Banking transformation special: The agentic AI panel

Tony Clark: This is the London Fintech Podcast, bringing you practical insights and approaches, stories and inspiration from the innovators who are building

Speaker B: the new world of financial services.

Speaker C: It's our last session, at least on that stage. It is called Panel podcast. Meaning it's going to be recorded and stay forever. Exactly. Um, with that lovely note, the name of the panel is the Identic Bank. What's really working. And it's a really interesting topic and really up to date topic. That's why we have forum. So Tony, over to you. Enjoy it.

Tony Clark: Thank you. And panel Podcast or podcast? Panel. So it's a first for me, um, probably the same for the folks that taking part as well. So we'll see how this goes. But just to introduce Tony Clark. I'm founder and CEO, uh, of Next Wave Consulting. I also host something called the London Fintech Podcast, which, um, I don't know, quick show of hands, anyone heard of the London Fintech podcast? Look at that, about half the room. Okay, so a few more future subscribers out there, but uh, at least we're getting out there. So I'll tell you a little bit about the podcast in a moment. I also want to give you an AI statistic to kick things off as well. But firstly, just to acknowledge, I've got Stuart here from Standard Chartered, I've got Mark Bernard, uh, from Close Brothers, and we've got James from Salesforce. And our challenge is to try and make this a bit podcasty, which is going to be a bit difficult because a really good friend of mine who actually does a bit of proper journalism and anchor work for one of the big channels said make it like a conversation in the pub, which is going to be a bit difficult because we're all lined up in front of you and we have no drinks. Anyway, we're going to see how we go. So the Agentic bank, what's really working? So I thought, well, I'll tell you a podcast story or maybe I'll have an AI statistic and there's a little bit of both. So just if you bear with me, humor me for 60 seconds, the podcast. Um, really excited to be here collaborating with Banking Transformation Summit and LFP as I like to call it. London Fintech Podcast has actually been been running for 10 years and I'm a bit of an accidental podcaster. So we've got 260 episodes, but I can't take the credit for the first decade essentially because I really took over the microphone only about 18 months ago. But at the same time we took the podcast onto Video and we set up a YouTube channel and we put the videos on Spotify and We now have 183,000 subscribers and starting to get noticed which is really interesting and it does lead to things like this. So we issued a uh, show special before the conference a week ago. We've had 25,000 views on that. Some of you I hope may be here because you saw that, I don't know, I'm doing a similar thing with money 2020 in a couple of weeks time. So the pod's gathering a bit of momentum and a conversation around that is an entirely different conversation. So I won't go there. But in terms of AI, the statistic that struck me last week was Anthropic's announcement that their quarter on quarter revenues were up 8,000%. So that's 80x uptake and 80% of that is coming institutional uptake. So we're, and we're seeing it all around the conference. I've walked around, I've seen amazing presentations from 11 Labs, from Salesforce, um, from Lloyds, uh, Bank, from NatWest, from all the folks here. There's 175 firms represented at Banking Transformation Summit and some amazing conversations. The content has struck me as super rich and there's a really strong focus on the practicalities of transformation, the human elements in all of that as well. So real quality event and please be taking part. So that's kind of my intro and now we're trying, we're going to try and do this as a conversation. So hence we're not passing a microphone around but we'll see how we go. So perhaps if someone can jump in, just tell us a bit about what you cover at your firm and how you're using AI. Uh, I'm going to see who's looking at me first. Mark, where are you going?

Mark Barnard Knowles: I wasn't looking. Um, so my um, name is Mark Barnard Knowles. I work for Close Brothers and Close Brothers has been really present in the news due to motor M commissions, complaints and the fact that the organization is having to rapidly evolve. And our CEO went to market and said we are, are rolling AI out at pace. Um, and a consequence or an outcome of that is also we're reducing our colleague footprint by 20%. So we are really living that dynamic of you know, we need to get AI into the organization to deliver real value, real benefit really quickly. Um, I guess that's my.

Tony Clark: Well we have to go Stuart, because you're not the only firm that has announced something about some AI driven changes in the in the workforce.

