ING Groep N.V. ($INGA)

Good afternoon, everyone. I'm here today with Marnix van Stiphout, COO of ING. Marnix, hello. Thank you for being with us.

Thank you. Pleasure.

But before I start with my questions, I have the usual following question. So how will AI Impact ING's earnings growth trajectory over the medium term. Strong positive, moderate positive, neutral, negative impact due to the higher cost of deposits or is it too early to tell? Well, okay, and zero negative impact. That's quite interesting. All right. We will get back to this call. But first, we don't often see a COO at Morgan Stanley Conference. And I'm very excited to have you because I think this will be a really interesting fireside chat. But what led you to come?

Well, the invitation joking apart, Look, it's -- there's so much happening clearly. And I think operational excellence the link to customer experience, the take-up, the rollout, the take-up of AI is all sitting in our remit. So yes, this is -- if there's one time is relevant, it's today, I think. So that's the reason.

Absolutely. And you published the presentation this morning about the scalability of ING's operations. But before we go into the details of this presentation, can you perhaps help us understand your role at ING and how crucial the operations team is in ING's strategy.

Sure. Yes, because it's good to go about it because the definition of COO differs across the industry, obviously. So what we got excluded from COO at ING is the technology itself. But what we include is all of the operations, business and support, so the wholesale, the retail and all the support functions. It includes all payments. So we've got a factory that sets is with us in COO. It includes all AML KYC. It includes all data and it includes all Analytics/AI that we do for ourselves, but for the bank overall, sits within COO. And we do all the transformation for ING. And it's in total, maybe just to give people a bit of an idea of scale, it's about 20,000 people, and that's kind of just shy of 1/3 of the group.

Very clear. And so staying on the topical ING growth strategy, how do you achieve scalable growth?

Yes. So look, it's -- we've got a pretty good trajectory over the last couple of years in track record actually in terms of growing wholesale and retail. We have been a digital bank for a very long time. That's how we've got known, I guess, to many people around here in the room and beyond as a challenger. And I think we still benefit from the fact we've been relentless in digitalization, in straight through processing, as we call it, STP. We even carry an index internally, which we also publish on how STP we are across our main journeys in the bank, 350 journeys we measure. Secondly, we have pulled together a lot of our capability in terms of people in hubs, 6 countries, 7 locations where we pull technology and operations teams together to work in an agile way, in a flexible way across the different priorities, across different countries, so not linked to a specific goal or a specific piece or a specific country. That has given us a lot of flexibility. It's about 14,000 people in total today, across, again, technology and operations. And of course, with the front runner that we are I dare to say, in AI. And I think if I give you a couple of examples of how this has made us scalable, right, in operations in a broader sense in the technology. We've added about 1 million primary mobile clients over the last couple of years per annum. We have gained productivity in, for example, mortgages, about 50% over the last couple of years per FTE, 28% on the other side of the spectrum in Wholesale Banking lending. We have been able to reduce friction for our clients, resulting in a 40-plus percent reduction in flow coming into our contact centers. And so basically a little bit of a long story short, we have been able to assume new clients more volume and our cost base and FTE count in operations has reduced. So that's what I call true scalability.

Perfect. Can I just go back to something that you said at the beginning, and quickly clarify because you said tech is excluded from the COO. Why is that? And how do you interact with them with the tax side of things?

Yes. So it's good that you asked. I mean when I talk, for example, about analytics, which is in our scope, right in my scope, then the technology side of analytics is within the remit of our Chief Analytics Officer. So he, in this case, works with a full team of about 600 people globally, which I call direct people in COO, but also people in technology. But the functional responsibility for technology sits with the CTO of the bank. But everywhere, we need technology colleagues to help us in, for example, KYC, they work together, of course, with these teams on an integrated basis.

Perfect. So then -- and let's go back to the cost base of the bank. Some banks are particularly bloated or have a very wide branch network. We call them this these, the low-hanging fruit, but it doesn't look like ING has this sort of low-hanging fruit. So where do you see the biggest opportunities on the cost base?

