From ATMs to AI: The history of personalized banking

About the show

Hosted by Nikola Mrkšić, Co-founder and CEO of PolyAI, the Deep Learning with PolyAI podcast is the window into AI for CX leaders. We cut through hype in customer experience, support, and contact center AI — helping decision-makers understand what really matters.

Summary

Personalized banking didn’t start with AI. It started at the branch counter.

In this episode of Deep Learning with PolyAI, Nikola Mrkšić sits down with Nathan Pearson, GM, Financial Services at PolyAI, to walk through the history of personalized banking and what it can teach us about where AI fits today. From local tellers and early self-service ATMs to call centers, mobile apps, and intelligent agents, banking has been reinventing personalization for decades.

Together, they discuss:

The big takeaway: while tools keep changing, the goal hasn’t. Customers still want banking experiences that feel personal, reliable, and easy — even when they’re powered by automation.

If you’re thinking about how AI should support (not replace) personalization for your business, be sure to tune in.

Key Takeaways

Transcript

[00:00:00] Nikola Mrkšić: Hello, everyone, and welcome to another episode of deep learning with PolyAI. Today, I have Nathan Pearson, our GM of financial services

[00:00:09] Nathan Pearson: on the podcast. Nathan, welcome. Thank you very much. I'm very excited to

[00:00:12] Nikola Mrkšić: be here. Yeah. Likewise. And I'm excited about this episode. I think in the history of this podcast, the most kind of, like, surprisingly high performing series of episodes was Oisin Glenn, our head of solutions consulting, and I talking about the history of telephony. So as we thought of topics, I thought that one that we're really interesting to cover with you is the history of banks kinda, like, trying to both make their experience better and remove humans from the loop, cut costs by using technology and removing, well, humans from the operation. So, you know, you having deployed conversational AI on behalf of very large global institutions, I thought there would be a great idea to kinda, like, move us back through history to, like, where it all started. So what's the first place that, like, a bank ever really tried to use technology to get that leverage or improve experience?

[00:01:05] Nathan Pearson: Yeah. So I mean I mean, firstly, big shoes to fill in, you know, meeting Ashen's expectations. But, you know, the first actual really mass implementation of, like, technology to transform experiences was the the humble ATM. You might not think about it now, but that really did, like, change banking. Before that, everyone had to, you know, go into a branch like QR, talk to a teller just to get access to your money, to check your balance, to all those good things. The ATM actually provided the automation that you could, you know, walk up to a machine, fingers, and they put in a check back in those days to get access to your money. But that was the early sixties, mid sixties. So what's that sixty years ago? That's a long time. And banks and experiences have changed a lot since then.

[00:01:51] Nikola Mrkšić: Yeah. So, like, do those first ATMs even have, like, cards associated with them, or what did it work in a different way?

[00:01:57] Nathan Pearson: No. I think it was like a check. I think I think there were I think it's one of these, like, technological horse races, a bit like, you know, VHS versus Betamax and, like, different modalities were created in in different parts of the world. So I think the very first one was here in The UK in North London, Barclays Bank. Congratulations, Barclays. It was the first one to launch, but I think that was quickly followed by, you know, banks in The US, in Asia doing similar things with different technologies. I think it was a few years before, you know, the PIN number emerged, before the actual card emerged. But yeah. So there were quite a few different varieties back in those days. That is so interesting. Yeah. I mean, as you were talking,

[00:02:37] Nikola Mrkšić: I checked in, yeah, 1967 in Enfield, Barclays. And I guess what's really interesting is they automated cash withdrawal before they had these, like, full digital identities for customers. And kinda like with any form of self-service, there were concerns about security. In this case, physical, where someone can jump you in. I think kinda like how people don't yet maybe accept that they'll be doing everything with AI agents. Similarly, I think it was quite a quite an ask of customer base to, like, not be in the safety of a branch getting money, but instead kinda like do it

[00:03:10] Nathan Pearson: in the middle of the street. Right? Yeah. Definitely. Definitely a big shift, but I think it shows the customers value convenience. They value the ability to, you know, get access to the basic transactions quickly and easily, and that you don't need it really did start to prove the hypothesis that we've seen, you know, really evolve that you don't need a human member of staff to do everything for you. You know, a human is a customer is competent at doing some of these things themselves. Yep.