Stuart McIntyre: Yeah, definitely topical this week. So Stuart McIntyre from Standard Chartered, our Ah, CEO made an announcement on an investor call yesterday that we're going to hopefully cut or not. Hopefully we will result in about a 15% reduction in workforce as a result of deploying AI that can be spun in all sorts of different ways. But um, absolutely, it's a key part of our vision for the next 10 years of our organization. We have a plan through to 2030 of AI being embedded in almost every part of the organization and agentic will be a massive part of that.

Tony Clark: Stu, you cover marketing and some adjacent teams. You've got about 500 folks reporting into you, is that right? So you've got a pretty sizable part of the organization.

Stuart McIntyre: That's right. So I'm AI lead for um. This is a mouthful but corporate affairs, brand and marketing cabm. So it's basically marketing comms function inside and outside the bank. So we have a portfolio of about 100 or so use cases uh, across the genetic genai and more traditional AI. And yeah it's a function of about 450 people plus a uh, million and one agencies that work on our behalf as well. So it's quite a board and that's

Tony Clark: what I really like about the diversity of this panel because Close Brothers, a smaller organization, we've got the creative, the sharp end of perhaps some of what's going on in AI represented with what Stuart's covering. And then to my left, James from Salesforce who are one of the biggest platform players.

James: Yep. So great to be here. So thank you. Name's James. I'm a principal solution engineer at ah, Salesforce and I have a slightly different role to my colleagues on the panel today in that I'm at the coalface talking and selling AI. So I'm aligned to our fintech enterprise accounts. So I'm seeing and hearing kind of the different strategies. A lot of our uh, great fintechs across the UK and across Europe are uh, adopting in terms of AIs. So.

Tony Clark: So I was commenting on walking around the halls last day and a half and really syncing up with what everyone's doing in the sector and part of that I saw the Salesforce presentation and you know I've not been paying that much attention. Um, the Next Wave team work with a number of technologies. We're not deep in Salesforce and so I was a little bit off the pace and then I saw what you've been and 100 agents, full agent canvas and a headless version of the uh, of the platform that Allegedly you've been working on for two and a half years. So a remarkable foresight there.

James: Yeah, definitely. So I think we've kind of, we've kind of changed strategy in a way. In the past. Salesforce really wanted to own that customer experience layer. But I think with the rise of these frontier models, a lot of our customers now are spending a lot of time in ChatGPT, include in Gemini. And I think the understanding is now a lot of our customers, they want to surface and query Salesforce data, but in the flow of work where they actually are, and it just so happens now that might be a cli, a command line interface, it might be a Gemini, it might be an OpenAI. And we're seeing now customers looking to use Headless360 uh, to query and abstract Salesforce data where they're working and not necessarily within the UI or UX of Salesforce. So yeah, big change but really exciting to kind of see that development and

Mark Barnard Knowles: close brothers is there with that. So whilst we are moving at pace and looking to catch up, that headless um, infrastructure, that headless ecosystem is really at the forefront of how we're starting to try to look to improve the experience, colleague engagement and lowering that barrier of entry of users to technology users, to AI.

Tony Clark: Yeah, and you know, there've been some really important conversations over the last 36 hours on the customer experience, on risk, on governance, on ethics. And what we hope to do in this conversation next few minutes is just sort of talk a bit more about practicality. So I'm not going to go deep on governance, but I am going to just wave my prop. So if anyone is uh, wondering about AI governance, this is the seminal work on the, on the topic Governing the Machine. It's written by, by um, Paul Donger, who's the head of AI ethics and safety at NatWest Group, with um, a couple of co authors. So we won't talk about that too much, but if you haven't read that book and you're into AI governance and control, you should, um. Stu, you're a bit more. While we're learning and keeping up with and understanding how to leverage the platform play into the infrastructure of the enterprise, and I might get this wrong, but you tell us for your function, you're kind of more on the sharp end. And I think tools like Adobe, I think tools like Figma, I think design systems creatives, um, all the cx, all the things that you're doing, is that the world that you play in and where does AI um apply?

