Name: Brainstorm AI London 2024: Strategies To Stay Competitive
Uploaded: 2024-04-15
Duration: 17 min 38 s
Description: Edward J. “EJ” Achtner, Head, Office of Applied Artificial Intelligence, London, HSBC Alexandra Mousavizadeh, Co-founder and CEO, Evident Insights Brian Mullins, Chief Executive Officer, Mind Foundry Moderator: Massimo Marioni, Europe News Editor, FORTUNE

Edward J. “EJ” Achtner, Head, Office of Applied Artificial Intelligence, London, HSBC   Alexandra Mousavizadeh, Co-founder and CEO, Evident Insights  Brian Mullins, Chief Executive Officer, Mind Foundry  Moderator: Massimo Marioni, Europe News Editor, FORTUNE

Brainstorm AI London 2024: Strategies To Stay Competitive

Photography by Joe Maher / Fortune Brainstorm AI

Conferences

Edward J. “EJ” Achtner, Office of Applied Artificial Intelligence, London, HSBC, Alexandra Mousavizadeh, Evident Insights,  Brian Mullins, Mind Foundry, Massimo Marioni, FORTUNE

OK. Hello everyone. Thanks for staying with us. Um So financial services and Gener A I Banks have a huge opportunity to significantly reduce the time that it takes to perform banking operations, uh financial analyst tasks and that hopefully then empowers the employees by increasing their productivity. Now, Alexandra, I know your company has done a lot of research in this space. Tell us who's leveraging generative A I. Well, and what can we learn from them? You mean who's leading, who's lagging? I, no, no, I mean, we do um the, the index that we create and I'll get back to what that is, looks at not only generative A I but sort of the whole A I spectrum. So yes, we um we produced, it's now 14 months ago, we released the first um index measuring the biggest banks in North America and Europe on their A I capabilities or their A I maturities. And um it was uh interesting to see who, you know, who leads and why. And uh essentially we, what we do is we do a full 360 of the entirety of the bank's A I footprint and we look at elements like um the A I talent stack, we look at innovation, we look at leadership, we look at operating model and then we also look at the transparency of responsible A I and um put this index together. Um And the leaders were, it was interesting to see that in the, in the top 10 or if you look at the top 20 they were, they were dominated by North American banks. There's some um uh some reasons I think that uh that could be, we can get into that in a minute. Um The top bank was uh JP Morgan, um was followed by capital one and then R BC and they're very different banks. But what they did have in common or what they do have in common is having been out the gate and very forceful on, on, on, you know, being very clear about what that A I vision is. And then all of that fo that followed was um pushing very hard to hire uh A I talent um setting up research labs, uh really thinking about uh their innovation strategy and structure, uh research and patents and partnerships and vendors and so on. And then looking at the operating model all with a view to sort of how can we get the time from ideation to production down. Um So that's what we created, that's what we established in, in January last year. And then we did the second iteration of the index in November and it's on an annual cadence. So be very interesting to see if anything shuffles between now and the next update, which is in October. And you mentioned JP Morgan got ahead almost by being early adopters and, and, and getting out the gate quickly, obviously, you know, we can't, we can learn from that going forward, but you can't be an early doctor anymore. What can financial services learn from what those top banks have done apart from getting to it early? What, what are they doing really well that can be replicated? Yeah. Yeah, it's, it's, it's a question around sort of how can one catch up now or um, I mean, one thing I would say is that the, the banks that are leading are really doubling down. So there is a bit of AAA gap that is growing between sort of the, the leaders and the index and those further behind because there is such an advantage, there is an advantage in being first mover because you've established a reputation to draw in A I talent. But that said there's certainly um any bank that decides to really, to really focus in on this now and sort of looking at what are the leading banks doing and, and taking what is best uh approach for that specific bank? Absolutely. That can be um, a bank can catch up in terms of being very clear about the vision, being very arti you know, articulated both internally and externally to make sure that A I is the absolutely, you know, most important thing for any leader of the line, any line of business and putting in place. What uh what is um important for talent acquisition and retention. So what is it that makes the bank attractive? A I talent has a lot of other places to go than banks. So um one needs to make the bank a really attractive place to work. So that is, you know, can you, can you show your research, is there uh access to be active and sharing on open source communities? Are there really interesting problems to work on? Is it a priority of the bank? Uh and so on. So there are many ways to make that, you know, skills and training ongoing in the bank and so on. So those are sort of one of the, you know, some of the many ways that banks can, can focus on it now and catch up now. We've got EJ here from HS BC. They're very aggressively making moves in the A I space. I think I've heard you say that HS BC has nearly 1000 applications of A I within the bank. Could you talk us through them all, please? No, not sure. Here we go. What are your favorite EJ? What are your favorites? So that is that is accurate. Um We do have uh approximately 1000 applications across HSB CS operations that use artificial intelligence. Um the oldest of which go back nearly a decade with some of our original machine learning models. Um And as you might imagine, for our 62 market