[00:03:38] Nikola Mrkšić: So I guess the overhead there is or, like, the benefit of that first level was you didn't have to wait in line at a bank physically. You still had to go to the bank because they were associated, firstly, I think, with branches. Then later, they kinda, like, started showing up everywhere, offering you the convenience of, well, having to travel less. So I guess that's that's a benefit. What what came next in terms of, like, self-service journeys?

[00:04:04] Nathan Pearson: Yeah. Well, I think, actually, the the next big shift is not necessarily self-service in the traditional sense, but it was it's telephone banking. And that really broke the the thing that you just talked about how, you know, a customer still had to go to a branch to go to an ATM. The telephone banking actually meant you didn't need to leave the comfort of your home or your workplace, and you could just call the bank to do basic banking transactions. So, yeah, that first launched in the in The UK at least with First Direct in the late eighties, and that was, you know, revolutionary at the time. You know? This this whole idea that, you know, banking wasn't just about branch physical in the location of the banking organization. You could do your banking from anywhere, and therefore, you could put banking staff in other places as well. They didn't need to be on the high street. They could be in these vast aircraft hangars that have become, you know, contact centers we know today. But those those early telephone banking days, there was no automation. It was actually, you know, human to human, quick response, but it it's built that trust because, you know, you can't physically see them, but answering the phone, just talking to another human being, they've got all the access to, you know, your banking services, your systems, etcetera, etcetera. So that was, I think, the next big shift. You know? The the ATM was that, you don't you don't need a human from a member from the bank to do the service for you, then the contact center or the telephony telephone was that shift that you don't need to go to the physical location. So you're starting to see the shift in actually what those experiences look like.

[00:05:35] Nikola Mrkšić: No. Absolutely. You know, my late father, like, till the end of his life, would call his bank to check kinda like the the amount of his account whether if his pension arrived, whether the payments were taken. And he was, by no stretch of imagination, a technical blue dyke. He was on YouTube, you know, watching conspiracy theories and just, like, consuming all sorts of technology voracious, though I still kinda, like, want him to call and have that. I guess it's a personal relationship, but then again, I guess it's not that personal because it is a streamlined thing with the contact center with a lot of people. I guess post that in evolving, it comes

[00:06:10] Nathan Pearson: the dreaded IVR. Right? Yeah. Exactly. So IVR was next. That was bringing the best. I don't know if what best is the right word, but it was bringing automation plus remote interaction into a single place. And so that really landed in the early nineties. Interestingly, just back to the the previous example, First Direct is the one bank that still very proudly does not have an IVR. They were strong believers that that human to human relationship should exist. Whether Whether we agree with that or not is a different matter, but every other bank in the world pretty much now has those dreaded IVRs. And obviously started with, you know, the worst of the worst with the, you know, the DTMF touch tone, the press 1 for banking, press 2 for cards, etcetera, etcetera. Customers pretty universally hated them, but banks pretty universally loved them because, you know, automation, you know, deflection, cost reduction, containment, all those, you know, those classic words you associate with automation initiative. That's what they gave you.

[00:07:14] Nikola Mrkšić: Yeah. No. I mean, absolutely. And if you think about it, I know of many a a large bank, and the stats from those things can be, like, 70% containment, where if your customer base has no choice but to learn how to use these things and you keep it on for a decade, well, like, it becomes quite functional. And then, you know, as you evolve the interface to a more conversational one, it's not always easy to match those numbers right from the get go, especially if in the newer instance, you stop insisting on there being no way out of the of the self-service labyrinth. Right? I guess what's also interesting about those applications, especially, well, whether they're voice driven or DTMF, is that it is an application. Right? So kinda like as we think of the next step, which I guess would be mobile banking well, Internet banking. I think Ally Bank was a pioneer in the early two thousands of this kind of, like, you know, web interface where your banking would live. With that, I guess, it starts being an application. But even before that, after you open an account, you could do a lot of the basic banking over the phone even though we still didn't have smartphones or maybe even Internet. Right?

