Transcription

• Speaker #0

  Welcome to FinTrends, the podcast series where we explore the hot trends and news in the financial sector with experts. Earlier this month, SBS launched SBS AI Foundation, the first product in its SBS AI portfolio. So following this launch, I am happy to welcome William Remak, head of SBS Data Platform. So he's here to talk about how AI is moving from experimentation to real-world banking use cases and what that means for financial institutions. William, welcome to this podcast.

• Speaker #1

  Thank you.

• Speaker #0

  So William, can you briefly introduce yourself and tell us how you are involved with SBS AI Foundation?

• Speaker #1

  Yeah, sure. So I'm William Remagler. My main objective indeed is to help customers to leverage AI in their existing ecosystem. So as part of SBS AI Foundation, I'm indeed in charge of the data part to make sure we are building, extracting the right data foundation to enable AI use cases.

• Speaker #0

  So interesting fact, only 34% of banks have scaled AI for a core process. So from your conversations with financial institutions, What is it? the biggest thing that is still holding them back.

• Speaker #1

  Yeah, indeed. That's a feedback that most of them share. They are building pilots, and then when they want to bring it, to deploy it in production, then comes the complication, the challenge. So I will take an exact example that the bank director shared with me recently. He wanted to enable relationship manager with AI. So they built an AI assistant for the relationship manager. They built a pilot. They started to demonstrate it, like give me a summary of my corporate customer I will meet tomorrow. Give me the recent operation, the risk, and eventually can I cross-sell another product to that corporate customer. And so it demonstrated very well. This is the deposit amount he has on his account. The loans have increased recently. The treasury is less and less. They just opened a new subsidiary in the United States. And so AI is suggesting maybe it's worth to propose them foreign exchange product to protect themselves against currency change. So it's a very good proposal. So in the bank, they were very enthusiastic. They saw that we need to give this all our relationship manager, this use case in their hands. And so they moved on the next step. Let's deploy and just realize and deploy in production. And there comes the problem because they've realized the data is there, It's fragmented because they have many applications in an ecosystem of a bank. And then even if they manage to unify it, At one place, the data, they realize that the data is of a certain quality. I will not say bad or good. Actually, they don't really know. But what they see is that they have data from core banking, from decades of operation. They have migrated from one system to another bank, maybe merge of banks. Some data is missing. They have a client from that product and from that product. they were not able to unify that client. So in the end, they have two clients in their data system. And so AI would understand they have two clients and not only one. And so they face those challenges and then they realize that it's not just building the AI use case, it's also what's behind and the data foundation that is really key to work upon. And so it's really important that... The data that we have, even if it's curated, clean, that we remove redundancy and so on, that we also have what we call AI-ready data.

• Speaker #0

  What do you mean AI-ready?

• Speaker #1

  Very good question and very important. It's not just about taking data and expose it to AI. No, we need to extract the data. We need to manipulate the data to make sure we have good quality, that it is clean, that we don't have redundancy of data. If I take a very quick example. In core banking, we maybe have 20, 30 times the data balance account, how would AI know which one I need to select? And this amount of data in a core banking makes perfectly sense. And so we are going to add data on data as well. So for the data, we will select the appropriate data and then we will describe it to the AI. So we bring the context to AI so he's able to understand what the data I have behind so I can play with.

• Speaker #0

  SBS has been working with banks for 50 years. How does that legacy translate into an AI advantage?

• Speaker #1

  So indeed, during those 50 years, we have built a really deep understanding of the banking business. We know what is a payment, we know what is a loan, what is a mortgage, and all this knowledge is embedded in our product. So we have our product, we unify. the data coming from those products. And we have built a banking semantic layer, an enterprise data model, which is described and so AI-ready to be exposed then to AI. And the AI vendors, they are not able to build that. They don't know our product, so they don't know how to get the data from it. And they don't know the banking domain either, as we do. And so they are not able to provide that. They are able to provide really good AI and models, but we are not providing that either.

• Speaker #0

  So to be very concrete, how would you describe what does SBS AI Foundation in plain terms for someone who is not deep into the industry?

• Speaker #1

  Yeah. So we have the example of the relationship manager. We see that for the relationship manager, it's a gain of time for the bank employee. So the AI prepares, gets. collect the data, prepare it and give it to the bank employee. So it's a gain of time. So the bank employee can focus on more complex tasks or focus on relationships with their customers, phone call or whatever. I will take also another example, still with bank employee. This time it's the support desk. They receive a call from their customers. a bit in panic. I've made twice payments, but it's a big amount. I'm really afraid. Can you do something? And then the support desk is also using an AI assistant. I have a client on call. He made apparently twice the same payment. Can you find it? He finds it. And then, say, the support desk person is saying, I recall one of the two payments. And during that time, he still... on the phone with the customer. And we all had that call with our bank, please wait a few seconds before I can come back to you. Now it's not happening anymore. So for the bank employee, it's about operational efficiency. But it's not only about that. It's also about SBS AI to help the bank to grow as well. It's to help them to give more personalized interaction to their customers. It's about taking faster decisions. If I take another example, I like example. Let's think about the marketing department. They want to, well, there is a new product they want to release. They want to TV spot. It costs a lot. They can see in real time the impact on that TV spot and take a decision accordingly. The day it is launched, either they stop. because there is no impact. Either they continue or they can even accelerate or just wait more time to have more data. Another example, which I would love, is to detect moment of life. If there is a wedding or a newborn or whatever, we could also detect that in the system, depending on the different merchant or whatever. And then raise an alert to the relationship manager again. Attention, this person. it's one of your important customers, by the way, maybe something is happening. Please, maybe send a message or a little gift or whatever, just to show that as a customer, I'm not a number, but I'm someone and my bank cares about me. And even more than that, I was speaking about growth, we can also imagine we detect certain patterns in the system that previous people... doing this pattern, at the end they left the bank. So of course, if they remove all cash from their account, it makes sense that the customer will leave the bank. But there are also other patterns that we could, I would say, detect way up front something happened. And so the bank can take action with their customer, maybe a promotion or a gift or a new product, whatever. It's the bank to decide to build more loyalty from the customer so they stay with the bank. And so we see that the number of use cases, it's... Actually, it's unlimited.

