Ep314 George Beattie Head of Innovation & Chris Mullan Head of Data & AI CFC: Choose your adventure

0:02George Beattie: I'm Mark Gagan and you're listening to the Voice of Insurance podcast, produced in association with Advantage Go, now part of Sapiens. Release your underwriters to underwrite with the Sapiens Advantage Go underwriting workbench. Over the past couple of years, we've been talking a lot about the potential applications of AI in insurance and where in the value chain it's going to make the best impact. But what we haven't done is follow the logic further and start to look into the medium-term consequences for the industry. This is why what follows is so interesting. My conversation with George Beattie, Head of Innovation, and Chris Mullen, Head of Data and AI at CFC, starts off as an examination of the innovative MGA's application of agentic AI into its underwriting process.

0:50George Beattie: That in itself is interesting enough. For example, the 70 to 80 times speed gains for delivering bindable quotes are eye-opening. But it's what follows that makes this podcast special. George and Chris have done a lot of thinking about what the application of AI can really mean for the sector. They've done the intellectual homework, and some of their conclusions will surprise you. For example, they don't feel that AI exclusions are going to be workable, and are only likely to lead to an embarrassing future climb down for the industry. But it's their views on how the age of AI gives us a chance to completely redesign the insurance product from the ground up, that are perhaps more insightful and offer a cautionary tale for incumbents.

1:32George Beattie: Those who simply think that AI will make the existing business process quicker and more efficient will almost certainly be right in the short term, but they might be outmanoeuvred in the longer term as whole new business models emerge and leave them in their wake. When smart and articulate, experienced insurance people really grapple with tech, It helps the rest of us understand much better. George and Chris are great communicators who can translate tech jargon into practical insurance ideas. What's even better is that they're both very grounded, and are really good at deflating the tech hubris that puts so many of us insurance folk off.

2:07George Beattie: In short, this one is a must-listen, and alongside the recent fascinating episode with Clyde Bernstein of Aon, ranks as one of the most insightful discussions I've had

2:22George Beattie: Enjoy the podcast. George and Chris. George, welcome back to The Voice of Insurance. And Chris, welcome. This is your first time. George has been on the show before. Yeah, I'm very jealous. He's never had to share a platform with anyone. So I don't know if you're jealous at all. I'm not sure if I'm used to it yet. So, well, George, I think you've been on the show before, but I think at the time I was doing it, I was probably just sort of so young and earnest, I would just go straight hammering into the questions. And these days, the last few years, I've realised that it's probably a good idea to get people to introduce themselves so that people listening who don't know you might get to know you.

2:53George Beattie: So my way of doing that is to run us through your career, tell us how you got into insurance, and then run us up to date where you are now.

3:01Chris Mullan: So it's a depressingly familiar tale that lots of your guests have outlined before, which is one of falling into the sector. I went to university and thought I'd do a really useful commercial course and ended up doing ancient history and specialising in ancient Sumerian cuneiform tablet translation. Worked out towards the end of my career at university, that wouldn't be the most transferable thing. So I did a law conversion. Well, suddenly you could do cryptocurrencies or something. Well, yeah, funnily enough, I think you're right. I think you're right. And actually, it has come up a few times.

3:28Chris Mullan: So, you know, who knew? But I did a law conversion to be a lawyer, did a disastrous work experience thing in the city where I nearly got ran over by the partner that I was spending the week with before meeting him for the first time. So that was good. And worked out that actually insurance would be a great place for me to be because it's sociable, allows you to talk about interesting things and lots of other reasons as well. And I started doing the Willis graduate scheme in early 2011. And I spent about four and a half years there before coming to CFC for the first time.

3:56George Beattie: That was the first time I met you because I think it was called Young Meteor. When I was at the Insurance Insider, we had an award ceremony called the Honours and the sort of Young Broker Award was Young Meteor. But I think Meteor is quite bad because they kind of rise precipitous and then they fade away. Well, they're transient, aren't they? So yeah, I wonder why it should be like a rocket that keeps going upwards. Well, it's a comet that's just exited view.

4:19Chris Mullan: Yeah, exactly. I'll come back at some point. So I've basically double boomeranged in my career. So I've been at CFC twice. This is my second innings. And I've been at Willis twice. And in the middle, I was at Beasley. So that kind of explains the last 16 years. But it's not a bad CV. What do you do at CFC now? My job is head of innovation, and it comprises three main things. The first one is breakout product innovation, which means coming up with products the rest of the market hasn't thought of yet, ideally. The second piece is product development, which is very much the more incremental side of innovation around keeping our existing products up to date and tradable across our different mediums like APIs, etc.

4:53Chris Mullan: And then the third piece is a more generalist hat, which is just trying to lead and push innovation along in all forms at the company. And that's a kind of random bucket in which lots of different things like this fit.

5:04George Beattie: Wow. Chris, tell us about yourself. I laughed at the beginning because I too fell into insurance. I graduated physics. As long as you didn't hurt yourself. I didn't. I didn't. landed on my feet. I graduated physics into the global financial crisis. So the classic physics to finance pipeline was suddenly very dry. And so I ended up taking a job as an internal audit administrative support at Aviva, working for the financial management and actuarial team. And through that, I became an actuary, started down the actuarial path, went through a few more insurers, fell in love with programming and data science, moved to Deloitte as a capital optimization actuary doing like proper hard maths trying to squeeze money out of balance sheets around Somncy 2 and then got really big into the python and the data science got asked to run the data science team did that for a couple of years that was fun then got asked to join the first three people that were trying to pull together an insurance strategy practice at monitor Deloitte had a great time there kind of ended up running product innovation for insurance at Deloitte which sounds weird but how could Deloitte build things to help insurance clients got a bit tired of telling other people how to do it.

6:07George Beattie: It's had a go myself. It's where we first met as CPO at Eigen, selling intelligent extraction products into London market. And now I'm here at CFC as head of data and AI. We're here to talk about AI probably more than anything else. So we might have to go straight into it. Two years ago, I'll be asking people, hey, what are you going to do with AI? If I'd had you on the podcast, then I think you would have already been doing things with AI. I think CFC, you've always been slightly further ahead on the tech side of things than everybody else, which I don't think anybody listening would disagree with.

6:36George Beattie: Nobody listening out there should be offended by that. So with you, I can say, well, you've done a lot of experimenting with AI in the last few years. What's actually worked the best? Well, I think the fact that we're live in production, doing insurance quotes using agents built on AI, built with AI, it's working, I think, better than we expected. That's a real line in the sand about what's possible even today. And it was clear that that was possible even a year and a half ago and so that's live in production now churning out quotes working better than we expected and we can't wait to prove out that we can scale it and better than you expected in what way is it better than you expected both of the hard and the soft side of it so the hard side like the numbers the stats like how it's changing the commercials and all that fear of it making lots of mistakes and doing something terribly off-piste i'm sure if you build these things poorly you can create incidents where it can all go wrong but the way we've designed it the way we've built it gives us a lot of comfort and safety.

