Behavioural Purchase Loan. Scoring based on behavioral data gives a chance even to people whom banks would not lend, says Ondřej Šmakal of Orka Ventures
10. 10. 2022
Reading time: 10 minutes
A virtual assistant, Kate has served retail and corporate clients at ČSOB for two years. It is integrated into dozens of banking systems that run in the background, and its development based on artificial intelligence took ten months. Today, she can help clients in every other situation and is still learning. "Unlike a classic chatbot, Kate works in a modular way. It can work with huge amounts of data, compose further dialogues from previous ones, and doesn't derail even if you ask it something on the side," says Michaela Bauer, the ČSOB board member responsible for innovation and digitalization, in the Adastra podcast.
Ivana Karhanová: She is the second most famous assistant in the Czech Republic, a virtual assistant. To be precise, the first place is held by Apple's Siri. The second place belongs to Kate, the virtual assistant of ČSOB. Not only about her but also about whether we will work with concepts like "Banking as a Service" or "Banking as a Platform" in the Czech Republic, I will talk to Michaela Bauer, a member of the Board of Directors of ČSOB responsible for innovation and digitalization. Hello.
Michaela Bauer: Hello, thank you for having me.
Ivana Karhanová: While Siri operates Apple devices, Kate operates internet banking. Is that right?
Michaela Bauer: Kate mainly serves the client. We are currently working on the design and development of her form. Today you will find her mainly in our three mobile apps. One is Smart, which is our mobile banking for retail clients. The second is our lifestyle app DoKapsy, and the third is CEB Mobile, our corporate banking app.
Ivana Karhanová: How long does developing such a smart assistant take to develop?
Michaela Bauer: Kate is not a chatbot, but it is, as you rightly say, a smart assistant, so it takes quite a long time. But, on the other hand, the speed was quite high if you look at it through the lens of the clichés common about big banks. We started two years ago, and I also associate that with the start of Covid because we never physically put that team together. We developed it sometime between January and by June. We had the first prototype at the group level, so after six months, we got to the first real solution and launched in November 2020. So the speed for a big bank of our type was dizzying. And as I say, it's not just about interaction on some question/answer basis. We've developed something that's reaching into the system, integrated into dozens of our systems running in the background. And even from that perspective, it required quite a bit of work.
Ivana Karhanová: What does Kate currently know, and what does she still need to learn?
Michaela Bauer: Kate has quite a lot to learn. We started with a question-and-answer logic, but next to that is more important for us: a lot of so-called use cases or situations that Kate knows end-to-end. Today, she's already helped in two hundred different situations from banking processes. It's a mix of both sales and service processes. She'll help you with little things like a forgotten PIN, which she'll show to you voice-controlled on your screen. It will help you change the limit on your card and help you pay. We can do features that maybe only Revolut can do, like split the payment when you're out to dinner with friends somewhere. So it can do many everyday banking things, but it's also helpful in insurance. Instead of the earlier forms where you would enter something, you can now arrange new products or changes on existing products, like credit limits and things like that, through the conversational interface.
Ivana Karhanová: Where are Kate's limits currently? What hasn't the algorithm mastered yet, and what does it still have to learn? And will it still take time before you can get it up and running?
Michaela Bauer: We are constantly working on understanding. The Czech that Kate uses is very good. I have to say that we are using a module that we buy from Google. She's sitting in the Amazon cloud herself. So it's not like we're developing everything from scratch, but we're using the good stuff on the market. And for example, we chose Kate's voice so that it would work for Czech and equally well in other KBC countries so that we had one solution for the whole group. So it has to suit Hungarian, Bulgarian and similar exotic languages, and Flemish. By the way, we don't have classical Dutch there, but since our mother country is Belgium, we also have Flemish Kate. I would say that the limits are not in comprehension, although I started talking about that, but in voice recognition. So we teach Kate in different situations to understand what the client might want, whether they speak with different accents, and of course, we compose sentences differently. To give an example, it's relatively easy to mix up a request for a card. A live banker will understand from the context what card the client means, whereas Kate may not always be clear whether the client is asking for a credit card or, say, a green card related to insurance. So there is a room from that perspective. I wouldn't say it's completely the limit. It's just about time and energy.
