Ep304 Andrew Johnston Gallagher Re: AI will come to its own rescue

I'm Mark Gagan and you're listening to the Voice of Insurance podcast, produced an association with Advantage Go, now part of Sapiens, enabling enterprise scale underising through a single pen of glass. Today's podcast is one of those conversations that stretch my ability to keep up with my interview to the limit. Andrew Johnston is the global head of in short tech at Gallaghery and has been at the forefront of understanding and chronicling the in short tech phenomenon since it emerged over 10 years ago. Gallaghery's quarterly reports on the topic have been required reading over the past decade for anyone wanting to understand the intersection between insurance and technology. In that period, digital disruption, blockchain, parametric and embedded insurance are themes that have come to the fore. Now AI is the dominant force, so much so that in short tech and AI are now effectively synonymous. Gallaghery's latest quarterly in short tech report is one of the best pieces of research into AI and insurance that I've read. It goes way beyond AI related in short tech investment opportunities and digs right into the profound risk implications of AI as an emerging casualty peril in its own right. Andrew's a dream guest, incredibly sharp, bright and fun to spend time with. He's that rare kind of person who makes you feel more intelligent for having spent time chatting to them. Listening back, the reports in the conversation were whole new avenues of questioning were opening up that I failed to explore, but that's the mark of a strong interview. It always leaves you wanting more. It's a bit like when you wake up the morning after a big debate or argument, with terrible pangs of regret that you fail to ask the killer question in the heat of the moment. I highly recommend that you go to the podcast notes and download the report as an accompaniment to this conversation. It's the sort of publication that you'll keep coming back to for reference long into the future. The developments in AI moving so fast that I feel certain it won't be long before I have Andrew back on the show to help put everything into context for us. But until then, this is the next best thing. Enjoy the podcast. Andrew, welcome to the Voice From Insurance. Thank you very much for having me, Mark. It's really good to be here with you. First, tell us how you got into insurance in the first place. It's a long story, so I'm a fail academic. It's a university for far too long. You can't fail. I definitely agree. I do. You know, there's no past rate at that point. Well, I mean, I just couldn't sort of tell it into a professional I suppose. And then I ran a couple of startup businesses in agriculture, including an alpaca business that I ran. I had a CQ cumber business. Were we breeding them? No, I was shearing them. So I went to various farms and would say quite fashionable. So it was a growth industry. Yeah, it was very lucrative and I had a great time and I got to see the world. But it's definitely a sort of younger person. Do you source them? Is it Bolivia Peru where they come from? Originally, yes, they're in the Camelot. I have to go and source them from the main. No, I mean, people own them in all over the world. So there are farms in the states and we're shooting Germany and Australia and they're all over the place. An alpaca is a nice of the llamas. Is that the? Yeah. Jarmies are horrible and alpaca is a nice. Is that the thing? Yeah, I think you can make that sort of general rule. What academic summit were you trying to be academic in? So my PhD was in political violence, specifically in genocide in Cambodia. So that was picked up by the UN, went to live in Cambodia for a while and then that job came to an end and I returned to the UK with my tail between my legs a little bit and my dad who was a Lloyd's person said, look, why don't you think about re-insurance? And I said, absolutely not. And then unemployment lasted a little bit longer and I graduated. So I thought re-insurance breaking was a good option. Yeah, but why not political violence then? I think when you look at your career trajectory, you sort of think what's going to ultimately make me happy and it's quite a dark subject. So I think I think I made the right choice in doing so now. No, maybe you're right. So do you've got us right up to date now? Yeah. I mean, it's been a wonderful thing that you've made this investment actually as a business. It's over 10 years ago now. And I remember because I was the editor of an insurance magazine saying, oh, this insure tech stuff is sort of really happening. Now it's got a name and we were starting to do conferences about it. And thinking we're going to make our own kind of investment. We're going to have to designate one of our reporters. We'll have to have this insure tech beat and sort of be around it and make sure that we know what's going on. I think Gallagher's a broker made the most investment. You've made that leading investment in your report is the standard literature. That's very close. Well, it's not that I don't have to be kind because it's just fact. You know, I'm just a journalist. I deal