Stuart McIntyre: Yeah, very much so it's as the owners of the creative gen AI tools across the bank. So um, that's as you say, things like Adobe Firefly, but various uh, image and video and audio generation tools. I mean we're challenged by the governance piece. To go back to your book, right? As a global bank operating in 60 markets, 90,000 people, we have to have a really stringent AI governance layer. And we have that and we built that over the last five years. But um, it's how do you turn that from being a hindrance or a blockage into being an enabler? A lot of the use cases that we deploy in the more creative field are very low risk, right in terms of uh, they use very little data of the banks, they don't touch customers. And so from our perspective we want to be enabling and be right up there on the forefront of what our uh, competitors are doing and small organizations. And so we're continuously challenging that governance layer. It absolutely needs to be there, but we also want to have a route through it for innovative use cases that allows us to actually be at the forefront and be challenging the whole time with what's possible with AI.

Tony Clark: And Stuart, are you using AI now more in the workflow? So it's three and a half years since ChatGPT, the first year or so. It's what I refer to as sort of widget AI which is you can get it to write a nice bit of marketing copy or a landing page or something. But where you really move the needle for a business outcome is end to end workflow. And in the marketing flow is that now starting to happen?

Stuart McIntyre: We're on the journey there. I think by the end of this year we will absolutely be there. So at the moment we have a lot of niche AI solutions. There are parts of the bank that are starting to do much more with agentic and building those workflows. So things like HR for managing um, seconds of people across different functions or managing moving people between countries, that's now being built into agentic workflows where all those pieces are being called automation autonomously with a layer of human oversight. On the marketing and creative side, we're starting to deploy within our optimizely tool which we manage, uh, campaign management through that will pull off all the individual creative actions using agents and build it into an end to end workflow. So we're just beginning to get those pieces together and as I say by the end of this year I hope that we will be able to basically press the button on the campaign and pretty much Map it out end to end with all of the individual components being generated using the different tools.

Tony Clark: Yeah. And uh, comes to market in a moment. But just on that tooling approach. Anything particular you're using? And I mentioned Figma and the share price halved in the SaaS apocalypse and apparently there's a SaaS recovery and there's a little bit came back on because they've been using more AI tools and the revenue is actually up. That's one that springs to mind if you think creative marketing. But is there anything really interesting in the toolbox that you're seeing that's adding value to your area?

Stuart McIntyre: Yeah, I mean Firefly is a massive one for us. Right. So that's Adobe's tool. What they've done quite creatively is to pull in a vast number of different partner models they call them, which is basically OpenAI, Gemini and so on, all under the covers. So we have one user interface into being able to create images and videos using all those different gen AI tools, all with a layer of um, legal coverage in terms of using those models and then they then building in the agentic layer so that we can then call those from all our different tools. So I think that's the one where we're putting our most focus on um, over the next six months. But we're also looking at things like Powtoon. I don't know if any of you have tried Powtoon, but that you can upload a single document and it will basically generate a video for you. So it's fantastic for training materials, for sharing corporate standards, that kind of thing. Just give it the document and it will generate end to end. Obviously it needs some tweaking but it saves us a massive powtoon. Yeah, Powtoon.

Tony Clark: I've not even heard of that one. Um, and when we were chatting last week just in prep for this Mark, you said, uh, you know, I don't know whether I'm on the right panel because I'm not doing, I'm still not sure. And then you thought about it some more and went actually tell us a bit about what's going on at close runs.

Mark Barnard Knowles: Well just it operates in such a different space and take the same level of automation but the outcomes move much more into that middle office. So we think about things. We're still talking about colleague productivity but in a more advanced stage and just I can create a PowerPoint quickly. We're talking about operational resilience. So using AIOps really to manage um, downtime, manage things like system resilience, or managing major instant processes, the speed of escalation, the accuracy of escalation and stuff like this and then the things like system of work. So we're bringing in a really connected ecosystem system of work that's empowered by and they're called um, AI teammates. We're using Asana. So the kind of repeatable work like the work of a Scrum master or the work of a PMO analyst, we're kind of handling that off so it's keeping that hygiene in place. So we are kind of moving that work through a bit more seamlessly. The human in the loop always exists but a check and balance. But we're trying to see how much of that we can offset so we can start to look at that value. Add that close brothers needs as it pushes forward his agenda.