Yes. We still got a lot of fruit, though. You might not see this low hang, but we still got a lot of fruit. And so just to give you some ideas. We still employ 6,000 people just a bit lower than that is a bit shy again of 6,000 people in KYC. We are multi country retail bank. We've got 10 franchises, deep franchises, and that has basically meant a lot of KYC work next to our wholesale. And there, we have actually got a big opportunity through what we call behavioral modeling, which we have rolled out, which is AI-based, to take out 80% of our manual fallouts. Well, of course, you can calculate that or at least you can imagine that the 6,000 people are not there to do non-manual work, they're largely doing manual work. So that's a huge opportunity for us, and that's what we are working on today as we speak.

You said 80 or 18.

80, yes of the manual fallout, we will be taking care of, i.e., they'll become STP, so to say.

Very true.

Okay. And the second example where we still got a lot of work to be done or a lot of benefit to be had is in our contact centers. It's about 2,500 people strong. We have, as I said earlier, reduced volumes by 40%. So it's already shrunk quite considerably. We've also got a GenAI chatbot that is giving us an 80% success rate in non-manual resolution of requests. But we're going to use conversational AI by the middle of this year to make that environment even more productive. And I think what's important to mention, both on the KYC side and the contact center side, there's always 2 conversations. There's the client impact conversation. So on the KYC side, it will also help us onboard clients far more smoothly. And then there is the productivity side, which helps us with the cycle of doing KYC. And on the contact center side, same thing. It helps us to convert clients. For example, with conversational AI. You draw in nondigital clients far more easily through that approach, then we can do that today. And of course, there is the productivity benefit on the other side, too. So it's always 2 sides of the same coin.

Very interesting, especially the 80%. Okay. So let's discuss AI now. Do you think this is an opportunity for ING? And how do you win? Or how do you use AI to win?

Yes. It would be quite funny if I would be the 0% change here. But no, I think it's a huge opportunity. And look, I think it's important to start with, we have been working, as I said earlier, on digital propositions for the last couple of decades, right? So we've done a lot of work on our data. We're sharing a lot of solutions like the app across all of our markets and our data solutions as well. So we are well positioned from that perspective. I think what is another important topic to talk about is how we organize for it, right? So we got a single team in analytics in AI across the world, as I just referenced. More than 95% of all the work sits in that team for us. We consolidate the work as much as we can in that team for consistency and quality and monitoring purposes. We also produce the work on a single platform. So we don't have leakage because people have got different platforms in different countries. We produce it all on a single platform. We got a very strict view on what work we work on. We've got an AI KPI that is carried by me as the COO and my teams by the business in retail by the business in wholesale and also by technology. So we got like we did with the digitalization STP, we got joint KPIs, which helps us greatly. And a further point is that we monitor this work on a monthly basis, meaning there's board involvement, including myself, with all of the expos across all of our markets. Every month to monitor what's happening, what is the progress? What are the new ideas and to vet new ideas into potentially add-ons to the portfolio. So my point, ultimately, it's very strictly organized and we look at this as normal execution. Of course, it's novel in terms of its content, but in terms of how we execute, it's very regimental you could argue. So I think our heritage and our approach gives a very, very, very important starting point.

Can I follow up on something you just said? You mentioned AI KPIs. So can you make it a bit more real for us? Can you give us some examples here?

Well, it ranges from the introduction of more machine learning models, right, in the propensity, but also our consumer lending models, our churn models, our e-banking models against fraud. So that's one end of the spectrum. And on the other end of the spectrum, we got in the KPI, the conversational AI launch, for example, right? So it's literally from machine learning to GenAI to Agentic AI all sitting in that space in that KPI.

Perfect. And so the KPI is how quick can you launch it and how effective it is?

We just got milestones on all of these topics for 2026 and 2027. So we just followed the plan.

Excellent. And can you -- you talked about scalability earlier, right? So can you talk us through how scalable your platform is? How much of a competitive advantage this is, how quickly can you scale when you have an AI use case?