7:30George Beattie: And there's many aspects in which it just can inherently perform better might be the wrong way, but more consistently than a human. A really good example of that is business activity. So it's really important, much like when you get your home insurance or car insurance, it matters what your job is as part of making a risk decision. Same thing for our submissions. If there's any ambiguity as to what it is, you're giving an awful lot of subjectivity to that underwriter about what to classify it. If it's three guys in an office and they're accountants. They're accountants. What does Deloitte do? There's not one thing that Deloitte does.

8:03George Beattie: You could pick lots of different things. It does tax, it does advisory, it does regulation, it does implementation, but there isn't a single business activity code that you can say that does. And that opens up subjectivity, but the agent, really rigorous, really understandable, really repeatable way of classifying that. If the insurance company, motor insurance, thinks I'm a journalist, it might think that I'm sort of whizzing around town chasing after stories and parking in bad neighbourhoods and town chasing the crime stories. Whereas in fact, I just have it down as sort of, you know, social use only and then it immediately forgets that I'm a journalist, thank God, because otherwise it would charge me 50 times what it must charge me.

8:39George Beattie: When I ask a lot of people and we get into the nitty gritty, maybe if we don't have a microphone, they often say, well, you know, it's a lot harder to train than I thought it was going to be. If it wasn't for you, is it because you're better at doing this? I think it's being really clear on what the exam question is, being really clear on what computers excel at and what humans excel at and drawing the lines very clearly around that i think a lot of people burnt a lot of money a couple of years ago perhaps getting over excited about ai and everyone believing that the way forward was to spend millions of pounds training a model i think the inherent power and capability of the frontier models is sufficient that you don't need to do anything extra there's not a lot you need to put in in the millions of pounds kind of level if you build the right system around it, put the right guardrails around it, it works out of the box.

9:25George Beattie: You know a dog is good at running off sticks. You can throw sticks, dog will go and pick them up, but it's not going to be good at typing. Nice analogy.

9:33Chris Mullan: It's a weird idea. I'm trying to unpick that in my head. But I think to use that analogy, the other thing is, what does a stick represent and why is it chasing sticks? I do think the sector needs to get back or maybe import a bit of design thinking, which is if all AI does is perpetuate the existing business model. Are we missing the point of what this thing allows us to do, which is to rethink and redesign critical elements of our infrastructure and how we work? So why do products look the way they work? Is that through a sublime process of evolution where it's now at peak performance? Or if you talked about it, it's been in the market for more than a few years, knows, is it actually a layering of inconvenience that has led to it being a certain way?

10:17Chris Mullan: The underwriting process, why does it have to work that way? We're still wedded to application forms. Why is that the case? The fear that I have with the analogy in talking about, look, let's get a dog to go faster and go get the sticks, is if we perpetuate the things that don't really make sense about our market, we are actually subliminally arguing for faster horses without even realising that we are.

10:37George Beattie: Yes, arguing for faster horses when automobiles are available.

10:40Chris Mullan: Exactly. And just in case it's not obvious to the listener, the faster horses analogy was before the combustion engine came along, if you spoke to the horse and cart industry and asked them what they wanted, they'd say faster horses.

10:50George Beattie: I think it was attributed to Henry Ford,

10:51Chris Mullan: wasn't it? Yes, of course. And he made some money out of that. So that was a good call by him.

10:55George Beattie: But sometimes it's not actually, if you're in the horse's business, faster horse is pretty good. If you're in horse racing, that is good.

11:01Chris Mullan: I think that's a good point. But if I wanted to make a bet at time of the combustion engine, I would have rather bet it on the combustion engine than the horses.

11:07George Beattie: Yeah, it did better. I remember reading a really, it was something, I think it was in the book for economics. There's a really good chapter about New York City in 1893, where it was absolutely at crisis point. Just how do we get rid of all this horse manure? There are more horses than people. It's a nightmare. Then suddenly, 10 years later, there was just no horse manure anyway, because everyone's got a car.

11:25Chris Mullan: Also, if you look at the news at the time, it's fascinating the analogies you get today around saying, well, what are all these horse and car people going to do? What jobs are they going to do? People couldn't comprehend a world in which the car was the predominant form of transport. The combustion engine was so ubiquitous. So it's interesting, right? What we find often, or what I often say, is that to be a really good innovator in anything, you have to be a really good student of history. Because there's not much that we see in the world that we haven't seen before. And just to roll this analogy back to the beginning point of where we started out, around the dog and stick analogy and faster horses and everything else, what is super important, a few years ago when Chris, you were saying people were throwing money and being disappointed at the outcome, if you're not prepared to ask the hard questions about what you're throwing AI at and do some design thinking, you're going to reach a terminal velocity at some point that you can't exceed.

12:08George Beattie: If you just say, yeah, I'm trying to teach my dog how to type, you go, yeah, well, good luck with that one. And that'd be a very expensive and very strange thing to do. If people were spending money on that kind of thing, thinking, oh, well, my AI can do everything, it's better to know where it stops. So where are those boundaries, Chris, between what humans are good at and what AI is good at? Computers are really good at reading documents. They're really good at doing maths. They're really good at doing the same process a thousand times the same way. They're really good at being consistent. They're really good at looking at hundreds of data points at the same time.

12:38George Beattie: What they're not very good at is going down the pub. What they're not very good at is picking up the phone. What they're not very good at is hearing what's not said, reading between the lines and appreciating what the broker really wants to get out of this conversation, to understand what the real life market dynamics are, what they're picking up on the street, what this broker relationship means, what the client means to that broker and all of these things at the same time. All of that relationship stuff, a chatbot can't do that. An agent can't do that. and one of my favourite things having fallen into insurance is a very pro-social human-to-human part of the value chain I appreciate it creates some limits but to George's point if we kept that part of it but ripped up some more of the past dependent ways that the kind of administration of all of this stuff worked it would be absolutely the best industry to be in in the world

13:24Chris Mullan: because if we think about our two types of customers Marcus CFC right we've got two we've got the broker and then we've got the policyholder who is somewhere downstream of the broker and we keep those two in mind's eye whenever we do anything. So this design outcomes thing is thinking, well, what do we have to deliver at the other end? And we look at the broker role a lot and we talk to brokers a lot about what they enjoy. What they do not enjoy, and this is commonly held and understood, is reeking information into multiple portals and having to work across multiple APIs and all this kind of stuff.

13:51Chris Mullan: That's not the fun part of the job. And the story here is in recent years, despite the fact we've got all this technology, all this promise, have brokers' jobs got more enjoyable as a consequence of all this technology. It's sort of rhetorical in that I think most brokers would say, well, not really, particularly if you're trying to do SME business where you're bogged down by systems and it feels like just rinse and repeat and you're operating across a patchwork quilt of different fragments in the market to try and get fair market appraisal done to try and get your client what they need. The result of that is cross-sell is not as easy as it should be, but critically, they're doing about 70, 80 % re-keying information and the remainder is spent talking to other people.