in facts. It's a fact that you've made that investment over time. You could see that it was gray and semi and you've now got a fantastic chronicle of that going right to the beginning. So no one's better qualified than you to tell us about the insure tech funding because you've been really looking at it and studying it. Yeah. Quarter by quarter. Yeah. For many, many quarters now. Yes. The first report went out in 2017 and I think it was very early. Yeah. It was. I'm, you know, who knew where it was going to go. But academia and research was what I knew and publishing was what I knew. So I think for me it was almost a safe space and I knew that if I wanted to build this brand for what was then willisery in the insure tech space, I needed to get something published in print because it adds that layer of credential. And at that time, one or two fairly major investments had been made under the label umbrella of insure tech, primarily the investment made by Pingan into Zhongan in China. And the securities audience and the kind of ILS audience and the investment community were asking us a lot of questions about whether or not these were investment opportunities for the industry. And to your point, all of a sudden, those are sort of canberian explosion of conferences and everybody sort of talking about it and the conversation at that time. So we are talking about 2016 was very much about disruption being usurped. And we had quite a few wary clients, if I'm honest, you know, were these new digital MGA is going to disrupt them. So I felt like it was incumbent upon us to try and give this fact oriented objective view on the space. And the last 10 years from an investment perspective has been absolutely fascinating. We've gone through very, very high peaks, quite low troughs. And we're now starting to hit a period of consistency, which is very, very in line with the Gotten Height Cycle, if you're familiar with that, that shape. I mean, you could almost draw it onto the investment parabola that went into insure tech and continues to. Yes. I mean, so it's pretty healthy state. We've again, to the sort of realistic. We've been through the trough. We've gone over excited. The ones who are going to lose money, have lost money on slightly frivolous things that make any sense now with hindsight. And now we climb up the sort of, what do they call that in the garden? It's the plateau of comfort or something like that. Yeah. So I would agree with that. And I think that the companies that are raising capital are generally coming to the table with very sound business propositions. I think we as an industry understand the opportunities a lot better. And that's reflected in the great matter of those who are investing, insurers and re-insurers are investing it unprecedentedly, high rates at the moment. But they're generally into companies that are supporting back office efficiencies, rules based automation. And they're all quite sensible, pragmatic, mature decisions. And it's not speculative bets on the poster child of the moment. And I think that's an indication of a very healthy evolution of the space. But one thing though, I was reading your last report and we're going to have links to that at the end of the podcast in the notes. Of course. And I highly recommend everyone really. Because there's so much in it. I mean, I used to publish monthly magazine. And if you get over 100 pages, you're doing pretty well. And this is over under pages. Of course, in this digital age, you keep scrolling where before when you had an actual publication, you could see it was 100 pages. Before you opened it, where it's actually now you get to page one. You don't know when it's going to end. And I kept thinking, this is pretty good. It must be about to have the last page. But there was just so much more to read. I kept thinking, goodness, me. I need to read this other bit. Because otherwise I'm going to miss lots of good questions to ask you. But obviously the first question has to be, was it 95.6% of all the money? I mean, logically, is going into AI at the moment. And you know, you talk about poster children. And we could go back through your back issues. And we go, what was it? Parametric and then would have been embedded and there would have been blockchains. Yeah, blockchains. You know, we've all forgotten about blockchains. Is AI just the latest thing? Or is it there's something more substantial behind it? And it's almost as if you want to get funding and you says you have to have dot AI in your domain name. Yeah, I mean, I think you've hit the net on the head. I think it's probably both things. So I think that AI is already proving to be a transformative series of softwares. And that's what it is. It's not one thing because it's so horizontal, isn't it? I mean, it goes across the absolutely. It sits on top of technology. And I think that's something that people don't fully appreciate. And actually, when most people are talking about AI, they're actually talking about automation. But I think in terms of true intellectual AI, we are already seeing having an impact in our industry. And it is undoubtedly the technology that is in Vogue right now. But I think it has