Tony Clark: So you're managing, you said Asana workloads and Asana, but you're using Asana's agents

Mark Barnard Knowles: now so the deal is very fresh. Um, so we're just working that through now. But Asana have went live with their teammates concept at their last home launch event, um, just almost exactly a year ago. So automating things like Sprint planning, Scrum ceremonies, all of that stuff, really boring back office stuff to some really vital and still critically important to an organization that's delivering increments of software. The more we can offset that, the more we can focus on the intellectual capital to build new products.

Speaker B: Before we jump right into today's episode, a quick thank you to our sponsor, nextwave. Nextwave is an award winning consultancy that is helping many of the world's leading banks, investment managers and insurers deliver on their growth, efficiency, risk and control goals and transform their businesses. With a senior team who have come from MD level positions in major firms like hsbc, Barclays and UBS nextwave. Nextwave has both the deep industry expertise and the hands on capabilities in AI data and automation to drive real results, more business outcomes and less PowerPoint. From the consulting experience, as their clients like to say. Next Wave has moved two week manual processes to five minute agentic automations, rescued global banking regulatory control programs and digitized deal platforms, sustainability and regulatory reporting systems. And they do much of this with hands on engineering capabilities on leading technologies which include ServiceNow, Alteryx, Quantexa, UH and Camunda. So if a modern and specialist alternative to the big brand consulting model sounds appealing, one which covers the full life cycle of strategy specialists and solutions, but with small practitioner teams, rapid delivery and a better price point, then perhaps you should talk to next wave. Visit nxwave.com to find out more and get in touch. That's nxwave.com. all right, let's get back to the episode.

Tony Clark: James. So I think I know the answer. What are you using?

James: So I think one of the great things about what's happened with the Salesforce platform M over the last couple of years, particularly as we've kind of moved into AI and agentic AI is the expectation that we ah, as Salesforce we need to become customer zero. And we very much have done that. So uh, about a year and a half ago I think it was, we turned on our Agent Force SDR capability. So this is agency SDR sales development reps working on behalf of account executive. We turn that on in North America effectively to prospect to a load of leads that we couldn't typically go after. They might have not fitted our icp. We didn't necessarily have enough resource to go after that. And then with our agentic SDR agent we started to prospect and that's now opened up a number of thousands of leads resulting in millions of pipes. So that's just an example of us using our own tech, our own AI capability to unlock hidden value. We've gone all in on Slackbot, which is our Slack AI capability to do account planning, account research. And then we're now kind of using Slack UI as the interface to all of our agents. So things such as healthcare agent, wellbeing agents, uh, workday agents, all being surfaced within Slack as kind of the interface into all things agentic. And we tell that story a lot to our customers because it's really important I think.

Tony Clark: Yeah. And I saw your demo yesterday and I was struck by the sort of maturity of the platform and all the agents that have been built and they go uh, in my head it was Salesforce. Um, CRM. Yeah, hang on, you've got 100 agents and they're doing things like loan underwriting and uh, real downstream flows. So it's a lot more than that, isn't it? And also quite mature because I asked you again just before we sat down, I said, you know, where are you doing this for real? How long has this been in the product? I think uh, this isn't fresh off, you know, fresh off the truck on the demo stand. This is um, being um, embedded with clients for some time now.

James: 100%. I mean we've been on an AI journey for the last 12 years. I think it was when we first started to go into the predictive. We then moved into generative handful of years ago and obviously More recently their Gentec. But in terms of those hundred agents, we're now building financial services specific agents to do all of the heavy lifting. So things such as underwriting agents, KYC agents, fraud detection agents, all of this now is coming out of the box with our financial services cloud product. And that's there to exist to get you going quickly, quickly, uh, get you kind of 80% of the way there. You just fine tune in the model, the kind of the final bits. But yeah, we're really excited, excited to kind of talk about this today.