Yes, yes. Look, I think it's very much linked also to what I just said, right. So first all, is the organization on how we operate, which makes it that the choices that we make and how we then make them, 90% of what we start get scaled, right? So there's no hobbying, and I don't want to say that other people are necessarily hobbying, but there is not a lot of room for people to innovate on their own. There is an agenda. We follow the agenda and 90% of that get scaled. There is a very strict risk and compliance approach to it, which is also centralized. So we don't have that kind of decision-making looked at by risk, for example, jurisdiction by jurisdiction. We got that consolidated, too. So I talked about the single team, all the work flowing through it, the single platform and also the single risk approach. And just to give you an idea on the scalability of things, I talked about the contact center, but maybe also in product development. We have just launched Agentic AI in Dutch mortgages. In parallel, we're working on -- we're using that solutions for German mortgages, and we'll take that to other countries too. So my point is, we really aim and we're succeeding so far in making sure that we get the benefit of these implementations across the different markets.

Perfect. And a follow-up again. So Agentic AI in Dutch mortgages, how does that help? Can you give us some examples?

So we've grown market share quite considerably over the last couple of years. And we are a big Dutch Bank. We have 40% of the Dutch population banks with ING. And still, and that's something to be really happy about, at least I am. We're still add clients and we still add content and products. So we're still growing the bank is my point. On mortgages, actually growing mortgage book is quite cumbersome, meaning you need well-trained experienced people to add mortgages to your book. And actually, we can't do that anymore. We're just adding people. It's just -- that's too slow. The Agentic solution that we've just launched a couple of weeks ago and will continue to add to, gives us about 10% to 15% productivity gain. So we can add volume with our people, so to say. That will grow, and that's a fantastic delivery for us. I've seen it live a couple of times now. And yes, it's very good. And maybe to talk a bit about the German side of things that gets launched slightly later still in the next couple of months. And that gets us by early '27 to a below 1 day time to yes for the full German portfolio of mortgages. So that for us in the German large German market is a massive thing to achieve.

Perfect. And I guess the Agentic side of things helps you with the mortgages in terms of all the checks on the property on the collateral on the borrower?

Correct, correct.

Great. And so let's talk about FTEs then because earlier, I was quite impressed by the 6,000 people, 80% potential savings. But if I look at the FTEs of ING over the past 5 years, that has actually increased. So if we look forward, how do you expect AI to impact the overall number of ING?

So maybe just a bit of context. We have built -- of course, that's now in the past, we built quite a bit of capacity on the KYC side, given all that we have to do. We have built our hubs that I just talked about where we basically took external people out, third parties out and build captives. We are a big believer, I personally, but also us as a team that we want to do these things ourselves with our own people. We believe that's better for the continuity, its better for the experience and it's much higher quality output. So that has added to the FTE base. But to be very specific about today and tomorrow, we have reduced in operations in the largest in a broader sense of the world, I should say, about 1,000 people last year. We are reducing about 1,250 this year, and that will continue over the next couple of years. And you ask me, okay, for the group overall, we're still growing, right? We're adding segments content to different markets, business banking in Spain, in Italy and Germany, Australia, affluent. That needs people. Also engineering needs more people, but we are balancing that of funding that you could argue through reductions on the COO side.

So let's get to that 0%. There is a thesis in the market, although perhaps not in this room that deposits could be disrupted. And within deposits, specifically, I would say, the savings account. Not really the current account on that one is more sticky if you have your payroll. But when it comes to savings in terms where actually ING is particularly strong that lends itself, perhaps it's a bit more to be disrupted, especially if you are a digital native and so how do you respond to this clip?