14:30Chris Mullan: At a time where the whole market is getting excited and frankly, some people are going in the wrong direction saying, hey, AI is great because I can employ fewer people and I can automate everything and my margin is going to go up loads. I feel like they might be missing the point. I think the point of AI is to increase the humanity in the system, not to reduce it. If you can flip it for brokers and say, hey, you used to spend 70 % of your time on re-keying. Now we're going to flip it and you're going to spend 70 % of your time talking to your customers, which is what everyone joins the sector to do, right?

14:56Chris Mullan: It's to talk to people. I think you'll get to a promised land, which would be super high productivity, but super human led, where people feel empowered to do more exciting things and to command more money. The last point I'd just make there is penetration of global GDP by specialty insurance, which is what we fit into, is about 0.2%. Banishingly small. So I think it's really important at this point in the market that we don't just say, hey, we're happy with the size of the pie. All I'm going to do is increase my slice, my size in there. And we should be saying, well, what are all the markets that we haven't got to that are now possible because of AI, because of these productivity gains?

15:31Chris Mullan: And it's not necessarily about employing fewer people. It's about doing much more with those people. That, I think, is the exciting part.

15:37George Beattie: I just had a flood plus on that. They were having their 10th anniversary and it was really enjoyable. We went to a 10th anniversary breakfast the other day and it was lovely to see a Lloyd Syndicate showing me a big slide and filling in the zip code of some address somewhere in Alabama. And the premium came up, spat out flood premium,$100. And that's a Lloyd syndicate, a leading Lloyd syndicate. And I was just thinking, God, yeah, back to even in the 90s, that would have been reams of paper. That would have been a minimum premium of$5 ,000. And of course, there's no way that person in Alabama is going to buy flood insurance for$5 ,000.

16:10George Beattie: And why would anyone ever buy it? And why would everyone write it with that expense ratio? Completely insane. And now routinely saying, well, there was only a revolution that started 10 years ago, but you can already see where we can go with this and how much more sophisticated we could be. To go back to that horse analogy and the horse versus the car, what are the first sort of new kind of, you know, space hoppers that you've suddenly managed to create with this? Or what do you think are the most likely things that new insurance vehicles that could be created that we found hard to imagine because we're such a bunch of dinosaurs?

16:44George Beattie: Not necessarily by design, And it's just because we've had layers of compliance layered on us. We've had the way that the market, we've had all those limitations of technology. Now you have those brakes absolutely removed. What kind of things might we be able to imagine that we could do? And what new kind of vehicles have you actually been able to create? There's two things.

17:01Chris Mullan: And if Chris agrees, what I want to suggest is Chris takes on the one that's much more in his wheelhouse than I take on the one that's mine. The two things being obviously the authentic processes that we can run to alleviate the pain points that we're talking about. The other thing is a bit less fashionable, but it's to think about what an insurance product looks like and how it's compiled. I want to be controversial.

17:18George Beattie: I want to say there's three parts of the value chain. Distribution, product, and process. So how we sell where we sell, what we sell, and how we make the whole thing work.

17:29Chris Mullan: Why don't we tackle the two that are in your wheelhouse and then we'll get to product at the end. Are you sure? Yeah.

17:33George Beattie: So distribution end of the chain at the moment, like we said, is very human to human, expert to expert. We're seeing an awful lot of demand for the next version of that. much like the phone allowed distance selling, the internet allowed distribution that way. That's how we were founded, clickforcover.com. It looks like this kind of agent-to-agent, computer-to-computer medium will be a new plane of distribution. We're already seeing all the demand for everyone to be in OpenAI and Anthropics chatbots. And we see a lot of human behavior moving to that as the place where they do all kinds of things.

18:07George Beattie: I think SEO is dead. It's all about, I think it's called GEO, where it's you need to be the credible source for the model. We know of holiday companies who've gone from 80, 90 % of their sales being Google search based to 89 % of their sales being based on chat CPT recommending their holiday. So human consumer behavior is being inextricably pulled to that domain as a place to interact. And we can see that the distribution of insurance might end up that way. And we're making sure that we're ready for that from a technology perspective. Yeah, and there's been some interesting announcements recently on the broking side.

18:42George Beattie: In fact, I've got a podcast coming out on that. The one that Aon was talking about, I mean, that will be out by the time this one's out. Do you think there's going to be more of a sort of digital marketplace where effectively, yeah, that agentic broker is talking to your CFC agentic underwriter and they're finding the best deal? And a lot of that's happening, obviously, with the lower premium stuff. It's all happening without you having to do anything.

19:03Chris Mullan: I think we can say 100 % that it will be a shape in the future. I think the thing that we need to be careful about here in CFC is nothing if not pragmatic about all forms of innovation. We don't do headline innovation for the sake of a headline. We do it because we want it to make a difference and add margin, add value to our customers that we were talking about before. So will agent to agent be a component in the insurance market? Yes. I don't think we would sit here and say it's going to be the only thing because Chris, you alluded to it just then. There's so much stuff doesn't fit in it. So much stuff doesn't fit in it.

19:31Chris Mullan: And if you think about what happened to personal lines and you think about the automation there, What you might find is that these types of AI systems lead to some kind of creep of that kind of capability up the commercial and specialty spectrum, if you like. But it's still going to be captured at a low level of complexity. Because beyond a certain size, there is good reason to believe that businesses like humans in the loop and they want to have that advice. And the thing about agent to agent is that might end up being quite limited in terms of the human interaction that forms part of that. So I think there's a place for all of this.

20:04Chris Mullan: And the secret, I think, to success right now is to spread your bets, but not too thinly, and not to make assumptions in absolutes. So to say, oh, the whole market will be agent to agent, that's not necessary.

20:14George Beattie: I don't think it will, because I think there's so much new risk.

20:17Chris Mullan: It can't be. But what frustrates me sometimes is that people love convenience, terms of convenience. They love to go to the simple example or the simple sort of outcome of everything. So I think it's just important to be pragmatic and say, based on the type of customer, based on the type of process and the type of product being sold, agent to agent will have a role, but it will not fit other types of circumstances.

20:38George Beattie: Do you think it's likely to be the more vanilla stuff?

20:41Chris Mullan: Certainly. Certainly.

20:42George Beattie: Where the human interaction is not as valuable. In the same way, price comparison website killed a lot of the high street broker. Yeah. There was a similar role. And I think there's also, that's the very direct to consumer retail version, but also from us serving our brokers. As our brokers become more agentified in their systems, we want our agent to talk to their agent. So skip the API, skip all of the complex systems integration. So it's just up on the screen and they know, they combine that here and now. Yes, so they can transact. Whatever technology our brokers have to make their world better will also be there to make it very easy to transact from their broker agent to our underwriting agent.

21:18George Beattie: So bringing everything back down to earth, in terms of productivity, what sort of productivity gains have you been able to get? I know it seems a bit base but at the same time there's nothing wrong with that it's great if everyone could do twice as much we could all be paid twice as much. It's absolutely incredible so I alluded to at the beginning right so we're seeing some incredible stats already in our pilot we're seeing inbox to quote so that means the point in which we receive the email to the point in which we're returning with a bindable quote under 15 minutes for quite complex risks we're seeing a massive uptake in conversion because we're serving our broker better we're getting an answer to them faster.

