074: AI and the State of Financial Services

VP of Product Marketing Brian Thompson joins Nikola to discuss the current state and future prospects of AI in financial services, including banking and insurance. The chat covers the importance of human oversight in AI deployments, the evolving landscape in banking and FinTech, optimization of back-office processes, and how financial institutions are leveraging generative AI. Brian shares insights on the differences between large legacy banks and smaller community banks, and the challenges and opportunities AI presents in bridging these gaps, while tackling customer expectations, regulatory concerns, and the significance of a strategic approach to AI implementation.

• Follow PolyAI on LinkedIn
• Watch this and other episodes of the Deep Learning pod on YouTube

Transcript

Brian 0:00

the overwhelming bit of advice is as you deploy AI and, agent ai, generative ai the consensus seems to be that the, you do that in a way that has. A great deal of human oversight, and. I like, I think that this is right, but it's right with an asterisk

Nikola 0:24

welcome to another episode of Deep Learning with Poly ai. Today I've got Brian Thompson, our VP of Product Marketing here with me to talk about financial services and the work that not so much we are doing in banking, insurance, but in general trends and the state of AI within that whole ecosystem. Brian, what's happening with AI in the banks?

Brian 0:45

What's happening? Great. Great question. Very broad. Look, so I've actually spent a lot of time in my career in various, asset, various areas of the FinTech and working closely with everything from, major banks to challenger banks and so on and so forth. And so I've had the chance to see things evolve over time. Right now we are in, we're in this period of. Not backlash against AI hype, but people are trying to think in really practical terms about how do we get meaningful value out of deploying ai and how do we do it in a way that scales effectively and is at the end of the day, very safe. And so this comes down to. Optimizing back office processes, data analysis those types of things. That's the principle areas of inquiry. But also at the same time, things are very much in exploration mode. It's POCs more so than major transformative initiatives.

Nikola 1:39

Okay. Okay. I've always liked banks and like the work we did with them, not because it's easy. But because it moves fairly predictably, and I think that while the requirements in the onus of, I think the number of hours we put in terms of answering InfoSec questionnaires and ensuring regulatory compliance is probably the highest of any industry we work with. It is an area where, CIOs are very commercially oriented and like when decisions are made that they tend to flow in a very organized way. What are we seeing with AgTech applications in banking? You mentioned the back office, but is there anything else happening?

Brian 2:14

Yeah, so again, I think I, I hate to say. Because it's not, it is not that there's a solution in search of a problem. But there, I think most of the excitement really does come around. When you think about the amazing amount of data that exists, especially in a large global bank and the oftentimes disaggregated backend systems where various bits and pieces are housed there's a lot of excitement about using. Using Agen AI to pull things together and to analyze vast troves of data and to get, meaningful, actionable insights. Of course, there's all the work that's been done with robo-advising and things like that. There's an extension of that now that people are talking about. Can we go a little more one-to-one? With agents and that sort of thing. And long tail of, again, process optimization and this and that. And that's like kind of the story particularly of the more venerable banks is that it's this long tail of moving things from manual processes to increasingly automated ones. And agent AI lends itself to that if it's, handled correctly.

Nikola 3:18

Okay. Okay. How do you see the whole difference between the very large bank and the kind of like wider set of, like credit unions community banks?

Brian 3:28

Yeah. Okay. So I. Like I said one of the, one of the major issues that, legacy banks face scope and scale of course, they're just huge. And that introduces a lot of really meaningful challenges. Additionally because, so many core banking systems are, built on legacy mainframes and other things like that, there's much more of a natural build culture because there has to have been a natural build culture. And because you need to make a lot of very modern systems work with things that are really suited for purpose for what they do essentially, but that are long in the tooth and maybe not the best citizens of those larger ecosystems. And so just handling those challenges is very much a top priority for the large organizations now. Versions of that of that, as you scale down into regional banks and credit unions, but nowhere near to the same degree, it's it's easier to wrap your arms around around that problem. Just because, you're dealing with fewer fewer customers. You maybe you have a smaller portfolio of product offerings, that kind of thing. At the same time, however there's I think a real opportunity or a real desire for, for those regional banks to differentiate more effectively. And so you start to see like maybe maybe agent AI is not necessarily going to change the world with. Massive data analysis or improving processes for thousands of employees that don't exist. But they're thinking about it more in terms of how can I use this in a ruthlessly competitive industry. How can I use this to win more customers from, our larger competitors offer a service that is more engaging, more personalized in some capacity. And so I think that's more the direction of thinking as you get smaller and more challenger oriented. Yeah,

