VIDEO TRANSCRIPT

So that was an introduction with the hardest name that you will have today in front of you on the stage. It's a pleasure to be here today. It's a pleasure and we're delighted to be the headline sponsor for Money Live again for another year. I'll be talking to you about how to rewire your bank, how to use AI at an enterprise scale and make sure that you're ready for the future. It's based on our experience of disrupting ourselves but also our work with clients across the globe in financial services. To do so, we'll look at three parts. A quick count of three of the opportunities and the challenges, and there are many challenges that we heard about, there are many opportunities that people talk about. We'll then take you through the path to the journey to reach that enterprise scale AI, before we share some of the key early next steps that you can take. If you're not done talking about this or listening to this topic, in the coffee break in the morning and in the lunch break, we also hold holding an AI exchange at the public SAPient Barista Lounge where you can have direct interaction with our experts and talk about how you get value beyond the pilot. Because as we all know, pilots don't compound. It's the impact on the business that compounds and it's the platform that compounds. Are we at the peak hype? Absolutely. We are past it, almost. But are we also at the opportunity? Yes. The hype is real and the opportunity is real. What you see in the next few years is that's going to be a rebalancing of AI is going to change everything to actually people having applied it and continuing being a bank. Right now it's used. For a lot of headline grabbing, firing forty percent of employees may sound more of a management redirection than a complete AI implementation. Saying that software development ends in twenty six sounds great but is unlikely to happen. However, the opportunity for enterprise application of AI is absolutely there and it's undeniable. And it's based on a few facts. One thing is that the LLMs and how you use them it's getting better and better. But to get the value and to make sure you use it, you need more than L_M_ you need more than the prompt, you need two more things. First you need to be able to integrate in an end-to-end work-flow of multiple steps and it also needs to drive action and results and decisions. For that we're seeing the development of M_C_P_s to exchange in a right workflow and context rich agents. To make sure when you go through the workflow, you have to write context in the right direction for the agent to take the action. The second thing you need is trust, that the output that it generates in that workflow is reliable, repeatable and fit and complies with the regulation that you have or the standards that you have. For that we have two things in here. One is the governance that's more and more developed right now. And of course the human in the loop is still the standard by which we d often design. And secondly is the RAC architecture and with that the knowledge graph that makes sure that whatever the LLM comes back with, you can redirect it to the requirements but it also can operate in the context that you want it to be. So the reason that we are The support of the idea that it's no longer hype but opportunity for enterprise scale AI is here is because it's not just LLM improvements, it's an integrated workflow that can operate in and you can create a repeatable and trustworthy outcome. However, where is the value done? Is it in disruption? Agent that takes your whole financial activities independently? maybe in five years, maybe in the future. And by the way five years means it could be three, it could be five, it could be ten. It's so far in the future I can't plan for it right now today. Bigger value that we see in the near term is in the productivity play, which has two sets. One is around optimisation, particularly automation of existing tasks. For me to scan a document, a person can retype the information or I can scan it with the help of A_I_ or L_L_M_s. Transformation is where the bigger impact can be done. That's where you change the operator model. It's no longer the system change through AI, but it's the process change. It's a people change. It's maybe a data source change that you need to have. You need to put them all together to get to the benefits. In the transformation, it's also easier to get those benefits out because in optimization, you quite often have a few percent. If I'm five percent more productive this week, what do I really do? I will spend my time somewhere else. As an organization, it's hard to capture that benefit. In transmission, you're looking at bigger value, potential value, and therefore it's easier to capture or to think about it. This is where we are today. Now I put two stats on there for banking. One is How do you create productivity in banking and what's the direction of people thinking about and the other one in particular around engineering within banking, which is technology optimization. And you see the ranges, but you also see that the red square means we haven't really scratched the surface of realizing those benefits. One thing is because new technology coming to market, the change that we saw in 2005 make enterprise scale AI more kind of... feasible, but we still need to build it, we still need to create that systemic change and a system of w in which AI is part of the driver of a catalyst of change. But there are also some observations we have around the challenges that we see our clients in actually adopting AI and b moving beyond the pilot and getting to an outcome. And some of the challenge that you might have seen as well I'd like to talk to them about the four things in speech, in scale, in the alignment in organisation, and actually confidence. And confidence is related to adoption. If I don't trust what it does, I'm not doing it. If I don't really understand how it works, I might not be using it. Speed is an essential part of issues we see within banks. It takes too long for the cloud environment to build. It takes too long for the AI service to be enabled. It takes too long to get an approval. It comes back to alignment. The business wants to do something, IT wants to build something. What does the risk function do to uh assign it and to approve it? And last is scale. Many parts of a bank will do the same thing. They'll enable the same service. They try to get access to the same data. They try to get the same approval within compliance to u to do a certain use case. That's a waste of time and it's a waste of speed. So what did we do to concrete these challenges? And this is how we get through the journey to enterprise scale and I. We put in three things. One is a switch to the main value stream thinking. I'll go into more detail about it, but domain value stream is not just a business domain, can be a functional area, can be capability, but it's there where the rubber hits the road, there where it actually gets used as part of a business process. Setting up a factory.