21:51George Beattie: We're giving them multiple quotes at the same time, whereas historically that would have taken more effort, more time, more cost for us. So by having a better outcome product for our broker, we're seeing way less time taken by our underwriters on the risks, more time to answer the follow-up questions, and better outcomes for us and better outcomes from our broker. And in terms of better outcomes, since that frictional, obviously time, time is money and maybe other costs, how much of that is in skill, i.e. getting a lower loss ratio? Presumably, if you have more time to ask more questions, you can avoid more bad risks and write more good ones, hopefully.

22:27George Beattie: I mean, we're slightly allergic to asking too many questions because in this day and age, you shouldn't need to. If you already know the answer, there's absolutely no point. Especially if you've got third-party data that you can validate half the stuff with anyway, right? Yeah, so we are seeing the extra time. We're seeing that with our system today, we could do the equivalent in an hour that an underwriter might traditionally have done in a week. And that creates time. And what we're learning is what's the best way to spend that time. So sometimes it is spending more time on the complex risks because the agent has done the easier ones.

22:58George Beattie: So conversely, all of this automation is actually about creating time.

23:01Chris Mullan: From a human to human perspective, it's like take time on the relationship building, because that leads to a lot of high quality business, right?

23:07George Beattie: I suppose you can't cross-sell someone something if you don't really know what else they've got.

23:13Chris Mullan: You don't have time to talk to them. Exactly. And that's the kind of product element that I mentioned. We were talking about three different areas. The product piece is we have an opportunity now to rethink how products are orchestrated. If you think about what the market did in SME, it brought together these packages of different coverages, normally aligned with industry subsector, to say as a point of convenience, where you can now buy your D &O and your cyber and everything else in one package. The issue with that, of course, is that the products ship with real net-sea built in. Meaning, if you take sections one of six of the policy, you still get all six sections.

23:46Chris Mullan: Which is almost without parallel in any other sector, if you think about it. An analogy would be, you do your online shop, and the supermarket delivers every product in the store, except that 99 % of the wrappers have nothing in them. It's like a crazy way to think about fulfillment. And again, I'm not bashing the market, because they were a wonderful way of shipping relevant coverage in a convenient way, But now we have this opportunity to say, well, look, what could LLM orchestration do to how you put a policy together? Could we get to a place where we have tailored policies for every customer without the required admin and lift that would have been required two years ago?

24:22George Beattie: That's really interesting. I had a similar conversation actually on the reinsurance side just the other day. Again, about building a reinsurance wording from scratch without having to sort of start with this massive suitcase full of stuff and then start removing bits. Yeah, because it's between additive manufacturing and 3D printing and chiselling a stone. We can just build it up instead of starting with the marble and chiselling down. That's a lovely analogy. You don't have too much stuff lying around on the floor. You just build what you need when you need it. Next podcast, I'll steal that one.

24:50George Beattie: Were there any drawbacks? Anything bad happened that you weren't expecting to happen? In the agentic stuff? Yeah. At the beginning, you said about our reputation. And it is really true that what made this easy and successful was the culture at CFC, the desire to do the right thing, to not be too held back by how we do things now. It was a genuine, this feels like an opportunity. We feel like we've got the right to play here. Let's have a go and try our best. And be it the technology side, governance, compliance, legal, and particularly the underwriting side. everyone's looked at this as a huge opportunity to learn win be at the front enjoy their job enjoy their career so there's been no real roadblocks we only started in september last year and we

25:34Chris Mullan: went live in april so it didn't take that long and bear in mind the whole point in the sandbox is things do go wrong in the sandbox that's the whole point and then when you ship it into production you would have ironed out the things that could go wrong and you put guide rails around it human in the loop obviously critical for how our processes work and the agent we've deployed in the cyber team very much augments rather than replaces. And that's key, right? The human reaction to this has to be one of positivity, which is, I now have loads more time to do more interesting stuff. And there's so much of that in the industry.

26:05Chris Mullan: There's so much fat represented by low value added tasking that brokers and underisers have to do. And we've all been there when you've been working in the front lines of the insurance sector. We've all been there with a paper-based application or an accord form that hasn't been filled out quite right and you're trying to extrapolate the data and you're trying to plug it in and the system crashes and all this kind of stuff. So it's really exciting because we have so much productivity gains to get out of this technology and all of this is possible today. It just takes a bit of bravery around asking the hard questions.

26:35Chris Mullan: What comes next around the truly space age stuff? What happens in five years time, etc.? We don't have to know that yet because there's so much rich opportunity right in front of our eyes just by doing the fundamentals right today.

26:46George Beattie: You've been applying this in cyber. What are the classes where you've rolled this out the quickest and where it's been the most easy to apply? Yeah, well, we're only live now in production in cyber. It's our biggest line. We know it the most. We've got the most experience on it. We're obviously exploring all of our other classes. We're looking both at the continuation of our SME volume lines as a natural extension of what we've already built. At the same time, we're doing a pincer movement and we're looking at those real specialty lines that we do and understanding how they work. We've already got some absolutely incredible AI powered automation around transaction liability.

27:19George Beattie: We're able to do submissions in like, again, 15, 20 minutes for some incredibly complex risks that come across our books. But we also know we're learning as we go. This isn't a solved problem. There isn't a textbook. So we're trying to manage that. How fast can we go safely without disrupting our operations, without disrupting our brokers, without messing anything up, but learning every week, we learn a little bit more about what we can do and how fast we can go. So I think if we We had this podcast in a year's time. We'd be saying pretty much anything that we could do was probably in scope and working, but we're taking it safely and slowly today.

27:53George Beattie: To George's point, it's still in pilot very deliberately. There's some things you only learn in production. You can't learn stuff on a PowerPoint. And by the way, did I get a specific number out of you on that productivity gain? You said, hey, you can do complexing in 15 minutes, but what does it take before? So what kind of gain is it? It's 70 to 80x. And there's a difference between elapsed time and work time. and I think elapsed time for us shouldn't matter. The amount of work it took us shouldn't matter. What matters is how long did it take the broker to get an answer and be able to move on with their job and make their money.

28:24George Beattie: And so I think previously we did about 70%, 80 % of all of our submissions we'd reply within 24 hours and like 95%, 99 % we'd reply within 48 hours. And we're now able to very, very consistently get an answer in 15 minutes. And what that means is that what would typically do a week's worth of submissions, we can sit down with the agent and an underwriter and get through them in an hour. Yeah, it's a 40-hour week. That's 40 times. So, yes. So, it's massively, we're not just talking about a percentage, we're talking about thousands of percentage.

28:55Chris Mullan: Yeah, it's transformative. And then, look, just not to put a stick in the spokes at all, but to think about other lines of business. We started with cyber because we can stand off from cyber and enrich the required resolution of an entity to work out whether it's a good risk to write or not. When you look at other specialty lines, particularly liability-focused ones, where the underwriting inputs are more difficult and more subjective. That's where you have to do the pretty unglamorous work that underlines these systems in order to rationalise what questions you're asking and why. And it sounds like a really strange thing to say because you'd assume that every question we ask is completely irrelevant.