Nikola 5:09

I mean I always found it fascinating, all our two main markets when you look at the US a long tail of banks and credit unions, I think there's 4,000 different entities that will classify as a one shape or form of a bank, right? You compare that to the UK where you've probably got like sub 10. Actual banks, including like the billing societies and stuff, maybe you've got up to 20, but it's a real d in Australia, I think you call them four or five banks and that's it. It's really interesting'cause I think both with the larger American banks and with the Europeans, there is a real appetite for the best of breed solution because of that scale. Everything delivers a huge ROI, right? If you've got a contact center of a few thousand people, if you've got a multinational one with different languages, it creates human inefficiencies that AI doesn't suffer from.'cause it's like on demand. But it's also been, I think, a space which. Has historically lent itself in conversational AI to vertical specific players because there's just a big market where serving them and understanding their pains and needs has been enough for people to get to a relatively small scale, kinda like Series A b, we've not really seen any of those players come to dominate the market or really emerge as the more horizontal solutions. And I can't help but wonder if it might be that whole like regulatory burden. And the fact that they're slower, that stops them from a competing activity. So for us, we really have more of a kind of like GM model where then there are vertical specific deployment teams and things, but the platform is horizontal. It's how do you think about all this and like in general, this build versus buy?'cause you mentioned that they do a lot of builds. I mean they do, but I will not classify them as. As build heavy as say a telco, right? I think telcos are like, the more, the most extreme form of like their tech companies by evolution, right? So they really aim to build and to engage that muscle. They just have to work that muscle'cause they have it, right? I think with banks you find their willingness to rely on services and GSIs and on others, but how's that changing with kind of new generation?

Brian 7:17

Yeah and I mean we covered a lot of ground there, but of course the closer you get to anything that brushes up against a regulatory concern like the, that increases the desire to own it end to end and to have like total transparency, so on and so forth. But actually stepping back a little bit here'cause you asked about vertical specific players and what we've seen And I think it's, because of the intense regulatory scrutiny that exists even for the smallest financial institutions. I. Of course people are they're gun shy a little bit, there's a lot of trepidation around ai and it's not really unwarranted. And so what ends up happening, of course is how do you build that trust? Some of these smaller brands can go in and they say okay, this solution is built just for you. And we really understand your problem space. To whatever degree. And that will hopefully, I think that's part of their strategy of getting past that first hurdle. Now, unfortunately, knowing the problem space or knowing the specifics of, regional bank regulations it's a nice to have. It's incredibly useful. But in terms of deploying, like AI technology, in particular agenda, AI technology, that, that actually delivers a measurable positive outcome. You need more than that, you need the much more sophistication in the tech. And I think, and maybe you can speak to this a little bit, because we've, we have very explicitly taken on a horizontal approach with, as you say, GMs like, and I think that's because, we want to understand problems, we wanna understand our customers, but we, I think we also have this sense of the core problems. We, we solve our like conversation and AI performance and, like significant scaled deployments. But what have you seen there? I like the term, right?

Nikola 8:59

I think I, yeah. The, in Fs the problems are, there are some call types that exist everywhere, even like cross country where people call and, ask for help setting up Apple Pay. And I think there was a change a few years ago, which just meant that you have to reenter the details of every card when you get a new phone. And trust me, like 10% of most contact centers that deal with like retail banking, have people calling in and just being asked to be taken through that process. Now they have choices around will they act, actually stay on the phone with them to help out and guide'em through? Or is it like you can go online or here's a link or it really almost depends on the agent and their personal decision around, how they feel that day and will they actually like persevere through? But those are simple things. Then you go into kind of like the higher if you're blocking a card, if you're activating a card, if you're transferring money, especially if you're adding new pays, it tends to be the kind of thing where, you really have to build and well get approval from InfoSec where it is sensitive. And weirdly, there are a lot of people who still call to do these things, they don't do it on the phone. Even though most banks, especially the bigger ones everywhere right, have invested a lot of money into their. Apps into digital self-service. There is still an unbelievable number of people who do phone banking who call to, check the balance to ask about a statement. Then you'll have questions about, like payments, outgoing payments, whether payments have been received. So there's a lot there. I think that. For us spiritually really, it's always been about building the best thing for any vertical, right? And there's a layer of kind of like agent that is agnostic to the vertical. But the more data you have, the better the speech recognition gets. The more sense you have in the platform to build out templates and use cases for these different things. We can pretty much get like a bank up and running in a month with almost everything if we are to like magically have all InfoSec. Done in advance. But we're seeing a lot of momentum there. We're seeing momentum with a lot of kind of lifting quote like tier one FinTech, right? Very large companies that have actually gone full circle in their build versus buy. They were always on the build side of of the spectrum compared to say, more traditional banks. We now see them coming back to us to co-create, right? And I think, to me, phase one of Poly was always very much about let's prove to the world that we can build 10, 20, 30. Outstanding voice agents because to most people when I ask them even today Hey, like what's the best voice agent experience you've had? It's just blank stares into, right? Because they haven't had good experiences. So to me, phase one was always built managed services for companies mostly with weaker it, those that don't wanna build, right? But where we're now seeing huge success is with companies that do wanna build, but realize that you're off the shelf Google thing or Amazon thing. Does not have what it takes for them to build something that would be worthy of a tier one financial institutions. So that's really encouraging. Now, of course, I think that data, privacy data residents, there are a lot of things in FS that just make it that bit harder for them to have the latest thing, they might have. Requirements about data not leaving EU or even a specific country or the us. For the US it tends to be easier'cause a lot of the subprocess at least will be in the US and in general, the US is a bit easier with all of these things, although in some sense, and with some FS related things, it can be harder. I think in the EU you have Dora the Digital Operational Resilience Act thing, which has recently just, I think, basically. To X the amount of paperwork you have to fill as a vendor whenever you do anything. So yeah, there's just I think it's harder for a startup to embark upon that unless it's really choosing to do it. FS is like my one true love of like our of our verticals and, we've persevered and we've won, big global systemic banks. We've won top three American insurers. Some other really exciting things that I can't talk about yet, but, it's taken a bit of time for it to move, but I do that when they move very systematically and in a very organized way so that if you get them to move and do the whole work of onboarding another vendor, then they're all in and then they do a lot and they tend to work really quickly. And I found that the bigger the bank, while you won't get any kind of like sudden overnight successes, the long term implementation of a very complicated project. It tends to work better with a larger bank than it does with a smaller one, because they've got the budgets, they've got that certainty. They're an institution. And I love that about them.

Brian 13:43

Yeah. And to that point, they've got the budgets, the certainty the institution. They have executional experience. And what I wanted to touch on is they have real significant motivation to do this. Because there's when you think broadly about, finserv, whether it's legacy banks, challengers, fintechs, all of the different stakeholders there, we've spoken about regulations as a constraint for implementing technology, but regulations and other just brute on the ground realities are also, they limit what you're able to offer. Like the interest rate is the interest rate broadly and so they're looking for ways to differentiate, and they're looking for ways to differentiate where customers are less loyal than they've ever been. Their expecta, their expectations are much higher than they've ever been because, they're being, rightly or wrongly conditioned by other people in the space. And so you have, they're staring down the barrel of okay, we can bleed ourselves, white, offering incentives to try to get people to, to come on board, or we can treat our overall experience as a product itself. And being able to deploy really effective AI that's responsive and engaging. And even your examples, there are so many there are so many use cases that really are, I think they're better if they're automated. Sure I can talk to a person and try to cancel my car, but I would rather be able to do that, 24 7 automatically pick up the phone whenever. And I think that there are plenty of those things that, that exist throughout these really complex, non-linear life cycles that, that you have interacting with with large banks, small bank FinTech doesn't even really matter.

Nikola 15:15

Yeah. No, I think that like there how would you say that the customer expectations are changing now with ai?

Brian 15:21

So this is the. Really it's a confluence of two things. Now we've said this forever it's trite, but it's true. It's that, between Amazon and Apple and Google folks have just been conditioned to demand a lot. And they want things to be more connected. I want, click this and Oh, my ID works, and so on and so forth. So that's one element, and that's been going for a while. But the new thing that AI introduces. Is again, another level of muddiness where you've, I think misinformation and misunderstanding is really high out there. And we see this a lot, there's this general sense of okay, chat, GPT can write a solid C plus B minus term paper. So clearly AI can do everything automagically and they folks don't necessarily understand. Everything that goes into like having a system that is really performant and works in a business capacity. And I think the solution there is not to educate people out of it because you can't the solution is just to really, lean in and focus on like high value deployments of ai, where it makes the most sense and where it can have the greatest impact in a customer's experience. Because they will see that will register and that will tick those boxes and they'll go, yeah, this is different. This is not like what I got down the road. Or at the credit union or with whatever app.

Nikola 16:44

Yeah. I think that's the one thing which again, favors larger institutions that can mobilize to do this thing over the long haul. There are some examples from the past that have been Mercury ROIs, for example. Erica is Bank of America's like assistant. It's using their branding. I think that there are tens and tens of millions, if not more that were spent on it. People when I ask people kinda Hey, what's a, decent voice, Asian experience? You've had people probably mention this one as. An experience they've had. They don't really call it good. But it is something that's had a lot of money poured into it. Yeah. And there are large insurers as well that have deployed something. There are some that are doing it with us right now and I think it will improve. I think, one thing maybe to look at as just general insurance versus banking. What are we seeing in insurance?

Brian 17:33

Yeah. Insurance insurance is interesting. Especially when you're thinking about qualifying for for something or an application, all that, you almost think of the application process as a considered purchase or an extended funnel. And that's one of those things where people fall out of that funnel or and you want potentially to bring them back. Or let's say you, you still, you want to have a layer of qualification before someone gets to a human because humans are expensive if they're taking people's information over the phone. And so there, there are, of course, there are implications for experience. On, on the insurance side where people can have a, callers can have a better time just giving their info. But there's, I think more broadly there's the opportunity and we've already done this a little bit, is to just change the underlying unit economics of, insurance qualification. And that has massive transformational implications. Beyond what you'd expect. Yeah.

Nikola 18:26

Yeah. Increase the leverage of the business, of course. Okay. What else does the audience need to know about FS and ai?

Brian 18:32

Okay. Most most important thing that I have neglected to mention here and this comes down to, all of your best practices that you're gonna get where you see the blogs and the think pieces and all of that. The overwhelming bit of advice is as you deploy AI and, agent ai, generative ai the consensus seems to be that the, you do that in a way that has. A great deal of human oversight, and. I like, I think that this is right, but it's right with an asterisk because I think that usually the way what's being proposed is have AI handle something and then have a human review every single individual interaction, approve it or not. And that, if you do that wrong, it ends up like. Take your kid to workday you're spending a lot of extra time and you're not necessarily realizing efficiencies. And so when you think about how to insert humans into that process of review and quality control and improving like the overall the overall state of whatever your AI deployment is that's a platform level question that you need to be asking to people. And you need to think about, your strategies for collaborating. Of course, internally within your. Your your bank, but your strategy is also for collaborating with AI agents and with your vendors who are deploying those. And so it's a more complicated problem, but if you solve it effectively, you can get a lot more return on investment much quicker.

Nikola 19:54

Yep. Yeah, that's a good point. I think at that point we're like at at time. It's a pleasure to speak with you. Thank you for sharing these insights and thank you for tuning in. thanks a lot.