with shared standards, assets and governance. The factories for speed. So the domain is for value and making sure there's business impact, the factory is for speed and consistency of what you do. And the last C_ is the point around people. And people, it's not just talent, not just hiring a new people, not just developing the talent, but it's also about driving adoption. Too many times we've seen where the initial attention on going live creates a good pickup of the solution. And then it peters off. People get a bit unhappy with it, we don't really see a reason for it. How do you drive that adoption going forward and how do you address the different kind of segments of people that you have on the adoption journey. So a bit more detail around the domains. Domains, as I said before, is function, business, okay ability area. So basic areas, SME onboarding, lending, mortgage journey, the journeys that you know from the past. In case of in function areas, you can think about credit risk, you can think in finance, you can think in technology, or in capability side, legacy modernization, where we see a big uptick and big potential at this point in time by using AI or agentic QE, where you actually replace a lot of the manual effort with agents. But it requires a full... Redesign. Now what to do on domain value streams? Again, this was around pivoting towards a business impact and therefore value. You need cross-functional teams led by business, but equally led by technology, with lines of risk in there as well. It needs to be impact and KPI based. So if you have an idea about the impact, make sure you have incremental KPIs. It's not about the grand picture in five years, it's about what will you achieve this year, what benefit will come out of it, and start targeting people in that direction. With some flexibility of course, because you want people to innovate. So if the targets are really harsh and completely built into the bonus, people will only do the things they know will work. So you need to get the balance in the KPIs.

that they partly allow for innovation. And lastly you have to go beyond AI, new data sources, new process design, different requirements for your products, that's where the value comes in. But it's a longer term journey and it's not done overnight. Let me give an example. This is a journey that you see now quite often. SME banking. I've instructed the cost of multiple clients. You see it takes forty days to cash. So I apply for a loan. I want one, forty days later I have one. That's not good enough. As we saw this morning, it has to be much faster. What you see in this process of a standard six-step process, there are all kind of friction points. Some of those friction points I can use AI for. Others are just driven by the fact that it is a very old system. The data is not good in the underlying systems. The training of the RM is maybe not the optimum. The product is too complex for the market it serves. So you need to deal with the different issues. In here you can look at two lenses. Where can I improve most productivity? So go to where most people are deployed or where do I shrink the time the most? So where do I lose most time? You get to different parts of the journey that you want to optimise. You go to underwrite or application for a loan if you want to reduce or you want to increase the productivity. You go to the collateral management or the documentation part when you want to speed up. What can AI do now today? It can for instance create your collateral portfolio that you need for the underwriter. So the underwriter can then use it through the standard engines they already have in their own review. And the good thing is you take care of some of the grunt work about collecting documentation and information that actually they're happy to pass off. So they get to spend more time on where they have value add, which is the underwriter component. So with journeys like these, where there are multiple root causes and not all root causes can be addressed by AI, that's why you need to have the domain view so that you make a plan across this. On the factory side, so factories is about velocity, is about getting speed. What's a part of a factory? Part of a factory is common risk and governance. You need it because everybody will ask themselves the question, I have an idea, I have to do a test, I have a POC, I built something, how do I go to the next stage and how do I make it easy for myself. I don't understand the form, I don't know why they ask these questions. There's an AI platform which has two components. Scaffolding, standardised away how you create the first experimentation, how you develop and how you deploy the solutions, but also offer repeatable AI services.

everybody needs. Why would I develop it twice? And then lastly, teams, for we deploy teams is the nicest word that we have these days, it will change again, but these are people from the technology side that can actually work in the business domain teams and are therefore forward deployed because they work on the business side. The scaffolded elements you see here, that's growing as ever. But the real reason to have the scaffolding is to have a repeatable and fast pipeline from idea to deployed AI service. On the people side, people is about two things. Making sure you have the people to operate in AI space. It's a lot about development of the existing people. Because AI is a component of what they do. It's a component of the domain space they have. It's a component of the capability they have. And of course hiring new people. If I were to hire new people, I would first hire them in compliance and risk areas because they are a significant delay that I have. But of course also in technology side. So broad up-skilling and specific new roles. The new roles, AI architect, AI engineer, but also here ranging the AI translator, what's the product manager that bridges the domain view and bridges what's required from a technology perspective. And lastly, around adoption, what I find is that you can find many different segments of people, how they behave and how you help adoption. You have executives, you have people operating the systems. But instinct one is to look at consumers versus specialists. There are pockets in the bank where you already have specialists that have lots of experience in in machine learning that deal with lots of data. They will pick up AI and what's required very very quickly because it's close to their capability skill. And quite often those teams like in Credit Suisse they define their work not in terms of productivity but important in terms of the impact they have, the effectiveness of how they operate.