[00:08:19] Nathan Pearson: Yeah. Yeah. Yeah. Definitely. And, I mean, I think I think there were even before, like, in proper Internet banking, I think there were some services that were sending out banking applications on, like, floppy disks or CD ROMs for people to do stuff on the computer before we went fully fully digital. So there've been some like there were lots of experiments. Like, I think banking has always wanted to experiment with elements of automation of leveraging the latest, greatest technology. It just hasn't always landed that well. But, yeah, I mean, Internet banking was, you know, became, you know, very popular in the, what, the probably early two thousands, I guess, when it started to become the norm. And then the twenty tens is when mobile banking became the norm. And I think mobile banking is when customers have really adopted, you know, that digital self-service. The fact that you've got, you know, a mobile phone in your pocket with all the the key key services available twenty four seven. You know? You don't need to boot up your laptop and do all the kind of login activities, you know, mobile with face ID and touch ID and everything like that that really made that kind of always on banking experience available. But, you know, really with the drive to self-service, you know, if you think about these, you know, web pages, mobile app designs, a lot of them are just, like, replicating the forms. Yeah. The paper forms that once existed in a branch, you would walk into and a member of staff would fill in all that detailed information. They just put that on a digital application and made the customer do it themselves that they're on the customer's own time, which is actually from an efficiency perspective, but, know, genius thing to do because, you know, therefore, you don't need to move your staff to do all that kind of stuff. It just builds some IT application that processes all of that, do some straight through processing. And, you know, the customer was working for the bank at that point, which is, uh, yeah, pretty good. You know, this is where I think,

[00:10:04] Nikola Mrkšić: like, you have to be somewhat nuanced. I've always struggled to kinda go, like, hey. Voice is the best interface or not. Like, even though, obviously, we've always been voice first, but not for everyone. Right? I think that the alternative to, like, filling out that form that's gonna have to be filled out is that, you know, you get a white gloved into a chitchat that takes twice as long and someone's filling it in for you, and you're standing there in the bank. Maybe you didn't wanna go there to begin with, or maybe it's over the phone, but still, you know, maybe you're on a train going through a tunnel. The line cuts. You call again. You can't find the same person. So, like, I feel like for younger generations, they might just want to punch through the form themselves. Whereas, like, someone older might be quite offended if they're forced to do it themselves rather than having someone take the time. So that's, I mean, that's really interesting. And that whole, like, equation between, like, pushing work out and you know? There are groups that will see that as functionality that they really want versus groups that get aggravated by the fact that they're asked to do it on their own. And, you know, especially when it comes to the channel mix for customer service, especially for banks that then become, you know, mobile first, but also, like, web banking available in case you're there. Maybe you have a branch. Maybe you don't. Increasing with fewer branches, I think, for even the largest banks and that, you know, the need is less prevalent. And then you look at all that and you're like, okay. When it there are banks like Revolut that have told me explicitly that they will never have a voice channel because that's for, like you know, no one wants that. And I don't think that's right either, right, for even, like, just the youngest of demographics because optionality is always better. Although there is research that says that having fewer CX channels tends to mean that you have better CSAT because, you know, you focus on doing those channels better is, I think, the truth. And you can't really know whether you'd have more of these or those customers if you invested in the other channels. But when you've seen these automation issues before and, you know, the the one big push from live person onwards has been to automate and chat and deal with automation there because it always worked better. What's your experience been with kind of, like, driving self-service through chat and then leaving phone as the nonautomated channel? Because I think it is largely unautomated. We we work with a few banks where we manage to go all the way through to complicated tasks, but it's an effort. And I dare say 90% of banks in North America and Europe do not have robust voice automation. Yeah.