Well, look, I see this a bit differently, to be honest. First of all, we've got a broad client base. And if you look at it per customer, it's around 15,000 deposits, right, on average. So these are small tickets. So less vulnerable for individual big shifts. That's the starting point. Second point is we've seen our deposit base very resilient over the periods, right? So even under considerable change in the markets, we have been very consistent. Further point is pricing-wise, we're sitting largely in the middle of the pack. We're not very high, so exposed in that sense. But I think even more important, as a digital challenger, we have got enormous capacity to detect. We got very good elasticity and churn models that pick up these kind of developments if they occur very early in the process. We got very personal response opportunities, because we have put also model AI developments there, which help us to have personalized marketing and offers to our clients, which basically gives us very good returns at this stage. So my point ultimately is that I think we are quite well positioned to be on the advantage side of this conversation rather than anything else.

But today, if such a tool exists if that Chat GPT, for example, launches a deposit optimization tool, could that operationally happen in Europe?

Well, I think -- look, it can happen, but even with the current tooling like raising offers, right? It's not as prolific or dynamic as people maybe believe it should be. I think it's also important to say, we're adding, when I talk about 1 million clients per annum, we talk about primary mobile clients who do a lot more with us than just putting some money with right? So there is a lot deeper relationship and more products that people work with at ING than just a deposit with a bit of an interest rate. So I think that also helps us protect what we've got today and actually build what we got today.

Got it. And so you have the responsibility for AI within the bank. It's quite a big role these days. So how do you ensure and innovative, but responsible, I would say, AI culture when you roll out new applications.

Yes. So I think first, back to what I said earlier, the way we organize for this is per our normal execution rhythm, right? So the risk we assess, the way we control things, just the consolidation of the work. So that makes it well controlled to start with. What I find very important in terms of controlling what we're doing is that we said, let's make sure that from a credit nonfinancial and compliance risk perspective, we have this single team in risk. So with the CRO, we said, let's build this. Liliano who is now Head of Wholesale Banking, the previous CRO, I was very keen to do this too. So we got at least a very well-developed capability in the bank for risk assessment, so to say, right, and decision-making. And that's all based on a process that includes 140 risk parameters that we always use for these assessments. So that's pretty solid. And last but not least...

Sorry, parameters, can you give us some examples?

Well, I mean, the hallucination obviously, is an important topic right? And just to give you an idea, before launch, we look at hallucination, but we continue to test a share of the go-live flows in perpetuity, so to say, on a manual basis to make sure hallucination stays away. So those are the kind of the controls that we put in place. So it's a very serious how we take that role also in production, not just before production as an example. And last but not the least, I think it's the whole education. We've spent a lot of time and money, by the way, on reeducating our workforce on the junior and the senior level, and that will continue also from a risk perspective.

Makes sense. So I want to turn to the financial impact for ING. But before I do that, let's see if there are any questions? There is one.

We've heard a number of banks talk about the potential for AI benefits to be competed away over time through get backs in rates or effectively being shared with the consumer in some shape or form. What's your view on how much could be shared what the triggers for that will be and maybe what the time frame for that will be, please?

Yes. So I guess the conversation about what will happen with these benefits, right? I mean there's a point of what will happen with the price of AI, I guess, is a question what will happen with consumers? Do you give something back to your clients? And the third point that's floating around, obviously, is -- is this a big reset for all of us? Are we all getting back to the same kind of starting point, right? And I want to start with that point. I think well, anyone who's telling me, let me speak for myself, anyone was telling me that -- this will allow you to basically do away with all your legacy and Sprint forward like everybody else who is well prepared, I think that's nonsense. My point being, I think the fact that we have spent a lot of time curating our data platform, data quality that we use across the bank rather than having all these different disperse sets of data and data platforms is giving us a very good and better starting point than many others. Because there's less legacy. I think that is important. And a lot of legacy means there is less consistent data, less data quality, less prepared to get this going at scale. So I think that's an important point, I think, to make first. Then there is a pricing point about, okay, you're using -- we use Google, for example, won't the price go up prohibitively and will eat all the benefits? Well, I think a couple of things that we are looking at. We're looking at using more different models than just a single Google LLM, so to say, right? So there is also a development towards smaller, more focused models for more focused business cases, that should make processing lighter and the business case better. But also, let's be clear, we're not going to throw Agentic AI at everything. It's going to be very expensive if that's what we do. So we're going to be far more selective than some people think maybe in how we apply this. So mortgages and in Germany, I just [indiscernible] is a big thing for us, right? So a big thing deserves a big solution from an income and a cost perspective. This is more income than cost by the way, on mortgages. And I think that will -- so when we go beyond the KYC and the contact center solution, when we get to smaller and I'm not saying everything else is a lot smaller, but when we get a small end of the business cases will take probably different decisions. And I think last, but at least on the consumer side, do we give benefit back to consumers. I think it's a bit too early to call, but what I dare to say is the speed by which we bring, for example, conversational AI and other solutions to the market will allow us to deepen our relationships with customers and that allow us to, I think, give some money back ultimately to customers with deeper relationships and more clients coming to ING.