and there are lots of activities they have that they can want to get rid of so they can add more value. But consumers are the people like probably all use within teams, what's the summary of the meeting? I don't really understand how it works, but it looks good. Then there it's more of a productivity play. And the adoption question there is very different. They want to be known that their job is not displaced, that they still have it tomorrow. The first group doesn't really worry about it. They find it instinct, they wanna work in it. Of course, we'll apply it to ourselves as well. Part of the learning is disrupting ourselves as a technology consulting firm. We've identified three domains where we think it can make a real difference. First one is on the authentic AI to update and to transform businesses' functional areas and for that we have Bode. So you can go from an agentic solution, instead of for months, to weeks. It comes with an underlying platform where the uh A_I_ agents run on. And of course the agents will sometimes need to be uh d uh adopted for your environment. But they are there because many people need the same type of agents. It really cut back on the initial deployment and design. And from there you evolve further. We have circuit string shots, which addresses the software development cycle. Within that we see a big, big impact on legacy monetisation. Two weeks ago we completed the test for two months with a client, where we cut back the time of what they had done manually using this solution by at least three x. And changing that speed reduces cost significantly. reduces the risk of the legacy modernization significantly and creates a different demand and opportunity. And then lastly sustain, and this is about reducing the cost of managed services, in part by removing the need for it. So removing the tickets and the disruption and in part by automating the processes. So our three domains where we reinvent ourselves to disrupt ourselves. So first steps to take if you're not organized in domains, if you're not organized with a factory and on the people agenda realize that moving to enterprise scale AI is only 20% about the technology and about 80% about how you organize and your priorities that you set. So pick one or two domains. Start with a domain that people find interesting enough and believable enough and complex enough that if you achieve something there they listen and they pay attention. But it shouldn't be the whole bank. It shouldn't be the whole retail division. Make it smaller. If you think about an onboarding journey and if that's relatively well-organised, you can do that by product, or if it's an onboarding journey for a client, you can work on the K_V_C_ element of it. It's big enough to be meaningful with more focus, so that you put a team around it. Then organise the factory and the people agenda in line with the initial selection that you've done on the domains. Because a AI platform can have so many services and so many requirements that you can build for a decent time. However, if you select the domains they need to be served first. They need to be served first in the AI services that you spin up, in the governance around it, and they need to be served first with the people that can operate it. So everything flows from the selection you make on the domain side. Then as you progress, you evolve the domain selections, you increase the scope of the domains, you increase the scope of the services you offer in the platform and in the factory, and you expand the number of people that you train and how you scale up. Every time you ask yourself the question when you scale up and evolve, does it address my speed, does it address the scaling issues I may have, am I fully aligned, and do I trust that I'm going into the right direction with my solutions and transformation. You will want to use the funding that you and the benefits you generate initially to expand. And this is where the model comes in. It's no longer a hockey stick curve where for a long time you invest, hoping that at the end a certain uptake will pay for the whole business case. You want, if it's now 26 and it's March, you want benefit this year. What it will do is that success creates success. People see the impact, you'll have some more funding next year, and therefore you can go to the next step. Also have the portfolio approach. Don't bet only on disruption or only on some kind of optimisation. It needs to be a mixed approach which you will leave to the domain teams because they have the KPIs to hit. And then lastly what you will see is if you in pockets achieve a significant productivity, let's say two x, you see sometimes demand for those pockets increase. Because when something, diagnostic for instance, becomes very fast and cheap, people want more diagnostics. So, the fabric of the firm will change over time with the productivity that you generate. In short, what we've concluded is the hype is real but the opportunity is also real. To move towards capturing the iPod opportunity, it's only partially about technology, but more important how you set yourself up, what operating model do you set out there. Do you have the domains in place, the factory and the people agenda. If you want to discuss this further, this is the plug I will do again, at 10:35 and 12:45 in the breaks, we do the AI exchange at the Sapient Bernstein Lounge. It's been a pleasure sharing this with you today. Thank you very much.

VIDEO TRANSCRIPT:

SPEAKER A: