An interview with Emily Prince
"We can really use generative AI to surface the most pertinent information to individuals as and when they need it"
An interview with Emily Prince
"We can really use generative AI to surface the most pertinent information to individuals as and when they need it"
Maurice:
Hello everybody, welcome to the latest edition of C&F Talks. This is a series of interviews with speakers at upcoming events, and today we have with us, Emily Prince who's the Group Head of Analytics at the London Stock Exchange Group. Emily's going to be speaking at the AI in Financial Services Leaders’ Summit, which is taking place as part of City Week at the Guildhall, London on the 21st of May.
Emily, welcome.
EMILY:
Thank you, thank you for having me.
Maurice:
Let's turn to the first question.
Maurice:
How would you assess the potential of generative AI to transform the financial services industry? Is this first phase of AI more about improving efficiency than anything else and what are the main applications of gen AI in this sector?
EMILY:
Thank you. So, I think when I think about the role of generative AI, I tend to think of it in phases. And I think the first phase very much as you're describing is certainly to do with efficiencies and I think because of some of the work that has happened in the past with NLP, I think there's a degree of comfort in terms of the role that generative AI can play from an efficiency standpoint. And I think for many organisations they have programs of work that are kind of well underway to be able to address some of that.
I do think though that there's a very kind of immediate opportunity in terms of assistance though, and we are starting to see that come through in the shape of things like co-pilots with the likes of Microsoft, but actually also in other forms where we're able to surface information in a way that is very helpful to people. And I think thinking about certainly my own experience in financial services, being able to have timely information exactly when you need it with the breadth of information that we have in financial services is a real challenge. And so, we can really use generative AI to make sure that we're surfacing the most pertinent information to individuals as and when they need it which really assists with decision making.
I think we'll see that very immediately and then over time we'll see kind of extensions of that in terms of how it enables more workflows to be autonomous. But ultimately, to be done in a much more scalable and comprehensive way that perhaps avoids some of the fragmentations that exist in financial services.
Maurice:
So, a sense from what you're saying is it's more currently personal productivity but turning to LSEG.
Maurice:
Obviously, it's predominantly a financial data analytics and news provider business, so the potential I guess for harnessing AI must be very significant for it. How is LSEG deploying AI and is it rolling out new products and services based on AI for its customers? And I read today in Financial News about your partnership with Microsoft that you hope to be offering more products through that partnership. So how is all that playing out in relation to AI?
EMILY:
So, LSEG has a very strong commitment to AI, and as an infrastructure provider with services and clearing, is it part of the exchange but also in terms of the content provision with the likes of data analytics and our workflow business, it actually brings together multiple parts of the investment lifecycle for many financial professionals. And so, AI can play a role in very discrete and distinct ways for multiple parts of that value chain. I tend to not see it necessarily as a single step change but rather multiple steps where we're aiding efficiencies for end consumers. I think though, what we tend to think about is how do we balance that though, because there are additional considerations we need to bring in when we're thinking about the use of AI. In terms of do we have the explainability of what is happening in the model and making sure that we're always surfacing information that is being validated and is trusted.
So, for example we tend not to use AI for generating, well, we do not use AI for generating information that is fabricated. We are only surfacing the information that LSEG has at breadth from a trusted perspective, and that is a real differentiator for LSEG and something we're paying a lot of attention to.
Maurice:
Okay, and in terms of the Microsoft relationship and that partnership, is that being built around the AI capability that Microsoft can bring?
EMILY:
In part, absolutely. So, Microsoft are absolute leaders from an AI perspective in terms of thinking about the role of AI and things like productivity from an end user perspective. But also thinking about broader more scalable frameworks that enable a lot of AI applications with the likes of Fabric.
So we work very closely with Microsoft as part of the partnership, and I think one of the great things is we think not just today in respect of what are the problems that we can assist with today in terms of things like productivity but also how do we future-proof to make sure that we get the best out of technologies such as AI that assist and enable financial services. So, it isn't a single thing, it is multiple steps and actually a lot of the products that we're building together with Microsoft, go towards solving very targeted problems for end users which are often points of deep frustration or dislocation in the market today and bring about much more cohesion.
Maurice:
Okay. I've read various studies about whether or not one should combine human intelligence with AI, the so-called hybrid intelligence. Do you think that this makes the best of both worlds or is the human element in many of these processes redundant?
EMILY:
Not at all. So, I think as many people know who've been building out AI capabilities, you need humans to be able to have the right level of domain expertise as part of that, certainly from the outset, but actually from an ongoing basis as well.
So, I see it very much as an assistant as opposed to replacement. Certainly in everything that we've seen we haven't seen it displace, we've seen it assist, and a lot of the tasks that humans are not necessarily enjoying often it comes to the correlation, the things that people don't enjoy doing are actually very well suited tasks for AI to assist in. And actually what that's doing is it's driving much greater enjoyment, productivity and focus of the humans that are using AI to solve for those different tasks.
Maurice:
Yeah so it's very much a partnership between technology and people.
EMILY:
Very much so, yeah.