29:29Chris Mullan: But the answer is actually that's not true. There is a sliding scale of relevance. And we have a lot of legacy built up in application forms and a lot of redundancy in application forms, where you talk to another, as you say... Why did you ever ask that question? Why do you ask that? Why does it matter to us whether a company has 24 employees or just to say less than 50?

29:47George Beattie: There was always a reason for these things, but often it's because someone had a loss that was something to do with that. And then it became a famous loss, and then thought, well, we need to ask about this in our proposal form, because my MD is always asking about it.

29:58Chris Mullan: Or it's even less exciting than that. It's just that the rating model that was built 20 years ago had an input in Excel, which, by the way, I'll survive all of this like a cockroach. Excel will be here in 50 years. But it had a cell that said you had to put in exactly the number of employees, and that's just fed through, and no one's asked the question. So application forms and the information we seek, there is innovation to be done there. And it's not sexy. It's not the kind of thing that you get up on stage at ITC Vegas and say, hey, we're the world leaders in rationalizing application form questions.

30:24Chris Mullan: But it's so fundamental because it's the whole rubbish in, rubbish out analogy in AI. It's like you're asking AI to chew through all of this stuff that you don't even care about. So there is a really good bit of work there to say, look, what risk questions are we asking? How are we framing those questions? The outputs that come back to those questions, do they actually move the needle? Because in my mind, when you're offering insurance to SMEs, there's very little difference in some risk measures. And actually it's more about choosing good risks versus bad ones, because no one really knows what the price point is.

30:53Chris Mullan: That's the question of demand economics. You want to get the right premium that sells, but is the right premium$3 ,000 or$3 ,010? dollars. The answer is the premium that sells is the right answer. So at the level that we operate, there is much more subjectivity, I think, around that. I think we could do a good job rationalizing underwriting questions, and that'll help us to feed it in. So in the future, when Chris's team says we'd like to bring the agentic process to the tech E &O proposition, that is a very different proposition in terms of the complexity of the underwriting data coming in. So we have to do that rationalization job in order to do it properly.

31:27Chris Mullan: Again, to bring this faster horses this thing back in. If you just took your traditional product or your existing product and said, throw AI at it, you're going to end up with a terminal velocity or a speed limit that gets imposed.

31:37George Beattie: You just can't go any faster.

31:38Chris Mullan: Yeah. You've got to do the unglamorous work in order to go as fast as you can possibly go. So it's really important we approach this holistically and we ask the right questions. In your wildest dreams, these are the wildest

31:47George Beattie: dreams that are about AI, where are you aiming for in five or 10 years' time? Do you have an infinite perfection.

31:57Chris Mullan: More realistically, you're going to get something called recursive self-improvement soon, which is a big debate in the industry, which is where you don't have to put any input into a model. It just gets better on its own. That's the breakout that you'll get towards the singularity.

32:08George Beattie: That's where the dog knocks on the door and pulls you some champagne and says, my first novel has been accepted. Yes.

32:14Chris Mullan: Well, except there is no door and you can't look in. The dog's having its own party and doesn't need it anymore. To invite you, yes. So recursive self-improvement, big debate in the wider world about whether we will see that at all, if it's even possible, But Anthropic has some decent observations on them thinking that it's possible between two months and two years. To be honest, I hope it's not two months because we have a lot of fundamental words. We're not ready either. We're not really ready to look at that. And again, not to blow it up into super sky high rhetoric, but that is why the guys that are leading these companies are advocating.

32:43Chris Mullan: And I don't think it's just marketing. It's why they're advocating for regulation and for everyone to agree to put down the tools for a bit to work out what comes next. Because we're getting a bit closer to the point where it just kind of runs away. So if we bring it back to insurance world and we think, so I'm not going to give you the five-year answer because no one knows. If I was to look two years down the line, or let's look to 2030 because that's how CFC thinks about a lot of things. If we are in a position where cross-sell between our lines is the norm, but to the point where you can get immediate gratification as a broker off the back of multiple lines of business, off the back of one underwriting record.

33:16Chris Mullan: So think about it from this perspective. I'm a broker. I have an entity that wants insurance. I ask CFC for a cyber quote. We give them a 135. but we also give them access to another six coverages that they didn't even think that we could underwrite off the back of the data, and we give them the ability to choose their adventure, that feels like a big change in terms of increasing cross-sell and share of wallet, which is what the brokers want. And because we've got all of that,

33:38George Beattie: one of the things I really look for in that cross-sell, up-sell stuff is, hey, broker, you didn't even know that your client had registered IP. We know that about them. Here's an opportunity for you to go and sell some more because we know that they've got IP. They didn't even realise, you didn't even realise, those, but they have it. You can protect them from the risks of that help.

33:58Chris Mullan: Because our history has always been about making the broker look good. Go back to 1999, the whole thing was we're going to sell a new type of insurance no one's really thought of via a medium that no one really understands in the form of the internet to a target market that's never been able to access specialty being US SMEs in the main. We want to keep that going. So our whole job is making brokers look really good. So to Chris's point, if you're giving a broker intel that they would not have come across themselves, you're enabling lend the grow share of wallet. That's an extraordinary thing to achieve.

34:26Chris Mullan: And we can see it. At the end of the tunnel, we can see that. So that's really exciting.

34:30George Beattie: This kind of, yeah, we've been expecting you, Mr. Bond, kind of thing.

34:33Chris Mullan: Did you know that businesses of the type that your customer are by this kind of insurance, these kind of levels, have you thought about these? And by the way, here's a quote. Here's a quote space that you can occupy yourself and choose your adventure, right? If you want to add in this, you can. So just breaking that referral mechanism in which the time taken to do the referral, most cross-sell opportunities get diluted down to nothing. and you may not even win the line you came in on.

34:54George Beattie: When I speak to the most progressive brokers are saying, of course, we're using AI to go prospecting for us, to be doing those similar things to that client that's not yet a client even, to say, hey, here's a quote. I know you never asked for it, but here it is. This is based on all the information we know about you already.

35:10Chris Mullan: One thing does spring to mind when you said that, which is just we've got to be a bit careful about not being too clever. If you're a small business... It can be a bit sinister as well. A bit sinister. I've heard about this thing cyber. I think I need it because I had a problem the other day, and they come back and they say, well, here's 10 lines of business and you wanted to spend 5k, but I want you to spend 500. We've got to be really careful with the power this gives us, because to your point, it could be quite sinister. So you've gone out and basically done an Intel job on us and learned everything about us.

35:33Chris Mullan: There has to be some level of permission within the process. A little bit. So I think this is really a point for brokers. Brokers will have to play that quite carefully in order to put the right amount out in front of a client that's come in on something, because you are changing what they expected for the better, but they might not see it that way straight away. So there's some sensitivity involved here.

35:50George Beattie: yeah you could hold their hand and work out what that initial handshake is and we'll give them what they asked for to start with and they'll come back as well anyway insurance is still a need not

35:59Chris Mullan: a want as much as people in the sector like to pretend that it's a want like a new iphone it's not it's a need so if we're doing our job really well we get in we get out really quickly we facilitate we get them to tomorrow i heard a lovely phrase the other day you know our objective is to make tomorrow another boring day that's kind of the thing the insurance so if you're out there trying to sell insurance as the next iPhone, you've got a problem. So our target market, let's assume they're aggressively disinterested in what we have to say. Because next to funeral care insurance is a very hard product to sell, let's be honest.