Perfect. Let me see -- okay, we have 2 more questions. So here, let's start at the corner and then we go on.

You've obviously mentioned how much AI can help in terms of cost takeout and kind of reducing head count. You're not the only Dutch bank kind of talking about those kind of programs. And I'm sure you're not the only corporation just in Europe talking about this. How much are you thinking about the potential risks to I guess, lending demand on the retail side, kind of mortgage customers being at risk of losing their jobs if a lot of corporates kind of just replace people with machines. Is that something you're already considering? Or if not, why not?

Well, we're certainly thinking about it. But the more immediate points for me, for us is I gave just examples of 2025, 1000 people, 1,250 in 2026, and that will continue over the next couple of years from a COO perspective alone. We are very serious about the social responsibility that we've got. So the other side, maybe the precursor to your question. And to take that role very, we do a couple of things. First of all, we reeducate everybody who is in scope and beyond the scope by the way, whether we believe they're going to be with us or not, we're going to give people some serious reeducation, reskilling today, tomorrow and the day after. So that's what we're investing in heavily. The second point that we're doing, the second point I want to make is we're making a lot of extra investment and a lot of efforts with third parties to get very well-trained ING people in operations who really sought after in terms of their skills to other sectors to health, for example. So that is something that's happening as we speak. We're doing this across markets. And we're trying to get people out of banking into other areas of the economy where these skills are really going to be very useful. So I'm not sure that's going to alleviate all the joblessness that kind of within your question. But at least, I think it's important that there is opportunity for these people beyond the job at ING, and that will help certainly this point.

Yes, we have another question.

So obviously, just given the developments in AI and how it can improve a lot in terms of efficiency, but also -- we got headlines, I think, last week or the week before, from Australia where a lot of cyber fraud was being used and AI was being used in cyber fraud. In terms of protecting yourself against that what does that look like? And also, is it only going to be manifested in better detection? Or will that be also reflected in provisioning and whatnot?

What you said -- what was the last thing you said also?

In provisioning or in terms of credit losses at LLPs?

So well, first of all, we're doing a lot of investments in fraud, anti-fraud and AI also is helping there, obviously. But the e-banking modeling and all those kind of things which are not always easy to do in terms of data provisioning, but there's a lot of money going into new technology for prevention, detection. Of course, PSD3 is coming out right across Europe, which gives us more of a liability for impersonation across banking, so that we know, obviously. But what is good to say when we put forgery tools in, for example, document forgery tools or when we put a new e-banking model in or we put stricter enrollment technology or process in it gives us a benefit. So we see that fraud numbers are coming off, are coming down as soon as we do these things. Obviously, fraud are pretty flexible and agile in their thinking. So we need to keep on running with them. But my point is, we know how to respond. We know where to invest, how to build it and to make sure that those flows reduce, but it is a serious thing for sure.

Right. Do we have other questions? Okay. If not, I'm going to continue because we get now to the financial impact for ING. So in terms of cost, you have high 93% cost benefit or $350 million. So which areas do these cost benefits come from?