[00:12:36] Nathan Pearson: Yeah. And I completely agree. And, I mean, I spent a long portion of my time in, you know, previous organizations, you know, driving that chat adoption as part of, you know, broader digital transformation programs. And the reality is is is what you said, right, that chat is chat is much easier to automate. You know, you don't have to worry about the complexities of voice, of speech recognition, of transcription, of so on and so forth. It was just, like, super, super hard to do, you know, before the, you know, the latest wave of technological change. So chat was the obvious thing to do. It's, you know, it's keeping customers within digital. Yeah. Everyone loves what's it's an easy story to sell, and you see really high customer adoption in most markets, like, across Europe, across The Middle East, across North America, kind of everywhere. It it has happened at kind of a similar pace. I think the the reality, though, in addition to the uptake, is you have not seen customer calls come down to net that off. So, you know, you've seen the you've seen a shift in the channel mix if you just look at percentages, but you've just that has resulted in, like, massive increase in chats and a, you know, a small decrease in voice as customers just start to self serve. So you've not really seen that substitution. You've just seen a growth in the one channel and a small decline in the other.

[00:13:57] Nikola Mrkšić: Yeah. Yeah. Interesting. So, I mean, yeah, I think that's, like, you know, it happened with, I think, kinda with petrol, like, Kevin's paradox where if you have more of a resource and then it gets cheaper, then you don't really decrease the overall spend on it. You just use more of it. People get, like, two, three cars and drive around more and more and consume more of it until that equilibrium is reached again. I think, equally, like, a lot of these interfaces have become more complicated. You know? You can do some things in I mean, with most banks, you can do some things on with online banking, other things with mobile, it's complicated. You get confused, so you end up calling more, and you need more support. So even though chat absorbs a lot of it, you still need more. Right? So that kinda human resource ends up still being used quite quite heavily. When When it comes to, like, large banks and the kind of, like, big successful initiatives, I mean, in terms of well branded efforts, I think Erika from Bank of America is probably the most prominent one, would spend estimated at least 50,000,000, maybe higher than that. Higher? Okay. Okay. Like, I mean, they've also branded it as their persona and stuff, so it's interesting. But do you feel like, you know, like, how would you assess the general state of play among, like, Western banks in in that self-service? And where do you think the Chantec AI is going to take over the next few years?

[00:15:16] Nathan Pearson: Yeah. I mean, I think I think most banks are on kind of a pretty similar footing, to be honest. I think mobile app has been the major focus for everyone universally. That's the story that all the consultancies have been selling. That's where most resource has been allocated in terms of transformation initiatives, I would say, across, you know, across the globe in terms of financial services. Everyone built chatbots, you know, five to ten years ago using deterministic models. All of those chatbots, you know, average to poor, I would say, in general, in terms of the, you know, the customer language, in terms of the usability. You know, the reality is and that's that's not down to, you know, bad implementation, bad design. It's generally just down to the limitations of the technology that was available at the time and the complexity of big legacy banks with big legacy systems integrating all their APIs and back end systems into them. So, you know, a lot of them will be FAQ driven with a small amount of fully, you know, fulfillment self-service use cases. But I would say some banks have gone bigger in terms of the persona and the kind of the PR around it. I think Bank of America have done a fantastic job there. The Wells Fargo have tried to kind of echo a lot of that with theirs, Fargo. Um, but it's still generally chat first and still generally deterministic. I think where we are, you know, now in early twenty twenty six, then we'll start to see that shift. You know, a lot of the conversations we've had with banks has been, you know, we're not ready for generative AI with customers yet. We will be, but not yet. Like, let's do custom let's do star facing things. Let's do pilots. Let's, you know, test everything out, and then we'll do it. We've got regulators. We've got all those concerns. I think the conversations over the last few months are really shifting, and I think it's now conversation is not we can't do this. It's we want to do this. How do we do it safely? And I think that's that's the shift we're seeing. That's the appetite the risk appetite shifting within these organizations. ChatGPT has been around for three years now. People understand it better. You know, the risk and regulatory teams understand it better. So I think that's where we're going to see, you know, a shift in the the appetite and the desire. And I think, you know, this will be a big year in in that in that shift.