Tony Clark: And uh, my observation is it's very much an ecosystem and a component play. So I was walking around and I went in the 11 Labs telephone booth. If you haven't done it, go in there, uh, and you can have a chat with an 11Labs agent about how great their product is. And it is quite interesting because there's 3 million agents out there and they've got a remarkable story. And I wandered back over to James and I said look, they sound like they've got um, very human like agents almost. You know, there's some huge uptake and perhaps the most human like agents but the guide rails, the guardrails for regulated flows in this sector surely is still the Salesforce's and the workflow uh, infrastructure. So wouldn't it be great if they had their agents on your platform and what did you say?

James: They already are. So uh, agent Paul's voice can be powered uh, from the text to voice, uh, piece by 11 labs. But the most important thing is when that handover gets to a human where obviously 11 labs might not necessarily be able to close that case. How do we then have a case management tool such as a Salesforce to close out that case as efficiently as possible? Because we're all in financial services. The most important thing is how are we doing fit and proper for the customer as per consumer duty guidelines. And I think that's always top of mind in terms of case management. And that's where hopefully the Salesforce platform continues to ah, play credence.

Stuart McIntyre: And I think that's going to be one of the challenges, right is who effectively gets selected as the um, agentic engine that controls all the other components. Is every vendor like Salesforce, but many others as well are setting themselves up as being both the engine and the agents. And it makes sense at the moment to pick maybe your biggest platform to be that kind of central coordinating point. But I think that's going to be a piece that's going to play out in our organization and many others. Over the next 12 months is actually who owns and runs that piece. The center of the agentic kind of universe.

Tony Clark: Yeah, it's that orchestration challenge which all enterprises have. You know, one of the big banks I worked with years ago famously had 800 different applications running in their global banking markets division. I don't know whether that's still the number, but that is the problem. It's the rewiring, it's the multiple databases, it's the multiple workflows. I've got a question about pace. I gave the anthropic stat on the extraordinary uptake and I learned a new word last week actually, which was claudification. Sounds like a record by the Chili Peppers. But actually, um, the claudification of banking, the qualification of the sector, Someone used that for real in a conversation I had last week. And uh, I thought, well, that kind of describes what's going on now with some, I've heard use cases of big firms who have been on the cusp of, um, vendor purchases, who then elected, oh, I'm not going to do that. I'm going to build it myself with AI coding. And so AI coding has become, you know, alongside Chatbot plus plus knowledge worker. And now a bit of agentic AI coding is that these are the two big use cases for the sector that have got traction. And just in the last three to six months, and particularly in the last three months, things have changed dramatically with the latest models. And the models are doubling in power every four months. As I gather, every enterprise will probably tell you the change transformation adoption cycle is something like 12 to 18 months for a major program. So that gap, gap is getting big. So how do you go faster? How do you go fast enough that. Because I can sit here and talk about Claude, but next door they were talking about Codex 4.5 and OpenAI, which was also equally remarkable. And the agentic thing is almost the same as the coworker on anthropic. I mean, how do you, how do you marry the pace of evolution of the technology, staying ahead with your organization with, uh, bringing the enterprise with you? Anyone want to have a go at that one?

James: So as I said, I've been kind of selling into banks and fintechs for the last 10 to 12 years. And the build versus buy conundrum has always been there and I think it continues to be there. And I'm going to have to face that kind of question. You can potentially build aspects of a CRM, um, but the kind of the four, the two or three kind of key, key Things to bear in mind is the maintenance cost. Are you going to be able to keep up with the latest security standards? And that's kind of where software as a service existed, was to effectively take that pain away. And I think that continues. And whilst it's you know, getting a little bit easier to obviously build with these super power powerful frontier models, I guess the question remains, do you kind of want to actually build on top of something like a sales tool? And that's why we've gone headless now we've got MCP A2A. So if you do want to build, doesn't it make sense to build on something that's already been proven for the last 20 odd years but context has

Mark Barnard Knowles: changed a little bit because buyer used to be uh, a fully purpose solution build used to be Java coding and now the alternative is do I bring something that's low code but then I can manage and configure that? So you're not, you don't take a course across that level of risk you used to have from tech debt and the total cost of ownership because there's still some level we're purchasing something but we're in control of what we build and what pace we build it at.