Ivana Karhanová: But they're learning that context from the data, right? And so, as the data increases, Kate's understanding will increase.
Michaela Bauer: That's right. We have a lot of data, of course, like most institutions of our type. Part of it is voice recordings, but part of it is the areas that Kate is gradually processing and that we're teaching her. So, yes, Kate is all about data, which is another big difference from, say, a traditional chatbot because Kate can work with a huge amount of data, and even the conversations we have are not classic chatbot questions. Still, Kate can work modularly, composing other dialogues from previous dialogues, and she doesn't get derailed even if you ask a side question. She can get back into the main flow of ideas.
Ivana Karhanová: What percentage of customers are using it now?
Michaela Bauer: I can't tell you the percentage off the top of my head right now, but our ambition for this year is to get to 50 percent of all customers using mobile banking because that's where Kate mainly lives today. Fifty percent, in our case, means half a million clients. As of today, over three 350,000 clients have tried it, and many of them are active users. To make it comprehensive enough for every client in different situations, whether they're many or the few you deal with once in a lifetime, our ambition is to get to a few hundred of those processes. We have both retail and corporate Kate, and if you break it down into those two segments, for example, we still don't have enough processes to have a solution to help a client in virtually every situation they can think of. Today, from what we've taught Kate so far, I dare say it will help in every other situation. And that was roughly our benchmark for where we are today. Our ambition is to service eighty percent of all client requests by voice or chat without any intervention from a physical person.
Ivana Karhanová: At ČSOB, you also measure an innovation index. The last measurement showed that people miss innovations in public administration, health care, education, and infrastructure the most, while in the banking sector, they welcome innovations. And even according to the survey, they don't seem to be afraid of them.
Michaela Bauer: What is pleasing is that more than three-quarters of Czechs welcome innovation in general or change and novelty, depending on how you define innovation. Here we asked a thousand individuals from five hundred companies. I'm sure each of us imagines something different through innovation. Unfortunately, this is probably not the innovation of the type.
Ivana Karhanová: Amazon, cloud, nuclear science...
Michaela Bauer: These are more like everyday improvements, everyday innovations. So in that sense, three-quarters of Czechs are not afraid of innovation, and only three percent of clients and people in the country reject it. And as far as the banking sector is concerned, where the situation is satisfactory, the index compares the difference between expectations and reality in many areas of everyday - everyday human or corporate life. Indeed, in the banking sector, even reality seems to exceed people's expectations. So it is not a question of being scared or not scared, but it does show that banks are innovating, and clients tend to be pleasantly surprised. This is probably most evident in the payments area, where the Czech Republic has traditionally been one of the countries where people have used various payment innovations historically since contactless payments. We were number one in the world, and maybe we still are, because it has taken off so quickly in our country. Clients like to get used to these innovations and work with them.
Ivana Karhanová: According to your press release, the index reflects the need for innovation and shows that people are demanding much more from the government. One of the things that have been successful in that state administration is, in part, to integrate Bank ID into various login systems. But in the beginning, it was a collaboration of all the big banking houses. How were you able to get across those houses, which on the surface look like relatively competing entities, to come to the table and push a solution that was able to meet one of the steps that we would want from the state government?
Michaela Bauer: Well, that's exactly the other side of it, where, of course, much more is expected of the state government. Technology today is making it possible to digitize and make available to people what they are already largely used to in the private sector, whether in banks or elsewhere. And so far, we are not succeeding everywhere. Now I mean in different sectors of the state machine. Bank Identity was intended from the beginning to leverage what we have in banks in some way for what clients need elsewhere.