36:29Chris Mullan: It involves bad things. People don't like to think about it. SMEs have got a lot to get on with. They're not thinking about insurance every day. The opportunity we have to grab their attention and to solve a problem for them is very narrow. If you're having to educate them on a problem they didn't know existed, it's even more narrow particularly if it's money they didn't want to spend because let's be honest the economic situation is not good SMEs are not lining up with their pockets full of money to spend on insurance so we've got to be really good at how we feel these products but really good at how we educate them as well and how we make it real a bit like your example with the flood

37:00George Beattie: insurance one of the beautiful things that this technology is allowing is such greater efficiency and efficiency breeds capacity if the expense loading has gone from four thousand nine hundred dollars to one dollar suddenly the actual cost of the risk transfer becomes much more palatable so by bringing on more of this technology becoming a much more efficient organization we can offer more cover to more people because it becomes much more economic i do love your line about the insurance because i mean no one ever invited everyone around to have a party and then to show off their new insurance so they you know but maybe if you're not hanging out with factories if you've got a swing pool i try not to apart from you i love reading the terms of conditions look at my exclusions it's good to talk to you george particularly with your innovation hat on because a lot of the innovation that i've been reading about the newest of the new has been ai being attacked by insurance as a new emerging peril and i know you george i know you must have been thinking about this i'm looking forward to asking this question about how are you viewing this potential new standalone class of business which some people saying to me in candid moments It's a bit like a new cyber, isn't it?

38:04George Beattie: It could be. What do you think?

38:06Chris Mullan: So look, I hold my hands up and say that I don't know how many people were talking about this a few years ago, but I was certainly one of them in saying, I think it might be a new standalone class of business. My view now has totally changed as a consequence of CFC doing its homework. And the thing that I came to realise off the back of doing this homework, which took us six months to a year and an awful lot of partnership and everything else with other people to work out how AI risk works, is that AI is going to be impossible to derive from normal business in the future, which means it's like a utility.

38:35Chris Mullan: So insofar as a business uses water and electricity, AI will be much the same. Maybe even seen as a kind of right, right? Like access to water.

38:42George Beattie: Well, a bit like in the way that the internet, well, we go on holiday with teenagers, then no water comes out of the tap. That's really bad. There's no heating or there's no cooling. That's really bad. And also if there's no Wi-Fi, that's a total disaster. Can we go somewhere else, please?

38:56Chris Mullan: But it's even more ubiquitous than that. So in the future, Wi-Fi, so internet will be everywhere, whether you want it to be or not, and that'll be a big debate in the future as well. And AI will be part of that, right? It's a transportation mechanism. So an architecture firm of the future that uses AI still has its existing liability obligations towards its clients. And deriving the use of AI from their previous or their traditional obligations is going to be very, very hard because there is no legal precedent in the courts for how AI works from a causation perspective. And it's inherent to all businesses.

39:26George Beattie: Ultimately, for example, if it's good old-fashioned E &O for a professional indemnity of some kind, if the output is wrong and it should have been right and it was definitely your mistake, I don't care how you made the mistake. You're still liable, right? Well, exactly. There's a risk of being a bit too clever here. So why would I just exclude your mistakes whether you did them as a human, like a dumb human, or a human using dumb AI? It doesn't really matter. The outcome's the same. You still forgot to file my claim in time or whatever it was.

39:51Chris Mullan: So let's play it forward. I basically have the view that the market has no choice but to cover AI in these existing forms of coverage because if we believe that AI is going to be ubiquitous and used by all companies, so split the world broadly into two camps of terms of convenience, AI users being companies of all types that have AI somewhere in their business, and AI specialists who sell AI as their business. So a big difference. But let's say 99 % of the world are AI users. So to your point, proving that a loss has come about through the use of AI in any one of these companies is going to be extremely difficult because the causation isn't there.

40:25George Beattie: At some point, the general liability insurer and the reinsurer of that general liability insurer, if it's a bad enough and big enough and terrible enough loss, they may want to subrogate against whoever's AI they were using. But they won't be able to.

40:37Chris Mullan: They'll just say, I don't know which bit was AI and which bit wasn't. But even if they could, because if you talk to one of these foundation model providers, who, by the way, are the largest companies in the world and don't care for this kind of conversation, they, on a contract level, absolve themselves of all liability for anything that happens. They say, hey, use our model. Anything goes wrong. It's your liability. So it brings back to the previous point, which is you already own it. So you can't prove that it's down to AI, really. Even if you did, the foundational providers say, well, don't present us with the bill because we're not interested.

41:05Chris Mullan: So what you end up with is something much more subliminal, which is actually much like it's impossible to prove that a loss comes off the back of a business's use of electricity. It'll be impossible to prove that a loss comes off the back of the use of AI. So our approach at CFC has been to say, this is coming, and we don't want to have a second order of products that are AI, D &O, AI, cyber. Because to our earlier point, there is limited tolerance for increased expenditure on insurance. And past history has shown the insurance market needs to be very careful about reacting from an exclusionary perspective in the face of change.

41:39George Beattie: And also, if we're going to a total aggregate level, looking at really big picture, if AI is a net good, which it probably is, then that's probably net good for the whole world's probably a net good for insurance, isn't it? There'll be less losses, one would hope, if AI is controlling risk management.

41:57Chris Mullan: I mean, bear in mind on a fundamental level, the way AI works is designed to be the same way the human brain works, right? So Geoffrey Hinton, father of modern AI, neural networks. So AI at its very best will be like a human. I mean, you can't explain, if you asked five different humans the same question, they'd all come to an answer and they might be slightly different answers. You can't explain what happens in their brains to get to the answer. And AI is kind of the same. It's probabilistic. You can't exactly explain how it's worked, but it gets to an answer. So in a weird kind of way, what you end up with with super intelligent AI is sort of what you have with humans.

42:27Chris Mullan: The difference is there is 30 years of legal precedent as to how software bugs happen and humans interactions with them. And there is not any precedent as to how agentic AI goes wrong. And we've had lots of debates about this, right? The market must not look at agentic failure the same way they look at software bugs, because it's not the same from a causational cause perspective. Chris, do you agree with that? Yeah, 100%.