Speaker C: Yeah.

Tony Clark: Stu, how are you dealing with this? Because the creative sphere is probably the sharp end of all this and you must be uh, throwing new creative tools and possibilities every week. I mean, how are you sort of picking the best ones and moving fast enough to leverage it?

Stuart McIntyre: Honestly, we're not moving fast enough. Right. We're trying to make the right bets, we're trying to justify the resource you put into it, we're trying to find budget for the best tools. Um, but we're not moving fast enough and I would love to accelerate that. Um, I think what it. Because our governance is such a heavy process currently it takes us probably 6 months to 12 months to get a tool onboarded and through governance and into implementation. So that necessarily makes us slower than most other organizations. So I think that then puts us more focused on the tools that we've already onboarded when they have additional functionality. Right. So maybe we're not going for best of breed, we're going for, you know, choosing our stack and then trying to expand it and really sweat the assets as best we can.

Tony Clark: Yeah. And how do you see AI as a productivity accelerator and a coworker versus the point top of the conversation about the headlines around uh, replacing human workers? You know, I think it's fascinating.

Stuart McIntyre: It reminds me so much of where we were in the kind of noughties. So years, right when I, I specialized previously in employee experience tools and so I worked a lot with Lotus Notes and I don't know if any of you are old enough to remember Lotus Notes for all the sins of that product in terms of ui, what it did was put tools in the hands of knowledge workers to generate their own databases, their own applications and really create something impressive from a productivity perspective. I think Claude and others are uh, doing that today, right? Is it suddenly every member of an organization could come up some really clever and innovative of uses of AI. The challenge is how we avoid that kind of sprawl piece that we had with those kind of knowledge applications in the early noughties, um, and how we kind of keep the enterprise control over that. Um, but personally I don't want to control it too much. I want the innovation that every one of us has in our hands to kind of make the best of those tools and go into areas we haven't done before and then be empowered by them. But as a member of an organization leading AI, we have to keep that governance layer in place and control it. And I think that balance is going to be incredibly hard to strike over the next couple of years.

Tony Clark: It is. And I heard just this morning, first I heard the phrase which I really liked actually I think it's gentleman from Krida said, uh, he stopped using a human in the loop. And he said human in control. And I thought, yeah, that's really what this is all about. And that's going to be a continual balancing act, I think thing for all firms. Let's talk about timeline for a minute as well, because I said that, you know, the models are accelerating in power clearly at such a pace and the tech evolution is so fast. I went to an analyst briefing, I think it was last week, maybe a week before, and they were talking about identification. Multi sector actually wasn't just fs, but of uh, BPO industries and multisector. And they had a timeline on their graph that went out to 2040. And I put my hand up, I said 15 years. Really?

Speaker C: Really.

Tony Clark: You know, um, we're all hearing the world's going to be over by next year and we'll all be working for, working for a Tesla bot. So how do you marry that? I don't think the answer is 15 years. It's somewhere in between. But uh, Mark, where do you think we are on the timeline?

Mark Barnard Knowles: I think the predictability of that has got to be next to naught.

Speaker C: Right.

Mark Barnard Knowles: Like AI is inventing itself. Every five minutes. And I only realized there was an apocalypse last week. And you've told me we recovered from this, so I'm a bit behind the curve.

Tony Clark: Not sure where we are today, by the way.

Mark Barnard Knowles: Yeah. And you know, to the points, you're right. Like I'd like to be a bank that says they're going at the right pace. I just don't think it exists. I don't think we can be fast enough. So tomorrow's looking hopeful. I think that's what we can say. Right. Another 15 years. Let me know how you get on.

Tony Clark: Yeah. James, any thoughts on that one? Yeah, I mean it's a leading the charge on the platform play.