• Speaker #0

  Talking about use cases, SBS AI was launched with two use cases, helping customer relationship manager know their customer better and also letting banking teams generate insights faster. So why start with these two?

• Speaker #1

  So indeed, the relationship manager assistant and then prompt your data, these two use cases we made available. First, it's because we have listened to our customers. This is really a challenge that they are facing today. Secondly, it's also because it delivers immediate value to the customer. It's also, I would say, an easy use case compared to more complex, agentic use case that we are also working on. So I already spoke a lot about the relationship manager. So the objective there is to help the bank to gain in efficiency. So the bank employee can focus on more complex tasks and the relation with their customer and prompt your data that's actually requested by all our customers. And banks, they sit on a huge quantity of data and it's moving every day, a lot. And so the challenge they face, how can we extract this data and so we can use it afterward? So that's what we propose. But not only, we allow them to prompt dynamically. all the data they have there, but this data that has been prepared, not just a simple extraction, as I explained earlier in the previous question, so making them AI-ready. And so a bank from the business, he has no technical understanding to write a query in a technical system. He is asking in a natural language to the SVS AI, I would like to know the most. active customers in my base. And I want to have that on data now. So in real time, not data dating from a week ago. And what's happening today is that such business people, when they want this information, they need to ask one department. They have other priorities. It takes time. And most of the time they say, it takes us a week to get the information. Nowadays, it's already outdated.

• Speaker #0

  A lot of banks, well, I mean all banks, deal with huge amounts of sensitive customer data. How does SBS AI handle this responsibly? And what reassurance can you give a bank that is still worried about data security?

• Speaker #1

  It's a really good question and very important. We are in one of the most regulated industries. We need to make sure. What we do with the data, sensitive, personal data, making sure we have a system secure, that's really important. So first, what's important to know is that SBS AI is running where the core banking is already running, where the products are already running, in the same environment. The AI is not getting any interaction with Internet, other services, or whatever. It's locally hosted. on the same environment. And by the way, the customer can choose which AI they want to use, which LLM. And so the data, it's not moving that place. It's staying there. Then it's also a matter of there is encryption, access control, and so on. That's clear and that's there. That's important as well. But also the AI, it's not having wide access to everything. No, there is also... clear access right from the user. So bank, they have an organization, they have permission, they have role and so on. This is also the same with this AI. So if I come back on the relationship manager example, he will only be able to prompt or use the assistant on the data he is having access to. So this is already, this is configured also. on the data side independently from the AI, so not able to have access. And the last point, and that's more related to regulatory, the AI answer we have is not enough. We need to explain how the AI managed to get that answer with which data. And we should be able to, from that answer, to trace all the information, which data, which way it has been transformed from which source, who is owner of it, and what kind of governance we have around it. And that's what we do. Today with SBS AI Foundation.

• Speaker #0

  SBS AI Foundation is now available to select new clients with a broader rollout in early 2027. So what does getting 1,500 institutions into AI look like?

• Speaker #1

  Yeah. So indeed, we are selecting a couple of customers with those use cases. We want to learn, get feedback, working in parallel with... already other use case. And beginning next year, we will indeed start to, I would say, onboard other customers that would be interested. What's important to know as well, there are many use case AI. And all the use case AI is not going to fit all the banks at once. Each bank, they will have, okay, I want to work on this kind of priorities and other banks on other priorities. So we will have to see progressively which AI use case could fit some customers. And so it would rather be a progressive adoption, not going to be a big bang. It's also important to know from our point of view, there is from one side, prompt your data, get insight, the next best action, for example. But also then it's about building assistance that each kind of personnel in the bank could use. But then it's also about agentic. So agentic, either with human in the loop or even more complex without human in the loop. And we have to pay very attention to that because again, we are in a regulated industry. But agentic, it's taking action for me as a bank employee. Like the example with the support desk, having twice a payment, recall one of the two. So we are working on that and we need to make sure that it's... a really good answer. And then why not, even later, predictive AI use cases? So it's going to be a journey with customers, which AI fits them the best and to start small. But then there will be also the type of use case that will also evolve over the time.

• Speaker #0

  To conclude this podcast, what would you say to a bank leader that is still on the fence about AI? to convince them that now is the right moment to move.

• Speaker #1

  Yeah. I think everyone hears about AI every day, whatever banking, but all the domain, all the industry. So indeed now it's time because the market is also moving. If a bank does not move, competitors will do. And it's... They will be left behind because it's going to be a competitive advantage. And that's not me saying that. It's customers that already implemented successfully use case really see the value of it. But also, it's all the analysts. There is a consensus about that, that if you don't move with AI, you will be left behind and you will decrease with your growth. So it's quite important to do that. However, we don't need to start with all the use case at once. We can start small with a couple of use cases. And progressively, once we validate one or two use cases, indeed, it gives me a return. Then I start to move forward with other use cases. So this is going to be, I think, now it's the moment to start. That's also why we come with SBS AI Foundation now. But then also, it's going to be a journey for the customer with the use cases.

• Speaker #0

  Thank you so much, William, for being with us and sharing your insights. Thank you. you