a staying power that probably other technologies don't for the reason you just express it. It sits on top of existing technology. But I do also think that for funding purposes, there is very much a sense from the community of founders that to be taken seriously, they have to have AI in the name. And I think from a valuation perspective, it probably does add value, but you can't call yourself tech without having AI. But the 1999 companies were referring to themselves as sort of dot-com companies. We would laugh somebody out of the room right now if they said that they were a dot-com. And I just think that there is an evolution of terminology that goes alongside all of this. But actually, a lot of the companies that are raising money, successfully, have been raising money for the last 10 years. And actually, they just have a new identifier as part of their company identity. And they may have had AI type functionality all along. They just wasn't the kind of social pressure to pay mention to it. So I think it's both. I think there is an identifier. There's a paint that needs to be on the wall. But I also think as a inhabitant, technology set. It is very important that we use machine learning for this or without the other. Right. It's just probably yes. Now it's absolutely explicit. Now it's the first sort of tick on the box for the due diligence. And then it'll go away again, right? Because it will be assumed that it's so synonymous that there's no need to call it out, which is what we see with every technology hype cycle. And what about those investors have been around for that time? You know, these were quite mature tech investors. So I suppose they already knew that they'd be a hype cycle and I suppose they were probably ready for it and they knew that some of those valuations they're investing out there were unlikely to get their money back. But should we say it's some of those really frothy valuations early on when insured it comes to venture capital fund and says, Hey, here's the pitch. Do they sit with their arms cold and say, I'm not you lot again. I've been so burnt by you guys and insurance. You know, I'd prefer biotech or whatever, something else. Yeah. How do they feel? Or is it quite mature at the moment? It's many things that I mean, there are the tourist vcs and pcs that did make very speculative bets that in some cases lost a lot of money that probably wouldn't want to touch this space again. You then have very, very nuanced sophisticated insure tech specific funds that generally speaking have done well and love this new wave of technology. It's my only point insurers and re insurers are now parting with more checks than they ever have. But I think that the underlying thesis hasn't changed in that there is a multi trillion dollar industry that is seen as being. I don't want to use the word ripe, but there is an opportunity to make inroads with technology that could return an equity play. And that's not lost on any investor. It's just about finding the right business that has the right opportunity that is adding long term value. I think he's quick exit. So probably a thing at the past. Yep. So what are the sorts of AI related ideas? Particularly, I mean, now that we're at a stage where every carrier, every broker has already implemented AI. A lot of the low hanging fruit is being plucked at the moment and the kind of really obvious stuff, the sort of structuring unstructured data. Presumably now, if someone's asking for money today for many millions today, what sort of picture they're coming with now? What sort of premise are they? Obviously, you've got to pitch it on the basis that you're going to make 100 times the investment. What sort of pitch today with today's technology are those young businesses making? I think the smart ones have come to the pitch table with a bit more humility than we saw about 10 years ago. And I think there's more realism. And I hear less and less these sort of grandiose statements around our industry being fundamentally broken. I mean, I just don't sort of hear that kind of stuff anymore. But in terms of companies actually looking to get into the space, I think that they better understand that it has to be embedded in and around a traditional business outcome. It cannot create a situation that isn't really solving a problem. And so the best use cases that we're seeing of AI at the moment are things like claims fraud detection. So the tool is already modeled on a business outcome or it could be sentiment ingestion from an email or quoting from an API file or whatever. But it's not just what we saw 10 years ago, which is we're all really clever. We've built this really clever tech. You guys figure it out. And I think that that again speaks to the evolution of people that come in with these cases saying that we know the industry the problem you have with claims is this or we've got this. We've got that. So it's the far more user friendly, should we say? Yep. So it's a much more mature pitch than there was. Yeah. And I think most of these companies now go to the effort of consulting with insurance people. Again, we've seen much more of the insured techs looking to raise capital have got that