Hello everybody and welcome to what is probably the first in our series of autumn C&F Talks. These talks are where I get to interview a speaker at one of our forthcoming events and today it's my great pleasure to be interviewing Caroline Wayman who is Global Head of Financial Services at PA Consulting. Caroline's going to be joining us to speak at the third annual AI Regulation Summit which is being held in London on the 11th of November. Caroline, welcome.

Thank you very much, great to be here.

Great to have you with us. Now let's turn to our first question.

Obviously, you have extensive experience as the Chief Executive of the Financial Ombudsman Service and as somebody who runs a team of 500 people across the UK, Europe and the US, so you're in a very good position to judge AI global regulatory trends. How do you think the approach to AI regulation is evolving in the UK and what are the latest developments?

Yeah I think obviously it will continue to evolve but I think the approach that has been taken thus far which I happen to think is probably the right one at this stage is not to try to create very prescriptive rules and rather to use some of the existing regulatory framework to create the kind of boundary lines and the overall outcomes that the regulator wants to see and that's kind of in keeping with how the FCA in particular seeks to regulate more broadly, is to be much more about focusing on good customer outcomes as opposed to on prescriptive rules and they relatively recently have restated that as their present position which I think is the right thing to do. Things are moving very, very quickly and I think it'd be very difficult for any regulator to try and give kind of prescriptive rules to firms about what they should and shouldn't do but we'll see how that evolves over time, I think.

Maurice

Yeah, I mean I suppose at the heart of that there's really how do you strike the balance between on the one hand fostering innovation and on the other hand protecting customers and making sure outcomes are fair and transparent. How do you think that balance is being struck now and is it about the right balance?

I mean, I think that is the balance. I think that the FCA has emphasised the importance of two bits of its framework in this. One is the SMCR so the senior managers regime and the other is consumer duty which see relatively recent restating and new set of expectations set out for all financial services businesses in retail context.

So that's the kind of overarching framework and I do think that that is the right way to go and that enables everybody to understand that the boundary lines really are about overall delivering good customer outcomes and prevention of harm that's foreseeable and that's the sort of fundamentals of those things.

I think that's the right thing to do and I think there will be some quite difficult questions for regulators as they start to see how this evolves over time and in particular if I think about you know my old world of complaints and the ombudsman the observability of decision making and being able to understand exactly what belies that I think is going to be important. And so, you know in some sense’s regulators going to need to know what the right questions to ask are and which threads to pull up so I think there's quite a lot to work through there.

I could imagine a scenario in which whether it's just financial regulators or regulators or governments more broadly actually do need to create some more clear kind of absolutes and guardrails, but I don't think that's just an FS regulation question that's probably a broader one.

Maurice

Yes, I mean that issue of explainability of decisions say taken by AI is going to be I think quite key and a lot of institutions seem to struggle with the ability to explain how the models work and the outcomes that they produce. But I guess you know that's looking really at gen AI but the next stage in the development is going to be around agentic AI. How do you feel about that? Is this going to be you know a helper or is it going to be a threat?

I think it must be both is my honest answer, and I think none of us really know quite how far and how fast this will go but I think we know enough to know that it is going to be absolutely transformative to the way that we work and the way that we live.

I definitely think and as a you know kind of somebody with a regulatory background of course there must be threats there for in there. There are considerable risks as a result of that but there are also considerable opportunities and one of the things that I quite often espouse is that I think we have a sort of tendency to think the counterfactual particularly when we think about this in a customer service context is that the human is providing something that's perfect and that's the counterfactual to an AI intervention including agentic AI and actually that's not really the case is it because we humans are fallible.

And I think there's a particular challenge in trying to make sure that we don't just codify and build in all of our existing human unconscious bias. The opportunity it seems to me amongst other things is actually trying to trying to bring that out of the system and not have the same degree of unconscious bias that we all have as humans and I think you know some really transformative things that I think we can do particularly for example in supporting vulnerable customers, that's a really tricky topic for every business certainly for every financial services business, and so you're thinking about ways in which actually we can do better and that has to be the aspiration not just as do as well as, it's like how do we do that better but yeah I think I think there have to be kind of guardrails to manage at the same time as well.

Yeah, I mean just a supplementary on that, I mean, obviously with the AI models that they learn as they go and so is there the danger that well whereas you may be able to remove human unconscious bias yet you'll begin to find an AI developed bias as a result of that learning process that autonomous learning process?

I think so I think that must be a very real risk. And I mean part of you know the human condition of unconscious bias is all to do with making sort of bypassing connections because you go yes well, I'll just go straight to this point because I don't need to do the working out and that's part of how our unconscious bias works isn't it? 

The sort of synaptic connections we make without even realising and so sort of by definition it seems to me that that risk must be very real, and I think the back to observability being able to understand the extent to which that is happening is something that you know particularly if you sit sort of at the top of organisational boards and in executive committees understanding how far you will be able to tell how far will you be able to intervene I think is going to be a really tricky kind of area for people to navigate.

Absolutely.

I mean, turning to how your clients are dealing with AI, how are they sort of viewing it at the moment what are the blind spots they have, say, around you know generative AI?

Yeah I think it's been really interesting over the last you know it's kind of exploded hasn't I mean agentic AI is sort of a topic that you know has exploded into action in the last kind of few months I guess from those of us who aren't practitioners of AI I would say.

And I would say that probably even if you went back perhaps a year or so you would see that most of the use cases were people trying it out and kind of back office functions and so it was quite sort of removed from the customer and I think we are seeing that shift and a lot more confidence in deploying various solutions in a kind of front-facing customer operations.