Each bank that eventually became either founding or joining member solved a similar problem. Or maybe not a problem for us, but a similar situation. And they had a similar idea. Inspired by various Scandinavian, Canadian, and other models, we thought it was a real shame that clients identify themselves in different ways in different places. Of course, we were not achieving the kind of digitalization that has been achieved in some of the Baltic republics. And as banks, we had a unique opportunity, an opportunity and we were glad that the state had heard about it that you could use just that secure physical identification that happens in the bank to transfer that identity to any other point in a person's life, whether it's to the government or other commercial entities. I do not recall the exact moment when and how we agreed, but it is true that at the level of the Czech Banking Association, the subject was discussed for quite a long time before the agreement was reached. Of course, the idea was to harmonize the different entities with different starting situations. In the end, however, the common idea prevailed that if each of us produced our system, it would only partially work. In that sense, only the interests were interconnected, and not the client databases were integrated. The overlap was mainly between the first three big banks, and a few smaller, perhaps, even more, digital banks came in. In doing so, we have achieved something the state has not managed for several years, although it may have tried. We say this is the simple motto of banking identity: you already have that password or identity. Every one of us who is a customer of one of the ten participating banks has an identity. And when that portfolio overlaps, you have virtually nobody falling out of the system. Each of us has or can have that identity. And you don't have to log in again or go anywhere. We saw an opportunity there for digitization, and for our banks, a digitized economy is thriving and profitable.
Ivana Karhanová: From your perspective, what is the biggest technology or innovation challenge for banks right now? It's quite related.
Michaela Bauer: There are probably quite a lot of those things, but it depends on your starting position. If I speak for the larger and maybe more established and older banks that didn't have the advantage of building on greenfield sites, there's a huge historical burden. We're working with many different, even legacy, systems that we're trying to change over time.
Ivana Karhanová: But these are often backbone systems, banks.
Michaela Bauer: Yes, they are often the bank's backbone systems. At the same time, of course, we must develop new things in parallel, like the ones we discussed. But few of us can afford to stop operations for five years and devote ourselves to housekeeping. So we develop in parallel, and we clean in parallel. So I would say this is one of the big challenges. The second thing, which is then a bit related to regulation, is the various cloud technologies we have also mentioned here today. Because we're an extremely regulated environment, we have a huge responsibility for client data, and even shared responsibility is quite difficult for us. So I dare say it's not that much of a technology challenge. I don't want to make light of the move to the cloud in any way, we've been through it ourselves, but maybe that's the easier side. But how do you cope with all that new technology has to offer? With regulation, which is largely - and for good reasons - getting tighter. And at the same time, I would mention this as a third one. There are various security threats. I don't want to bring it up. I'm going to knock on wood here. Still, those classic muggings don't happen as much in the banking sector today because it's, I don't want to say, easier, but at least it's perceived as more secure to sit anywhere in the world behind your computer and try to attack various banks as well as other institutions remotely. So it's a balance between what the technology can already do today and, at the same time, a centrally given sort of environment to protect the user. At the same time, our responsibility is to protect the user and the data because, of course, that's at the core of trust towards banks in general, which we logically need to maintain at all costs.
Ivana Karhanová: You mention the protection of the user and their data. On the other hand, from my point of view, this is sometimes countered by the users' demand that the bank does not bother them with irrelevant messages. But this exactly requires a thorough knowledge of that client's data. Do you also perceive a contradiction between what clients would like and who, on the other hand, fear that the bank or any other entity will misuse their data?
Michaela Bauer: I think the times are generally very unpredictable. Two or three years ago, we all thought we knew what was going to happen, but it turns out that we don't know anything. What individual clients want is logically one thing. Here too, the situation is unclear because different philosophies of life and approaches are clashing. I have been in banking for twenty years, and I don't want to say things are any worse now. It is different than it used to be, so we are dealing with different issues again. But the way we partly have to protect our clients because regulation tells us to, and partly want to because we don't even want to admit any unnecessary wrongdoing. It doesn't pay off in a positive way sometimes. It all comes down to the fact that the reports and the various disclaimers should be longer and longer. At the same time, we are trying to make them shorter and shorter as part of user-friendliness. The same applies to the various acknowledgments, second factors, etc. When you talk to any compliance department in a bank or audit or any other risk department trying to protect us from those risks, it would be better to do nothing.
Ivana Karhanová: Now it occurred to me that since we can all decline cookies, we acknowledge them anyway because it's a pain in the ass to click what I don't want.
Michaela Bauer: Yeah, it teeters on edge sometimes, whereby overwhelming and overprotecting, maybe we're reducing real security. But let's think of a more standard situation. Our goal is not only to surprise the client in a user-friendly way, that is, to make their life as easy as possible, so that with a few clicks, you can have practically anything you need with the bank, but the various security measures go a little bit in the opposite direction. So technology is there as a great enabler, but unfortunately, the easier the access is for the client, if it falls into the wrong hands, then the easier the access is for anyone else.