James: Yeah. And even us, it's so hard to know kind of where what it's going to be like in the next couple of years. But I still think we at Salesforce, we have very much the view that humans and AI and agents work together to achieve better outcomes. Ultimately, we've made a commitment to hire a thousand, uh, interns across, across, across the globe as part of our futureforce program. And mainly that's because these interns, they're AI native. Right. They're not digital native. They're AI native. They've only known AI, which is making me feel quite old all of a sudden. But that's the idea. Now we're going to have a new wave of workforce coming through and it's an untapped potential to start to kind of use those types of, uh, young, young talents to help the business drive.

Tony Clark: You almost said next wave there, which I thought was a sort of sideways reference.

Stuart McIntyre: But Stu, a couple of things for me. One is that, um, it's a classic cliche of everybody always overestimates what you can accomplish in a year, but underestimates what we can accomplish in 10 years. Right. And so I think probably we won't move as fast as we think over the next six to 12 months. But actually when we look back in 10 years time, we'll have changed the enterprise in a way we just can't imagine today. Um, and then the other piece is I think there's a lot of talk in the investment kind of media about how much large organizations are spending on, um, AI, uh, hardware. The capex spend is just enormous. That's then leading on to them needing to generate additional income to be able to fund themselves. And that's then asking them to ask us for more funding for our use of those tools. And so it's how do banks like ourselves justify that increase in tech budget to pay for these AI innovations when we don't really know what the value is going to be. And so I think it's going to have to be incremental just to cover that.

Tony Clark: I think there's still such a Runway but uh, the sector is somewhat polarized.

Speaker B: Right.

Tony Clark: You listen to what folks have been saying. I think Lloyds bank said there are 57 live use cases and NatWest said 25. Um, and there's examples all around the hall. But on the other hand I talk to organizations who sometimes equip the most digital thing in compliance as a PDF and uh, you still hear stories like that. And there's such a Runway for rewiring and automation. Not necessarily B2C but in the front to back operation which has been untouched for decades. So yeah, for my money it's not 15 years but uh, it's not 12 months either. So um, it does sort of creep up on you three and a half years since ChatGPT and we're talking about the examples uh, that are evident here at the event. Um but on the other hand it still feels quite slow in a certain area areas. Any predictions for 2027? Anyone want to have a stab at

James: uh, that I think yeah, the rise of the synthetic worker. The AI agentic kind of org chart is on its way. I know few companies have already started to document AI agents alongside human people so that'll be an interesting kind of way to play a how people.

Tony Clark: You're not allowed to predict the salesforce share price based on consumption based pricing instead of seats. Not quite because that's the thing now as well.

Speaker C: Mark.

Mark Barnard Knowles: Look, I think same thing I just came from talks exactly about that. I think we see the AI teammate actually become a teammate, an ally that's understood defined that we work with with clear operation.

Stuart McIntyre: I think there will be some significant mistake made with AI this year that will cause massive fallout in the media and then there will be incredible analysis on or ah, focus on how organizations like ours uh, is spending our money and investing in AI next year. I don't think it'll be enough to stop it but I think we haven't really seen the fallout yet in quite the way that it's of kind.

Tony Clark: That's really interesting actually I interviewed um, Paul Dong who wrote the book I was referring to and uh, when I was researching that I don't know whether folks have seen it but there's something called the AI incident database. You can look it up and there's a logging of all the major AI incidents there's 1500 incidents on there. And uh, and I think the first one was in 1983 and it was a machine learning model that was flying a jet, jet, jet fighter that went off uh, piste. But more recently it's all deep fakes, it's scams, it's chatgpt advising on terror activities. It's really scary. And uh, um, I think we are going to see more of that and positive crossover into the cyber threat as well which uh, folks have been talking about at the event as well. For my part, I think, I guess we might see some robo consultants next year but uh, hopefully not too many and those that ah, are um, working alongside us. Okay, so we have recorded this. We'll put it out on the London Fintech podcast channel as a banking transformation special. We're also talking to the conference organizers about a post conference event of some shape Shape. So if you see or hear about that, you want to take part, that'd be great. Not quite sure what shape that might be, a roundtable or another podcast session, something like that. But um, I'd just like to thank everyone for their engagement and thank the panel.

James: Sa.