Well, it's mainly COO, has made the things I talked about, but it's a technology and the cost of transformation gets smaller. For example, we are redoing our tech landscape in Belgium in the Belgian bank by ING. We call it the new banking platform, which is core banking plus the whole of the periphery. The speed by which we do this has gone up because of the engineering benefit we see. So overall, a bit of a long-winded answer. It's mainly COL but also a bit beyond. We think this will continue over the next couple of years at similar levels. there could also be other areas. I agree, but let's not forget, we're growing the bank. We're growing the bank. We're investing in business banking in Spain, in Italy, in Germany, in Australia. I talked about it earlier, Affluent is the same thing. That needs investments, that needs people. So we might see other savings, but we'll see other investments and still income will grow faster.

Yes. I was -- actually, that was going to be my follow-up question because [indiscernible], we like net cost saving numbers, not gross because the growth then gets hidden away with growth and so the cost base continues growing. So -- you mentioned it's already 350. You see it as a good run rate, can that accelerate? Can we see a bit a step-up in cost savings initiatives?

Let me -- allow me to say 1 thing first, because it's the first time I've seen this in my professional life. COO operational costs are coming down in absolute terms in our case. And the FTEs, as I said, we're also going down in absolute terms. I've never seen that before. So something is happening, right? For sure and that's why we are growing the bank. So I think that's an important point to add to this. The scalability is really novel. Now your question, could the 3% be more? Yes, it could always be more. But now we are committing to these numbers based on our growth and based on what we see and RCI should improve with 1 percentage points in '26 and '27.

Gross income. I was actually going to ask about that because that's the other key metric. So we're talking a lot about efficiency ING being particularly scalable, yet the cost income is in the 50th, which is a bit higher versus the second quarter. So if you think about AI and all the initiatives you have talked about, do you think this helps more on the income side or on the cost side?

Well, just by the sheer numbers. On the income side, we're growing faster than before. So our CI is going down, whether then in absolute terms, it helps more on the income side or the cost side that remains to be seen. I think, look, I can add that everything and anything we do in COO and beyond is now being looked at through the lens of AI, right? So when we can actually apply it, we will apply it. That will have an income impact potentially and a cost impact. I think the best thing is it will improve our CI both ways.

Very clear. Do we have other questions from the room? Yes, we have one in the middle.

Just one more since you mentioned mortgage underwriting. What is the regulators involvement and appetite for the use of these tools in credit decisioning, if anything?

Yes. So let me be very clear on all that I've talked about in KYC, but also a genetic mortgages, the regulators are fully involved since day 1 because otherwise, you don't get this done. Now more specifically, on mortgages, the AI act tells us this is a high-risk process. So there's always a human in the loop, right? So when we do this, it's productivity for the people working with, but there's always a person assessing the output. So that's appreciated by the regulators. I think that's really evolved, but that's my personal prediction on the AML side on the KYC side that has already evolved that except for some increased risk, that's a pure nonhuman in the loop automated process.

Perfect. Do we have other questions? Yes.

This isn't really AI-related, but just generally, when you think about the growth on the income side, so as you grow deposits, how do you think about remaining may be disciplined on growth there, where you're not growing deposits just for the sake of growing, but focusing on profitable growth. And if that mindset maybe shifted in the more recent years to be more disciplined over time.

Look, I'm the ops guy. But let me say, we have been disciplined over the last couple of years and will continue to be disciplined going forward. We're not just growing deposits for the sake of having money on the balance sheet. So you can assume that ING will be as disciplined tomorrow as we are today, and we were yesterday.

Perfect. We have another question.

From the outside, it's quite hard to judge, which banks are well invested and well progressed on AI and which are the laggards. What would you advise us from the outside to be asking each bank, either a metric or some sort of ratio which would help us to understand how advanced a bank is -- or what sort of measure would you tell us to convince us that ING is a leader as I think you suggested earlier?