Ivana Karhanová: I mentioned at the beginning the two concepts of Banking as a Service and Banking as a Platform. If I understand correctly, the bank can offer its environment to various fintech startups, which can offer their services to clients through the bank. But on the other hand, Banking as a Service is an approach where individual startups can use the bank's clients. Do you think something like that will be developed more? Even in the Czech Republic, it seems to me that there is a bit of a Chinese wall between fintech and banks.
Michaela Bauer: I guess there are more definitions. The one you offered is one of them. I think a lot is going on in the Czech Republic. I'll start with the fact that most large and medium-sized banks are working with fintech and startups in some way. Each of us has chosen a slightly different path, but most of us in the banking sector have at least two cooperation models. First, most have invested, so we co-invest in a fintech depending on how it fits into our strategies, whether in terms of different gadgets for corporate or retail clients. And most of us work together on a commercial basis. The third thing we do at CSOB helps incubate startups. We do that as part of our CSR pillar. We don't ask for anything in return and don't take a stake in that particular company, which is why I mention it as a separate category. But it still makes us happy because we really manage to kick-start a lot of young and older startups and steer them in the right direction. I also consider it a success when some of them eventually give up. For example, thanks to our incubator, they have verified that their business model or idea is not viable. So before he or she drowns a few million and...
Ivana Karhanová:...a few years of life.
Michaela Bauer: And a couple of years of life, so they realize it. With the experts they have at their disposal, whether from us in the group or even outsiders who advise them in a safe environment without the pressure of collaboration and performance, making money. All of this, I think, is very helpful. We have higher dozens of applications for each of those incubation rounds. We do it twice a year, pick those startups out of eight or ten, and some great companies have gone on to get multi-million dollar investments. So I'm really happy about that. I love it because it's still a little bit of a separate activity from some of our startups. We're also looking at fintech more broadly and establishing commercial collaborations. So, yes, it is true that by partnering commercially, they are getting to quite a large base, a large volume of business, because when their solution is deployed in a group like CSOB or KBC, a lot of startups can only dream about that.
Ivana Karhanová: Can you mention startups' services or products in this way?
Michaela Bauer: I can mention a specific company called Resistant AI, which detects fraud based on artificial intelligence. Resistant AI even won a tender. Not because they were in our incubator but because they are so good. We only helped ourselves by knowing about them. Today, they scan millions of documents, not only at CSOB here in the Czech Republic but other countries as well. We can tell if a document is genuine or if it shows any signs of possible alteration thanks to artificial intelligence. But not to stay just with startups, I want to say that commercial cooperation is quite common, and investment participation - whether in our bank or other banks - is also becoming quite common. We try to not only look at trends and look for startups that fit the banking industry in some way but are expanding the area of interest and already have something that the bank doesn't have or would take quite a long time to develop if they were developing it themselves. If I give you an example from, for example, our rank, I get to Banking as a Service because one of those services is Buy Now, Pay Later, which is a deferred payment. This product comes from Scandinavia. Its most famous representative is the company Klarna, which has entered the Czech market. We have Mall Pay here, which we bought into the group two or three years ago. Since October, we have renamed it Skip Pay. With Twist, which started a few years before us, we have been the most established on the market so far. After some changes on our partner's side, we have bought a 100 percent stake from Mall Group. Skip Pay is a service that today if we were to develop it from scratch, would take us several years. This way, we had, I don't want to say, a finished product because the product is still evolving, but it can be used in different forms as a kind of embedded finance. They're financial services embedded in the normal, whatever other processes people go through. So the bank, in this philosophy and Banking as a Service, is to be invisible and there in the form of service and where the client is. That's roughly what Banking as a Service is all about. Mostly it is derived from payment as such, and at the same time, you can build on that with various savings or investment products and lending products.
Ivana Karhanová: This is how Michaela Bauer, a Member of the Board of Directors of ČSOB responsible for innovation and digitalization, describes the cooperation between the banking sector, startups, and the fintech scene. Thank you for coming to the studio and sometimes for listening.
Michaela Bauer: Thank you. Goodbye.