie's fine. He'll figure it out. But the ones in the middle could be stuck. And the ones who are not embracing, everybody says it, the ones who are not embracing AI will be victims of AI.

Dennis Gada:

And maybe I'll just add, you know again, if you look at the US landscape, there are 10 big banks or asset management firms, maybe another top 50 institutions who will figure this out. And, you know, they have money, they have intellectual capital, but there are 5,000 financial institutions, lot of small banks, credit unions, small firms. I think it will be extremely challenging for them to drive this adoption. And if they don't, they will get really left behind, right? So I think again, there for some of the Motive portfolio companies, for companies like us, I think we believe there’s a huge opportunity to help them uplift the technology evolution that needs to be done.

Rob Heyvaert:

It's interesting when everybody talks about democratization, but that is really what's happening. So it's on the one hand very hard as a legacy player to adapt, and on the other hand, I have never ever seen in my life so much innovation coming to us on a weekly basis. It's the super smart kid from MIT that listened to his dad who was a trader and now has a better way of trading. And he's never even been anywhere near a financial services industry. That level of innovation, I just want to put everybody on notice, this is unprecedented. And it's good news because it’ll wake everybody up, but it also makes me very bullish about what the art of the possible is with AI in financial services.

Skyler Mattson:

Okay, you started talking about people, so I do want to shift there because I think so many leaders in this room are probably struggling with what to do about talent and what skills and roles are we looking for in a world where AI isn't just assisting anymore but actually executing work. Dennis, what's your take?

Dennis Gada:

As we look at Infosys, everybody needs to move ahead in this AI journey. We’ve classified as AI aware, AI builders, AI experts. So across the spectrum, we will have different levels of proficiency, right? But skills is one part of it. I think it's also the ways of working and the operating model changes that will happen. We’re already working with clients on pods where, you know, normally it would be a team of eight humans and it will now be maybe six humans and two agents or four and four. There will be some humans in the lead. So this is a different operating model. Firms are actually classifying digital workers in their HR systems and working with them. So different models are evolving, but I think most important, at least at Infosys, what we've been doing is making sure that across the spectrum, people are getting up-skilled, but also we are consolidating certain roles, especially things like the traditional software development lifecycle from product to requirements to design to development to testing. A lot of those roles are getting compressed. And finally, I think the distance between technology, operations, and business is also getting a lot more closer, right? So, I met somebody earlier today and said, this wave doesn't seem like technology is doing something and we just have to follow. But from the business perspective, we have to be with technology right from day one or almost lead in some cases to do an end-to-end transformation. And that also needs different skills, not just for the technology folks, but also for people on the business side.

Skyler Mattson:

Rob, what's your take on talent and where are you seeing leaders get it wrong?

Rob Heyvaert:

Huge shortage of engineers for the next five to seven years. Not the other way around. Now, I say that and… You should speak to some of our investors and analysts. And I say that and then just one of my friends just got laid off as part of a big transformation and we didn't, whatever they call it, and he had a job within 24 hours. I mean, the stats speak for itself. There are more open positions for engineers than there were a month ago or a quarter ago. And this is where we are always wrong with predictions, to be clear. I think we are going, as humanity, that's a heavy word, as humanity, or as a company, or as a country, we're going to find ways to create environments where, but there's going to be a re-shift of labor, for sure, but nobody knows what that means. Remember, I had this little meme that I saw a few days ago, kids in the UK being interviewed in the... they were like ‘64, it was in 1964, and there was these young kids and they asked them, what's the future and what do you think 2000 will be? And they said, it's boring, there will be no more jobs. It's unbelievable, I'll send it to you, it's unbelievable. And we have the same thing. I go into a meeting and oh my God, we can run. And it's true, you could technically run, I'm going to not say JP Morgan because Jamie's not going to like it, you could probably run not Citi because Citi’s here. You could run company X with 10, 20% of the people that have today if AI is fully implemented. The question is A, will it happen? And B, what could those people focus on? 70% of financial services is some form of control, some form of reconciliation, some form of whatever you want to call it. It is not servicing the clients. So imagine that, yes, we will lose some people and some people will reinvent companies, some people will retire, some people will go into different sectors. I think organically I'm quite bullish that we will enter into a world where it might be 20% unemployment, but there will be a capital structure underneath that will allow that to happen. And the 80% I'm talking about will be very innovative. And for me, between the engineers staying here for the next five to seven years, long, long term, we have a history of reinventing ourselves. And you can say, ah, this time’s different, Rob. It's not the steam machine. Yes, this time is not different. It's the same human ingenuity that we had then we have now. And we have it better because everything’s democratized. Everything gets cheaper. We live better now. My neighbor lives, not my neighbor because he's very wealthy, sorry. Some neighbors live better than Rockefeller lived 100 years ago. So it's sort of like an interesting phenomenon. I think it's all going to be fine.