So for the scalability of things, will come to impact in a minute for the scalability of things. I would like to understand when you launch a product with AI, with the support of AI, do you use the same data structure across your bank -- or do you need the Germans to construct something different from the Dutch versus the Italians versus whatever other franchisees. So I take a lot of time to make sure that when we do these things, we've got data products that are basically reusable as a structure across the bank. So our consumer lending, right, in AML is now 65% instant that consumer lending data structure gets used in all these markets. So that will give you at least a clue about how scalable these things are for these people. And if there's a lot of legacy, they can't get it to a single product, then I think that tells you something. So that's on how to scale it. And in terms of how to see the impact of it, I would ask, can you show me in, for example, let's stick with consumer lending. Can you show me in consumer lending over the last couple of years, how much of your origination has become zero touch, right? So have you really been able to go from zero, if it was zero to 80% or 90% or 100% or you're still seeing somewhere between 10 and 20, like many maybe will tell you. And last but not least, and what volumes are you then creating through that increased straight through solution. So those are data products, what's your STP rate or instant delivery and what's the volume going through it.

I actually want to follow up on this question. I think ING is no longer in the mainframe in the Netherlands. Does that help or actually no the solutions even if you are in the mainframe because that's an easy question.

As good point. Very good point, and I'm glad you asked. So it's not only that we're no longer in -- on the mainframe. It's also -- and a lot of people talk about this also today at the investor meetings, obviously, people ask about core banking. So what are you in core banking? Are you migrating to a single core bank or -- what we have done over the last couple of years, we have gone from a core banking structure that is kind of rich, which does everything for anyone, for everybody. to just doing accounts in core banking, right? So we just have your my account in the core bank. And it means that we've got a payment engine for payments. We've got a lending engine for lending. So that has become a set of modules in our infrastructure. The benefit of it is it's easier to reuse that's point one. It's easy to change was you don't touch everything at the same time, right? And the third point is you can be a little more granular in what kind of features you build in those kind of dedicated environment. And last but not least, back to the question we just got on what would you advise in terms of measurement. It also allows you to have data quality at a certain level easier than when you put everything together, I dare to say. So that's what we have done over the last couple of years, and I think that helps our proposition. Right. So rather than replacing the core banking, it's about the middle layer, that becomes more modular. Perfect. Sorry. Are there any other questions from the audience? Yes, there is another one.

The capability of AI tools is obviously expanding or accelerating very rapidly. How quickly can you integrate new tools into your workflows now that we're seeing the plug-ins come out from Claude and Co-work and so on. As that continues, how quickly can you bring these new things.

Yes. So look, things can always change. But if you look at it today, we use Google, right, as a toolkit, so to say, and also for data -- but within that construct, and we also got an on-prem ING cloud. But within that construct, we can use different models. So we've got Google models, we also got Claude models, right? Or we use Claude, we can add Mistral, we don't use Mistral but we can have Mistral. So there's a library that you can fill with large or small dedicated language models. So in terms of the modeling and the diversity of models, I think we've got a lot of flexibility. And the tools goes too far. We've got a single engineering platform that combines a lot of tools from different vendors relatively easily, actually, absolutely easily. So I don't see that as a problem at all or you didn't say it was a problem that you were just asking a question.

Can I follow up with the question, when we speak with Fintechs and especially in new banks, they tell us that 1 of their competitive advantage versus incumbent banks is that they can release more quickly, like they do several hundred releases even a day, whereas banks operating batches, there is a batch a quarter or it's a much more slow environment to go live. Is that correct?

That's correct for certain banks, but not ING. We launched -- we got an enormous backlog every day of launches. So we don't have that kind of monolithic way of you got 3 releases per annum, not at all. We moved to agile quite a long time ago. So that also means that our release schedule is far more intensive than just a couple of times a year.

Okay. How intensive?

Thousands of releases every day.

Okay. Perfect.

And of course, different platforms.

Yes. No, no, I understand your present across many front companies so, but that's very useful. Thank you. Any closing question, otherwise. Not? Thank you very much. This was incredibly interesting.

Thank you very much.