[00:17:37] Nikola Mrkšić: Yeah. I mean, like, again, I think that, like, when you look at just the interest and, like, the adoption among the customer base, I think that a few years in now, there's a real kinda, like, taste for, you know, a persona behind, like, that interface and an understanding of what it sometimes does badly or not. Again, like, I think it's still evolving. I think it you know, just leaving it there in that shape and form without the human handoff is gonna be the real kinda, like, holy grail. I don't think we're close to that at all. But, equally, you know, we have a few banks already that have deployed fully all empowered systems. Now, again, what agentic means, you know, fulfillment and tasks being done absolutely as far as the integration kinda landscape allows in a given organization. But what about personalization? Like, how important is that for for, like, let's say, even just a a chat experience being automated?

[00:18:30] Nathan Pearson: Yeah. I mean, hyper personalization has been one of those buzzwords that's been banded around for, I don't know, five years, ten years. I I I lose track, and it's never been delivered. I don't think again, I there you see little elements of I don't know. You got a personalized push notification based on the offers that are available on your account versus me, but it's not, like, it's not personalized to you, Nicola. It's it's personalized generally to someone of your demographic, someone of your, like, product holdings, whatever else. So this, I think, is the big opportunity, like, to for a institution to, like, differentiate themselves from the competition. You know? What would the use case of, like, hyper personalization be? I think it's it's that proactive it's that proactive note like, nudging in terms of your behaviors. Like, obviously, there's, like, always a fine line between sort of, like, creepy and useful when it comes to personalization. But, you know, there's there's simple things that have been talked about for years and years. No one does it. Maybe Revolut do. I'm not a customer. But, you know, the stuff like, you know, you land in Dubai on holiday, and you get a notification saying, hey, Nicholas. See you see you in Dubai. Did you know this is the local FX rate? Actually, if you use this card, you're gonna get you pay no FX fees while you're here. Just those little things that are saying, hey. You're here. This is the thing that's gonna help you. Or, you know, hey. You've actually got some excess money in your current account. We know based on your bills, you know, you can transfer some money across to your savings. You get higher interest rate. Those kind of sort of behavioral things that everyone knows they should do, but it requires a bit of, like, cognitive load to do that. I think being able to do that, being able to, you know, go in, start some conversations. You know, the thing when it comes to calls is, generally, a customer who calls in has actually been doing something digitally beforehand. They've they've seen something. They've seen a Ford on the they've seen a transaction they don't recognize. They think it's fraudulent, and they call in. And then you call up, you go through IVR tree, you land with the Ford team, and they say, hi. Why are you calling today? And, really, you should just be able to say, hey, Nicola. I see you you've been looking at that transaction on your account. Is that something you don't recognize? Can I give you a bit more information about the transaction? Those are the the things where it's just, like, contextual. It's helpful. It just sort of, like, takes away some of the time of a conversation, but it also just takes away some of the emotion because it's like, oh, shit. They actually know me. They know what I'm looking for. They know what I'm doing. Oh, yeah. I mean, like, even if you just think of probably, like, one of the more frustrating call types for a