Maurice:
There's been a lot of concern about the risks that AI presents, as well as some of the ethical dilemmas that may cause. How do you think these should be handled for both a company, for instance, LSEG looking at this, and also a national level in terms of regulation?
EMILY:
Yeah so I think there's multiple pieces to this. I think the first thing is you know when we think about risk from an AI perspective as here at LSEG, we have developed a responsible AI framework with our risk team. And that builds on a lot of where we have very specific areas of interest from a risk standpoint in relation to AI, but also thinking much more holistically around, well there are actually elements of risk that we were already considering as an industry.
So, if I take model risk as a very good example. Model risk is something that we have very well established frameworks around and have done for many decades. They're pretty sophisticated, very refined and there's a lot of learnings that we can gain and glean from that that we can apply in this context. Different use case, different considerations but certainly there's a lot of learnings that we can gain there but we've also introduced a lot of tests.
So, how do we know that the system that has been built is fit for purpose, is a lot to do with use case. Which problem are we trying to solve for and how do we make sure that the application of it is in the right places. And so we safeguard it in many respects in that type of capacity. The other thing if I give a very specific example is, when we're developing things like chat interfaces we always ensure that citations back to the information that it was sourced from.
So, even if we can generate information that is not available, we don't do that. We only come back to information that we can source and we can stand behind as an organisation as validated information by LSEG.
Maurice:
Okay and what about the issue you mentioned earlier of explainability, are there any instances where it's hard to work out precisely why the algorithm has come up with a particular answer? How do you trace that back, if I may ask?
EMILY:
Of course, so there's various things that we do within the models themselves to achieve a level of explainability into the system. But it's not one thing it's multiple pieces of that system that will drive that answer. And actually we take a very, as I was saying a second ago, we take a very use case orientated approach to this, and what I mean is if we are solving for tasks where we think a deterministic model and not a probabilistic one is needed, then that is what we're going to use in that use case. Because we can't achieve the level of model explainability that would give us comfort in performing that task.
There are other tasks where we can gain a lot of efficiency and actually the risk of where we aren't able to explain everything is within a tolerance, so it is very carefully done and then on top of that because it is built as an overall system there's a lot that we can actually do in terms of understanding the modular components and applying a lot of tests in terms of are we able to explain the information that is being the topology of the information that is being presented as a result of that system.
Maurice:
Fascinating.
Maurice:
Finally, looking beyond gen AI what do you think will be the next stage in the development of AI and what are the implications and opportunities that it might have for the future of the financial services industry?
EMILY:
I think what we're going to start to see is that there's historically a lot of distinction that has grown up because different parts of financial services have grown up in microcosms, very much so, and there's a lot of persona distinction within financial services that comes really from a heritage of where it happened to grow up, which country, which type of segment of the service industry.
I think we will see much greater collaboration across financial services because we're going to see much more commonality in use of different information. What I mean by that is, historically the ability to link supply chains within financial services is a notoriously difficult thing to do. But actually when we start to take the benefit of generative AI we're going to see much greater ability to link information end to end through the value chain, through supply chains and so forth, in a way that we can actually gain much deeper insights and those much deeper insights are going to allow for us to be much more accurate, much more transparent about what is happening and what the drivers in financial services really are.
So, it's a real enabler in terms of information as well. Another thing I'll just add to that is I think one of the things that I find very exciting about AI is how it democratises access to information. So in the past, you had to have a certain level of skill or maybe you had to be a coder or you had to be something or you had to speak in English. There was usually some kind of barrier to accessing information in financial services.
What we're seeing with generative AI is a leveller in terms of or the potential of a leveller where all you need is the ability to ask a question and in your language. And that is a very powerful thing when it actually starts to enable the access to the information that people rely on as customers of financial services or the participants in financial services. That's hugely exciting.
Maurice:
And presumably when you go beyond gen AI or the next generation that that ease of access is going to increase further and the degree of insight you get is going to increase. Is that what you're saying?
EMILY:
Absolutely. I mean there's many areas where we would love to understand better the relationships of different types of information.
We have a lot of streaming signals, pieces of information that are happening continuously across the globe. Our ability to absorb all of those and translate those into actionable insights point in time when ideally, we'd like to apply those points of information is something that is a very hard thing to orchestrate. By bringing together each of these different pieces we again go back to that point of we're able to surface the right information to the right participants at the right point in time.
And in financial services where there is just so much information that is an incredibly, incredibly enabling thing for humans.
Maurice:
Absolutely so, it's an amazing development, isn't it?
But our time is up unfortunately so, to our viewers if you'd like to hear more on this very interesting topic, please do join us at the AI and Financial Services Summit which is being held as part of City Week, at the Guildhall, London on the 21st of May where you'll be able to hear more from Emily and from an exceptional cast of speakers. More information available on www.cityweekuk.com.
Emily, thank you so much for joining us today.
EMILY:
Thank you for having me.
The AI AND TECHNOLOGY Financial Leaders’ Summit
Generative AI has already led to many use cases in financial services. However, the real potential may lie in next generation AI. At this summit, top AI experts will share their insights on this, global regulators will discuss emerging AI regulation and financial services leaders will discuss current and future AI applications, model risk management, and governance and ethical issues.