Dennis Gada:

You know, I mentioned this in my Milken panel. I'll repeat it here. In fact, I heard it from somebody else, and then I validated it, right? In year 2000, when Y2K and other things were happening, there were 1 million software engineers in the world, approximately. In 2026, there are 30 million software engineers in the world, right? Now, in the last 25 years, it's not that software engineering has become more difficult, or there’s not been any automation, or so on. All of that has happened. But it's just that the amount of software that has been written now in the last 25 years has grown so much. And if AI makes it easier, faster, cheaper, it will grow even more exponentially. Yeah, we will have eight billion engineers. So I think it's going to be an explosion in some ways in terms of how much work gets done. And then when the volume of work is so high, whether it's coding or something else, then it will still need systems, processes, and people to manage that.

Skyler Mattson:

Rob, you're going to give this room a 12-month playbook. Well, not all 12 months. We're just going to focus on the two to three things leaders need to do now to ensure AI delivers real enterprise value. Two to three things. Most important.

Rob Heyvaert:

They need to make sure that every individual in the company, all the way down to the person who welcomes you at the reception of your company, are immersed in what the power is of AI. And people say, oh, I want to be tangible. Forget tangibility. I mean, it’s tangible. So be it. Let's make sure that everybody understands that there is enormous creativity, because it's exponential. And we see it now in our company. We have a very small group of company. We manage 80 companies. We have billions of dollars, billions and billions of dollars. But I've asked everybody to spend at least three hours a day… it’s costing me a fortune by the way, for what it's worth, three hours a day with AI tools. No strings attached, three hours a day, a given. And then every weekend, every Friday we have an, what do call it, hackathon or whatever they call it, we come together, we explain what we found and what we did. And every month they have to build me a model and one guy... 27-year-old guy said to me, can you come into my office for a second? He didn’t have an office. Can you come into this corner? And he showed me that he had remodeled all our portfolio, all 80 companies, their cash flows, their forecasts, their history. And he was showing me how well they were doing. And I said, are you kidding me? He said, yeah, it's a Claude application that I used. And it's not perfect because my next link is I'm going to link it to the existing data. I'm going to go to Prequin. And this is our old data provider. I’m to go to that. I'm going to connect to SAP. And this is a 27-year-old associate at Motive. So imagine if you have, I don't know how many hundreds of thousands of people there are in city, and you could say it's chaos, or it could be the definition of what I would call your AI readiness as an organization. So, first priority this year is let's get everybody to understand the power of the model, and let's then take one or two very tangible, ambitious projects and deliver them within every quarter. I mean, the other thing that it democratizes, in my humble opinion, when it comes down to is you could get stuff done faster. I mean and it's funny that those people I asked and then the flip side is the 60-year-old investment professional at my firm comes back to me and says it's unbelievable Rob I used to be a coder but now they can code for me. So that notion is important. So implementation of awareness of AI implementation and then run certain projects to the ground in one year is my advice.

Dennis Gada:

Yeah, I mean, I would agree, go deeper in some areas and make it really transformative. But the other thing I would say is, and we heard a lot of this at Milken on both sides, but have a lead with a positive narrative, right? I think there's a lot of anxiety people are creating, doomsday scenario, nobody knows. But I think what you can, as somebody says, right, be an optimist, even if you die, you'll die happy.

Skyler Mattson:

I think we'll take one last great question, a real hard one, to end it on. We just had some news today about Anthropic unveiling AI agents for the financial sector. I'm curious what you think about that.

Rob Heyvaert:

So the whole notion of collaborating, it's happening. Now, we are living in the age of announcement, where to be honest, every announcement, I have a company, we have a portfolio company that made an announcement and took the industry sector down, the insurance sector, brokerage sector down by 17%. And their announcement was that we're going to build this smart agent and everybody responded. So it's interesting, but this announcement is real, because I know the company well. That is what we're going to see, because these LLM models are not meant to be purposeful for everybody. This is all about them becoming very powerful and other people, other organizations taking those models and doing complex workflow, going into proprietary data sets and actually basically doing what they need to do. If you look at FIS, they have tons of books of records and so on. So you’re going to see more of those announcements. I’m actually pretty close to OpenAI and they have 10, 20% of their company working on partnerships. That's what they do. Because they need it. There's so much need for experience and capability building for these models to be adopted. So expect more of it. It's pretty exciting. Pretty exciting in my opinion.

Skyler Mattson:

It's not a tough question. It was a nice question. It is, I think it was on everybody's mind, so I'm glad we got that in. Thank you. Thank you so much. I think they kept it grounded. I think there was substance instead of hype, so I appreciate you sticking to the plan, and let's enjoy the rest of the evening and keep the conversation going. Thank you.