42:50George Beattie: I actually find the personification of AI in the form of an agent quite helpful, because the more you think of it as just a person with all of the foibles and failures of a person, the more you will treat it appropriately. So the more you will, all of your existing controls, governance, oversight for all of your humans, you need to bring the same mindset to these machines because they aren't perfect. They aren't infallible. and much the same way that a bad programmer will make a mistake and that might cause a bug an agent can do the exact same thing and so if you've got the right procedures and controls to make sure that your day one developer can't delete prods you want to make sure you're doing the same thing with your agent but they're fundamentally the same risk it's there's an operator human or computer

43:33Chris Mullan: in an environment so you almost need like your human risk management characteristics which businesses have forever and that's how we underwrite or have underwritten historically they need to have a separate rule book which is here's how we manage our ai assets yeah good you

43:45George Beattie: got to keep them up to date yeah things will drift and get out of date and models stop working and just in the same way that you know the prominent vendor models update every year don't they and of course they do and that's what you pay a subscription for exactly so just to take this

43:59Chris Mullan: product conversation onto the next natural chapter we are putting in place affirmative ai coverage across our product suites and all that means off the back of what i just said before is that our policies will cover for AI as a cause of loss. So D &O is still D &O, but we're covering for AI as a cause of D &O. You're going to make it explicitly not excluded. Yes, exactly. And again, CFC's history has been one of saying that our target market does not buy ambiguity. So the old adage of saying, well, don't worry because it's not excluded, it's not going to work insofar as it ever worked for anybody.

44:30Chris Mullan: I think people want to see what they're covered for in their policy in normal English, not Shakespeare, beer, not some kind of back-end deal at the back of a policy where there's some clause that doesn't quite apply. We don't want that. And again, big part of how we do product at CFC strategically. So if you look at that, that's saying, okay, across these major forms of coverage, which everyone is quite happy buying, we're going to be saying we're covering AI risk as well, which reflects our view of what we think the world is going to look like, which is you can't derive AI from normal business, and essentially you have to cover it as such.

45:02Chris Mullan: And then where there is novelty, we think that that AI specialty, that specialist group that are reskinning compute and selling some kind of AI product, those guys really like any other subsector, they want a convenient package that gives them the coverage they need with language they recognize from their specialist area and probably with a counterparty that loves AI as much as they do.

45:24George Beattie: Do you think there's room for a few specialist products, almost like an AI product guarantee, that that vendor can say, you know, we've got this great model and I can guarantee it will do these things for this amount of time? I do, because the world's not convenient, right? Rather than having to exonerate myself and, you know, exclude, you know, you'll have to hold me harmless if you want to buy this product. It'd be quite nice to say, actually, you can sue me because I've got this much cover.

45:46Chris Mullan: Yeah, I mean, that's just E &O logic. I mean, I don't think we'd like any of our insurers talking exactly that way, but yeah. You know what I mean?

45:51George Beattie: Yeah, but in the way that, you know, if you want to sign on to that, even in construction, you know, you're a subcontractor, you want to sign on to this job, you need to show me your stuff and insurance, and we'll all agree not to see each other.

46:00Chris Mullan: If I use this analogy very lightly, because it's really not the same thing, and there are people in the market who would love to believe it's the same thing, because they can ignore AI then. If you look at what happened with cryptocurrency, cryptocurrency firms wanted to do business, and they couldn't, because they couldn't get E &O. So one of the innovations, one of the first things that I did at Beasley was to lead those products into cryptocurrency and say, look, these are basically normal businesses with some adaptation. And actually, if you look between the gaps. They're fantastic at things like cybersecurity because that's the way their brains are wired.

46:28Chris Mullan: So if you look at this, to your point, AI companies just want to contract. They just want to sell. They just want to get on with it. They don't need to have wildly unusual coverage. What they need on a foundational level is just the same everyone else has. And there might be some things you weave in there that are novel around novel terminology or risks that really don't fit anywhere. But the bigger part of the iceberg is how do you get existing coverage fit for the game, which is to cover AI. And then you have that piece, which holistically answers the question as to what you do for specialists.

46:58George Beattie: As someone explicitly including the consequences of AI use, how do you underwrite it? If you've got, I say I use ChatGPT, I'm risk A and risk B uses some other prominent model. How do you evaluate which one's better? Especially if it's got to the point where these models are rebuilding themselves all the time.

47:16Chris Mullan: Well, they're not yet. Yeah. But at that point, the risk will diminish into almost nothing because AI was so good at recursive improvement that it'll be better than a human. And then you might have a weird thing where people are discounting AI usage over human usage because AI is auditable and operates the same way every time and doesn't fail like humans do. So you do get to a weird future where what we currently believe is true, which is that AI is more risky, is not actually how it's perceived. And I can see that happening. But to your original question, if you look at something like cyber, one of the things we've been talking about is the extent to which AI dependency is the same as cloud dependency.

47:49Chris Mullan: And the reality is that most SMEs do and will consume AI via the cloud, because typically they'll buy it off the shelf. It'll be available via AWS or whatever cloud provider they have. So the question of what AI dependency risk looks like, you start with what cloud dependency looks like. So what you find is actually those policies are already pretty good at catering for cloud dependency with waiting times and all this kind of stuff. So what starts out being a very spicy and crazy kind of question to get your head around is often attached to something else you feel you know quite a lot about. So that's just an example.

48:19George Beattie: It's just another one of these risks that's probably beyond your control. You've got NatCat and you've got Man-MadeCat, and this is just another one of these Man-MadeCats.

48:26Chris Mullan: Technology Cat, yeah. And the way you navigate that is to provide for ventilation around a portfolio that means you put yourself healthily far away from catastrophes, but not so far away that you have no relevance, but just so you can gain experiential data. So that's what we're trying to do. The nature of innovation in the market is that people need to be more comfortable getting experiential data. The whole 0.2 % of GDP thing has come about because everyone's way to perfect. That's never been our job. Underwriting margin exists where there's uncertainty actuarially. So we kind of need to lean into it and be useful.

48:56Chris Mullan: So there is a real concern, I think, that the market may model its way to a suboptimal outcome on AI by wanting to engage with it in its own terms. And the reality is the AI train has left the station and the market needs to catch up pretty quickly.

49:12George Beattie: it's like we didn't exclude electricity and must have been a really shocking time to be a general liability underwriter when suddenly everyone started using phones because that is you know central and it's always has been and it must have been then it was so cutting edge you can leave the office and you can do a deal without being on premise yeah because at least before you know i had my copy you had your copy both signed each other's copy so we can at least say we signed the same copy and then we can misinterpret what that means that's what lawyers are for and that's what courts are for but at least we've got this contract but now if i'm suddenly sort of spouting my contract down the phone saying buy sell and said did you say sell or did you say buy it's 10 % off or 15 % off and you're thinking being an E &O underwriter around that time you must have been very tempted to say i'm excluding these wretched newfangled telephones because they're a nightmare and they probably were but but we didn't did we so an AI is way more sophisticated

50:03Chris Mullan: than that we have a lot more to lose as a sector by being on the wrong side of the AI change if you don't get it right.

50:10George Beattie: Yeah, we're going to lose all our customers, yeah.

50:12Chris Mullan: Well, customers will just say a policy that doesn't cover me for my use of AI when it is like me using electricity is not worth anything at all. It's going to be quite an embarrassing climb down for those that are using broad-scale exclusions. I was there during Y2K, and it became a bit of a joke, really. Exactly. In early policies said things like any use of computers leads to exclusion. That was soon found to be a kind of logic leap that didn't stand the test of time. So, again, the comment I made earlier on about good innovators are good students of history. We've seen this kind of thing before.