There's quite a lot of focus on the opportunity for being able to take cost out of operations and optimising processes and such as you'd expect that's all very sensible and very natural. I think we do see quite often that people haven't necessarily put in place the proper kind of governance to make sure that this is done in the way that they want it to be.

And the other thing we see is quite practically quite a lot of clients who quite a lot of people have deployed tools and you know given them out to all of their teams but not really actually given people too much by way of training or expectation, and I mean one of the things we've been quite recently one of our clients is just helping them to measure the benefit of those tools and what their sort of P&L impact should be rather than just saying here everybody have this new tool set.

So I think it's been a bit of a kind of explosion of many, many different activities some people running with like multiple multiple use cases I think ability to wrap your arms around that and make sure that there's a coherent strategy and governance would be the thing that perhaps I would say but I definitely think is a common feature in what we see at the moment.

And do you, how many of the clients or organisations you come across, how many of them are really using it to transform their business model? You refer to the fact most of them are using it for personal productivity or corporate productivity plays back office and so on, but how many of them are really trying to transform their businesses with a sort of holistic approach to AI and its potential involvement you know how to all involve in the future?

Yeah, it's a great question. I mean I see I think the ambition levels are actually quite high in quite a few financial institutions and perhaps retail banking in particular, so I think the ambition levels are quite high. I mean I think it becomes an interesting question as to how far what is the core business of a bank in this context, are they you know they an AI company that does banking or are they a bank that has AI?

And that's when you get really transformation what should they be what do we need them to be and so I think there's quite you know bigger than the normal sort of strategic questions to answer there and you know and given all of that do we have what we need at the top of this organisation and you know across our board and because those are you know there's a, I mean I would say probably most people are not there yet in really flipping that but do they get there yeah potentially I think so.

Maurice

Yes, I mean you mentioned earlier the need to have good AI governance which brings us back to the issue of boards and how well prepared they are. And also, I suppose you know it's a generational thing isn't it that the people sitting on the boards may not have as good a grasp of technology as some of the younger executives and their organisations. Do you think that there's a sort of skills and knowledge gap there do are they fully versed in the potential of AI how do you see boards at the moment and yeah, how well we get to grips with the whole thing?

And that's when you get really transformation what should they be what do we need them to be and so I think there's quite you know bigger than the normal sort of strategic questions to answer there and you know and given all of that do we have what we need at the top of this organisation and you know across our board and because those are you know there's a, I mean I would say probably most people are not there yet in really flipping that but do they get there yeah potentially I think so.

I think there's definitely some gaps to fill there I think that a lot of boards, I mean, I chair a board separately as well, I mean a lot of boards increasingly try to diversify across their non-executives in particular to make to make sure that it's not sort of archetypal just people at the back end it's sort of having concluded their executive career, you do want that sort of experience that's incredibly important.

But I do think it's going to be a challenge for boards to be able to as ever it's, kind of, as board members it's not really you know you don't need to sort of supplement you being the expert in AI but you need to know what the right questions are and whereas on many other topics they will have navigated that or something similar the extent to which people are prepared for this and up to date enough on development I think is going to be is going to be a challenge.

I think the other thing is the extent to which boards how will they know how much is being deployed in an organisation like how can they tell, what models are being operated and how are they functioning, so I think there's going to be quite an important element of helping organisations to both probably change some of the kind of makeup of their boards and potentially ex-cos as well. But also being able to upskill people around the table to be able to ask the right questions and pull up the right threads because in the end that's a big part what you're doing as a non-executive is trying to make sure that you hold the organisation to account but you know the right things to be asking and probing and I think that will be a challenge, yeah.

Yeah absolutely.

But final question, so if we were having this interview in I don't know five years’ time and you're looking at the stage of regulation of AI, what do you think would be the situation then what would be the biggest challenge? Do you think all these issues we discussed will have been sorted out people understand what they're doing, the definition of what a bank is has been you know actually answered by events over time and developments where will we be and where will regulation of AI be in financial services in five years’ time?Yeah absolutely.

Yeah, I think I think it's almost impossible question to answer I mean I think the pace at which we're moving is breathtaking and so will the position hold in terms of what we explored earlier that we that you know we have a framework we have our existing guidelines and or will regulate of regulators have needed to step in?

I mean I would anticipate that we will have had almost inevitably in the next few years we'll have our first kind of fairly significant thing go wrong that will make everybody go ooh hang on a second what does all that mean then and so I'd sort of relatively confidently predict that there will be firm or firms that have things that go wrong and that make regulators say well hang on a second what do we need to take from that and is that whether that leads to more prescriptive regulation.

My own view would be that that is still to be resisted because actually to a large part the main way to get good outcomes here is a combination I think of good culture and good governance in the organisations themselves and creating the conditions for that to happen together with the conditions for innovation and being able to help people to continue to evolve, but to do so very much with good outcomes in mind not just sort of for its own ends but.

So, I think um a slightly doom-laden prediction but I think we will just because that's life we'll see something a high-profile failure or a high-profile moment in time that will make select committees and regulators and governments go oh hang on a second, and then things will evolve from there.

Yeah, very interesting very interesting thoughts. Maybe we'll need AI to come up with the regulations for AI models in the future or be far too complex. We've run out of time, Caroline, thank you so much for joining us today.

For our viewers, Caroline will be speaking at the third annual AI Regulation Summit in London on the 11th November. Further information available on our website www.cityandfinancial.com.

Caroline, I look forward to seeing you in November. Thank you.

Thank you.