[00:21:09] Nikola Mrkšić: bank or for a customer of a bank is you're abroad, voila, your card is blocked. Right? So so, hey. Call us about it. And then you call and you're, like, call a number. It's, like, rarely done in ten seconds, and, really, it should be. And that's just, like, a point where you become a very high churn risk or at least, like, less likely. And very few people now, I think, have accounts with only one bank, especially with, like, the challenger banks and, you know, regular banks. And then you remortgage and you open an account with another bank I rep two. And then you've you've got real optionality that I think maybe wasn't as common as, I don't know, ten years ago, twenty years ago. And especially for a lot of these challenger banks, we've worked with a number of them. It's you grow a customer. You wait for them to start earning some respectable money. And, you know, I remember a bank's CIO told me very early on, and Polly and I's sister kinda like very proudly, it was the CIO of a challenger bank, and I compared them to another one. I won't name them here. And he was very proud that their average customer is five years older than the other bank's customer because that means quite a lot for, like, the value of that account holder in terms of their spend, propensity to get a mortgage, and things that actually make those max a lot of money. And when you nurture them for that long, and if they if early on, all they're doing is, you know, spending money at the pub, paying for their rent, they don't have a mortgage. They don't have, you know, school fees or anything like that. That's it's kinda like a bigger, kinda like mental burden. Things are great, and that app works. And then something happens. They have a large money transfer, and it's blocked for some reason. They can't reach anyone. Like, you just see a sudden spike of kind of, like, people churning to older school banks because they know that, you know, maybe their, you know, mother, father had a a Lloyd's account, a Barclays account, where, you know, they just kinda, like, believe that brick and mortar. You know? They've seen the that blue logo or the, like, black horse on the high street forever. And it's like, yeah. Like, okay. Like, that'll be there. Right? Although, I guess, 2008 taught us that that doesn't always have to be the case with all banks, but for the most part, it's held. Do you think, like, there is I mean, Revolut is obviously, like, a company worth 75,000,000,000 now. But where do you think, like, we converge, say, in The UK in terms of banking in ten years from now?

[00:23:34] Nathan Pearson: I mean, it's a it's a fascinating question because, you know, these these challenger banks have been around for, what, ten, fifteen years, some of them now. And they were always the great, you know, promise of, you know, they're gonna destroy and really disrupt traditional banking, and it's happened around the edges. Right? You know, people put, you know, open accounts with Monzo. They do some people take loans from Starling, but, generally, they you don't have these big banking relationships. I think because people like the trust safety, whatever, of the big banks, and the big banks have largely built apps that mimic the challenger bank apps. They all look relatively similar with different color schemes and slightly different designs, like, whether you're a NeoBank or a big bank these days. So the the disruption there hasn't happened. I think this is where the whole, like, agentic layer can completely shift it. You know? If you are now offering your customers, you know, your mass affluent customers or even just your mass banking customers, that kind of personalized assistance that is, you know, traditionally the kind of white glove service that, you know, a private banking customer only gets, where you've got a relationship manager who's calling you, who's practically doing stuff, who's managing your money for you, etcetera. If you're now able to offer that service to mass banking customers through an AI agent, if you get to market first, like, I think you're gonna get a big shift of people wanting to take up that service. And, yeah, they might only start testing it to begin with. They might not throw all their wealth at it and all their investments. But I think, eventually, like, being a first mover in this market, I think, actually can be something that is, you know, truly transformational within, you know, within the opportunity. Yeah. Yeah. Yeah. And I

[00:25:15] Nikola Mrkšić: guess for the context layer of knowing and having interface, I I guess open banking already paved the way for a lot of things that you would need from MCP or to do some of this. So I guess the question is just, like, who will move first? Well, the most exciting thing you've seen recently.

[00:25:29] Nathan Pearson: I mean, there's a few I mean, a few banks have done, like, things like being able to interrogate your finances using an agent, you know, asking, you know, basic things like, you know, how much money do I spend at Amazon last week or those kind of things, I think, are quite quite nice. But I think, generally speaking, there's there's not been that big moment yet. I think everyone is playing around the edges at the moment just because of that risk and the concern with using, you know, large language models and in a customer facing environment with a regulatory burden over the top is it's not quite there yet, but I think it's gonna happen soon. It's gonna it's suddenly gonna happen. Someone's going to do it. There'll be precedent, and then everyone will be diving straight in.

[00:26:12] Nikola Mrkšić: Yeah. Yeah. Yeah. No. Absolutely. Well, look, Nathan, thank you for joining me and talking about, you know, banking self-service in the history of of the field. And let's hope that we can do some really exciting stuff to write a chapter in it over the next few years. Definitely. Definitely. And thank you for having me. My pleasure. Thank you all for listening. Please like, share, subscribe, and we'll see you in the next one.