operational footprint, um all businesses, all functions are, you know, heavily invested. Um And really making sure that we're driving the leading edge of responsible and ethical A I with respect to generative. Um I think it's important to say a few things. So as you might imagine, we're testing and learning, you know, a range of generative A I, you know, use cases that are likely to scale. Um And I think it's also important to distinguish between proof of concepts, pilots and production. Um And what I see out there is a lot of great energy, a lot of great momentum. But I think what also needs to be stated particularly in financial services. And I know there's a many different sectors represented here today, but when it comes to banking, financial services, perhaps health care, ok. We do have a differentiated standard of care. We do have a differentiated standard of, you know, regulatory compliance and those are good things, ok? We should embrace that, ok. Um And so for us, the focus is on that fine balance, you know, between bridging from proof of concept into production that's going to take time. And as you might imagine, there's some lower risk types of use cases related to knowledge management all the way up to and including other types of use cases that candidly, even if it were in our risk appetite, it's our impression that in some respects the technology, the tooling is not yet mature enough for production grade applications. Um So we're really, I think proceeding cautiously, but at the same time, we are optimistic about this across, you know, all businesses and functions. Now, Brian with, with anything fun and, and potentially profitable. There comes risk. Unfortunately. Um What specific risks come with deploying generative A I in, in finance like algorithmic trading or fraud detection? What can you tell us about risk and how we can manage that? Yeah, I, I think that uh all the risks come with it. I think it's really important to understand what the models can and can't do. Um And, and whether or not that they're fit for purpose. Um I is the decision that is the highest risk decision to make. Um We see generative models uh which in and of itself is a little bit of a misnomer because all A I, all all machine learning models kind of generate an answer. But, but in, in essence, I think what we're talking about is, is foundation models that are trained on large amounts of data and, and are, are predicting tokens and they are statistically encoding knowledge that exists in the training data. What does that mean? Um You'll get some generalisability um but they're not truly Generali A I. Um they will, they will go off track, they will invent answers. Um This isn't a bug, it's a feature, it's how they exist, they invent these, these answers. Hallucination is not an accident. Sometimes we just like the hall hallucinations. But if you understand that, then you can see that there are some very powerful things that you can do with it. It's just not going to solve all your problems. Um And, and I think that once, once you know that and understand that you can choose the applications where they'll do the best or you can combine them with other machine learning methods to create a powerful solution that gets the flexibility from a user interaction standpoint of something like a large language model put together with a more deterministic model for forecasting to actually plug in the numbers that you'll use as, as part of the ultimate solution. Um So, so I think we really need to be thinking about it not as a silver bullet but as another arrow in the solution quiver. Um And, and then then really understanding um that uh the, the other silver bullet that doesn't exist is the idea of a guard rail. Um And, and, and we hear this term thrown around a lot. Uh But, but it's, it's used almost as a, we're gonna Sprinkle some magic guard rail on top and it's going to suddenly make the thing do what it can't do or stop it from doing what we don't want it to do. And, and I don't think we have enough experience, you know, with a I as, as an industry to say that that guard rails are improving things. In fact, uh often is the case when you add complexity to a system, you make it brittle, uh more likely to break. And, and so I think we should be skeptical when we see something that is not performing. Uh that the answer is we could add some guardrails to it and make it perform. Um We need to be cautious and, and, and really ask the questions about suitability uh on uh at the, at the beginning of a project before we decide whether we'll use something like a, a large language model. And do you think it's fair to say that one of the main factors involved with, with mitigating that risk is the people and the people that are operating these systems now with that? And ejr, I've, I've seen you talk about this before, you know, how can, how can financial institutions ensure that their, their workforce is adequately skilled in order to deal with A I? Because I've had a lot of people talk about a lot of very senior people in A I talk about, you know, it's not a I that's going to take your job. It's someone who can use A I who's going to take your job. So how can uh financial institutions as, as I say, ensure that their workforce is, is on top of their game when it comes to A I and that can mitigate risks on behalf of consumers. Sure. So I don't think we talk about this topic enough. I'm glad we're discussing it now. You know, you can come at this from the direction of the highest and best, you know, kind of productivity output. You can come at this from the direction of responsible and ethical use. You can come at this from the perspective of the law and you could also come at this from the perspective of what's the highest and best way for an individual, a teammate to achieve their personal or professional best, all roads lead to the same thing. And that's making sure you have an engaged, educated and formed workforce