50:43Chris Mullan: Yes, it's very different for lots of reasons. But fundamentally, from a theory perspective, the insurance sector has seen the traps that are being investigated now. By the way, I'm not saying that we're really clever and we have everything right. I'm not saying that for one second. But we do have the benefit of having done our homework. You've thought about it more than most people have spoken to. Yeah. New-jerk reactions are the enemy right now. And we don't have the political capital to trade as a sector on this. we have to get it right as a sector. So my hope is that by being pragmatic and doing the homework and saying what we don't know, that we can exhibit some leadership and for our small part in the sector, move the sector towards accepting this change that's coming along.

51:21George Beattie: It's very interesting that to be able to do that as an MGA, everyone said, well, actually, of course, you know, what if your paper just all of a sudden says AI exclusion number 12 and that's it. But you can lead them because of course, they're not putting 100 % of their eggs in your basket each time, are they? There's a relationship I think between the maturity of that product thinking and the maturity of the use of AI. I think someone who says that they can put a policy that says no AI is telling a lot about their business. I think part of the reason that we've got so far on our product thinking is our personal experience at CFC using AI in as many parts of the business as it makes sense.

51:57George Beattie: That teaches us how it's changing our business and allows us to really put ourselves in the shoes of our customers and go, well, this is bread and butter now. This is electricity now. This is going to be everything.

52:07Chris Mullan: What policy can we buy? It is a bit of a superpower because we can go in front of a carrier partner with Chris and I, and I can talk to the products piece, and Chris can talk to the operations user piece. And there's something in that combination that people haven't heard before. That's not always going to be true, by the way, but it is kind of true right now. So we're gaining a lot of trust by being pragmatic and calling out what we don't know, but we're also gaining a lot of trust or trading on CFC's trust by having done the homework and showing that homework to our supporters. Starting with fundamentals, which is the updates we're rolling out to our wordings are not war and peace.

52:38Chris Mullan: They're fundamentally quite simple. When you accept the reality that we do, which is that AI is not divisible from normal business activity, you don't really have a choice. We have to do this to remain relevant.

52:49George Beattie: It's funny because when you talk about cybersecurity, I was doing some recent things on podcasts on cybersecurity. It was fun. Of course, at the moment, voice recognition has been beaten by AI. Voice recognition is out. You can clone my voice as much as you like. I've gone, my voice has got 309 episodes. Anyone can download them. And I hope that they're not going to phone up my bank and pretend to be me and my bank would accept voice only as my security. And anyway, if it's a sensible bank, it has already ditched any kind of voice recognition.

53:19Chris Mullan: It happened to my aunt recently. She had a voice clone situation. The police turned up. It wasn't the police. They emptied her account. Really terrifying stuff. It reminds me of the movie Blade Runner with the Voigt-Kampff test, which is to test whether someone's synthetic or not. So we almost need a sort of Voigt-Kampff test for video calls to say, look, am I really talking to Mark Gagan or is it an avatar? And to your point, that is an immediate future where losses are being incurred because you can replicate boys, you can replicate the appearance of someone.

53:44George Beattie: This is already the proof, the proof that it's indivisible already.

53:47Chris Mullan: Yeah. And if you look at video conferencing software, the question you've got to ask yourself is, can that video conference be intercepted by a bad actor? Almost certainly, yes. Are there safeguards around whether you know you're talking to the real person or not? Not really. And that's how, you know, there was a breach recently in Hong Kong where someone called into this meeting with the FD and CEO and other executives in the company. None of them were real. They all had the right voices and the right visage. And they told this financial controller to transfer$20 million. And they did it because they're sitting there with the CFO and CEO, and they're not going to say no.

54:18Chris Mullan: It was all within the estate of the company. So that's a wildly different place than we've been.

54:23George Beattie: I can see an application for AI would be to have listened to everything we've just spoken about and look at my questions that I've got in front of me and then tell me which bits we've missed. But usually I just ask you, I think we've spoken about everything that was on here because we've done it in quite a roundabout way and a far more interesting and philosophical way than I thought we were going to, which is great. Do you think there's anything we should be talking about that we haven't spoken about? I think there's a point around the technology is actually the easy part all said and done and as firms adopt the technology they will get faster those transaction administrative parts will get faster but i think many of them might hit the actual organizational speed barrier much faster i think the terminal velocity of really adapting and using this ai stuff we're not even approaching yet but it's really fast i think you asked about how big can it all get i think the technology part's easy the organizational part the cultural parts really difficult.

55:15George Beattie: And if you don't get that stuff right, no matter how good the technology is, that company, that firm, that group of individuals will hit a speed limit, which is much, much lower than the potential. And that, I think, will be some of the defining competitive advantage that firms are trying to seek over the next 5, 10 years. If you're talking about 450, 60, 70 times faster and more efficient, and that's crazy.

55:36Chris Mullan: But not necessarily, because just going back to the lack of penetration point. So it would Darwinian if the pie can't grow. You're a business, though, won't you? Yeah, exactly. So if the pie never grows, it's Darwinian. But we're sort of slightly learning the wrong lesson if we think it's going to go that way. So yes, it will be Darwinian, because those that stay competing for the same pie will probably get squeezed out. But those that win will look at the new frontiers and say, well, if agentic AI doesn't care about language, which it doesn't, we can expand into new markets much faster than ever before.

56:02Chris Mullan: And look at Europe right on our doorstep. In so many ways, that's an opportunity for Lloyd's to grow into. And of course, North America will always be there, and it's a big part of our growth story. So I think the point that Chris is making is an amazing one, which is there is some really unglamorous but important work to be done on the operating system of insurance. Looking at why products operate the way they do, how policies are orchestrated, why underwriting processes work the way they do, why companies are structured in the way they are around products, which is not necessarily how the target market sees themselves.

56:33Chris Mullan: A technology company doesn't say I'm a DNO customer. It says I'm a technology company. so if you do the rationalisation work that means you can exceed this speed limit that people are going to face that's where the exciting stuff is otherwise you just have a bunch of cargo ships strapping these ai rockets on to the outside and yeah you're going to get a few more knots and that's going to really please people in the short term but you're going to miss the bigger picture which is increasing penetration at better productivity which means more margin more enjoyment and basically winning a race i've really enjoyed talking to both of you thanks so much

57:04George Beattie: Thanks, Mark. Well, I hope you enjoyed today's episode. If you did, don't forget to subscribe or leave a like or a review or recommendation on whatever podcast platform you used to access this programme. These really help get the word out. Before we go, just a quick reminder that advertising slots are available here and in other places in the Voice of Insurance podcasts. Podcasting is the fastest growing medium and attracts a high quality audience of key decision makers. It's also an intimate medium where you, the listener, are right in the room with me and the interview subjects. Needless to say, that means it's a great way of getting your message out directly to an audience because you know you've got their full attention.

57:44George Beattie: It's also very cost effective. So get in touch with Mark at thevoiceofinsurance.com to find out how you could be speaking directly to the industry. The Voice of Insurance is produced in association with Advantage Go. Release your underwriters to underwrite with Advantage Go's underwriting platform.