that's using these products and capabilities in a way that best meet all of those criteria. And the beauty of that is, you know, is that if you do it the right way, you build great products with product market fit, you have an engaged and informed workforce that differentiates you from a talent perspective, you're able to demonstrate to your key regulators that you're doing this and you have delighted customers. So no matter again what your starting point is, it's imperative that whatever sector you're in, you know, that you really have very thoughtful detailed plans on really employee reskilling. Now, I'll go to the audience to take some questions if there are any I just want to get some thoughts from the panel on what emerging gen a trends can business leaders monitor over the next decade, say and, and what steps are necessary to prepare them for these types of developments that we're going to see to start with. Brian, who looks the most worried. Yeah. No, II, I think that uh you know, I would, I would echo exactly what he said about um the, the idea of humans working with A I and, and really supporting them to learn how to use the tools. Um That's where you're going to get the most clarity. And I'm really excited whenever the user interface kind of gets broken down so that people without A I skills can start to use them and use them in their work. Because every time somebody in an organization uses A I, you get an answer quickly or it synthesize a whole lot of information quickly, but it kind of makes the information, you know, from an information theory standpoint, we'd say it increased the entropy. But really, it's just, it's making the picture a little bit blurry and the person looks at it figures out what's in it and then when they make a decision, they make it more sharp, right? Because they've combined a whole bunch of things from, from the rest of their life and the rest of the role that they do. And, and that helps correct for the weaknesses of these generative models and the more you can put those together, I think the better and, and I'm really excited when I see these U or user experience being developed in products that let people actively participate, not, not in a box ticking human in a loop exercise, but where I'm adding my creativity, then they're adding the creativity and more. And I think that's where we're going to see um you know, real valuable generative A I being deployed in the world. And Alexandra your, your top trend to monitor. So I just, yeah, I just wanted to agree with that. I mean, we, we, we monitor A I talent flows um in a, in a lot of detail in the and because the hypothesis when generative A I was, you know, when Chat G BT four was released was that there was going to be a huge um demand for prompt engineers, right? And so, but that actually didn't happen because banks were looking internally to do exactly what you're saying, you know, put, put it in the hands of everyone and, and figure out how we can, how one can use it for what and solve what problems. So there wasn't that there's been, you know, there's been other types of hires that we've seen instead and um looking at, you know, talent coming from academia, more looking at talent actually coming from big tech that have been part of, you know, behind the, the the growth of and the development of the tools so they know that but, but also not trying to build that in house, but trying to create the, the know how and the talent in inside the bank and upskilling and, and actually putting it in the hands of everyone to some have had that uh strategy, put it in the hands of everyone and then sort of see what surfaces from that and, and look at what, what um what problems, what ideas can come from that and ensuring that there is this, I know a lot of people talk about innovation mindset, but that's actually more important than ever because if you've got someone again and sort of inside the line of business or running it, that has a I to the front of mind all the time as a, as a solution, then then you're going to come up with many more ideas for use cases and use cases that are in the pipeline that then can be tested in terms of selection, which is looking at complexity and Roy that's attached to each of the use cases. So on uh on, on talent, right? So that's uh that's actually went in a slightly different direction than we anticipated, but upskilling and, and some that's, you know, what the banks are doing internally and across, you know, not just senior leadership, but across uh the whole talent stack is something we monitor quite closely interesting. Um I'll throw open to the floor. If there's any questions for the group, gentlemen in the middle. Oh, sorry. Hi. Um Ajay from uh sales force. Uh quick question and this is from a financial institution perspective. Banks and financial institutions have gotten really good with managing structured data and data is so critical to have the right A I output at the end of the day. Uh Can you give us a insight of what uh financial institutions or what you individually are trying to do when managing unstructured data? Because so critical to have a great outcome of any large language model? Sure, I'll take that very briefly. So, you know, again, another topic that's not discussed enough if you expect to get high quality output, even with a well trained, you know, workforce, you know, your house of data, that foundation must absolutely be as strong as it can be structured, unstructured where it is under what type of use, geographic, etcetera, full stop, especially in banking, financial services, health care, et cetera. So again, that strong foundation for your, your data is an absolute prerequisite, especially if you need to get demonstrable, repeatable, high quality outputs that you can stand in front of a customer or regulator to ultimately prove that you have product market fit. OK, guys, I think we'll leave it there. Uh EJ, Alexandra Brian, thank you so very much.