balance between tech expertise but also market expertise, which we didn't necessarily see 10 years ago. And I think that they're understanding it's in their interests to come to the table with case studies. And to get the best out of AI, you've got to model it. So you've got to have a use case. And so it lends itself quite well to expediting the conversation because you've got to do that anyway. And so they're coming much more as a service proposition to the insurance and so they're not coming to say, hey, I need a billion dollars because they also need an insurance balance sheet. It doesn't sound like that's the sort of thing. There's no, I've got some fantastic tools that the insurance industry is going to work. I mean, there are still some sort of MGA's looking to raise money and there are some digital full stack companies that like the label of insured tech. But I would say we are seeing a much greater move towards those companies supporting the existing landscape. Well, that was absolutely genius because of course, technology will get fantastic multiples of whatever you want to measure because basically it's difficult when there isn't much revenue or much profit or certainly no profit. And 100 times this or 100 times that way. And of course, insurance balance sheets are getting 1.2 if they were lucky at the time or hovering around book value. And of course, if you could apply the 100 times multiple just to a dumb bit of capital that is attached to risk taking in insurance, as far as I can sort of that is genius. Because there's the intellectual property capital, of course, which have a very high multiple such because it's very intellectual and hopefully it's incredibly valuable and brainy ideas should be highly valued. But of course, let's say capital isn't done. But one pound in my pocket, the same as my band in your pocket. So again, maybe if you're incredibly clever at allocating that in the risk insurance market, you also can be valued at three pounds and my value at 0.8 because I'm a dunce and you're very clever. Is that kind of order of magnitude? Not 100 times. I think as well, people are realizing what the value of that conversation is starting to look like. So again, 10 years ago, it wouldn't be unusual for us in one week to meet with say five or six companies that were all presenting more or less the same total addressable market figures to us. Just assuming that all of that business was right for the taking and all they had to do was put it into a tech enabled balance sheet type situation and you could then use the tan the total addressable market as some kind of lever to then say you're worth 20 billion dollars. There should have been more maturity and more common sense in some of those things. Big up your Swiss Rea Saidmann, it says PNC's worth what seven trillion. I think yeah, sometimes seven trillions. I don't read it every year because it doesn't change that much. But yes, you are complete fool if you think you can apply or I'll get 0.1% of that and then I'll be worth this. Exactly. But that's what we were seeing and I think it's been refreshing to see sort of common sense come into the conversation which been wonderful. Right. What was going to move on from each other because your report because I think you took that figure of 95 point four. I don't know what what are the other four percent being invested in? Who knows? That's probably still there. Also possibly AI. They just don't blame themselves. Because the rest of the report is going really deeply and really interesting in the AI and it's I think now is the absolutely the time to be doing this. So let's ask a generic question about how do you view AI as an emerging casualty peril? There's a lot in your report about this. It's really interesting and obviously there's nothing to see some insure to some new businesses starting to address this. I think we're going to see a trajectory that follows quite closely with what we saw cyber experience particularly in a sort of 2017-2018 space where there is undoubtedly a risk class that emanates from digital exposure or digital outputs and the market is going to have to determine whether or not the lost manifest sits in existing classes of business or in fact there is some sense in having a single business class i.e. AI liability that picks up the unknown that people lean into and I'm almost certain that that is what is going to happen. So some of the DNO policies, some of the ENO policies will probably expand some of their coverage and be clearer on certain language around AI but I think that AI liability as a risk class will stand up alone and we're already seeing that with companies like Testudo and Armilla and I think that will continue to grow and I think the markets are pleased because it's not opportunity to grow but also it's an opportunity to become a more sophisticated but be put more sophistication into those traditional classes that are now under a little bit of stress because they're being interrogated in the way that they probably weren't five years ago. You know when you pick up your general liability policy you've got the war exclusion in there and do you think there will be an AI exclusion? Possibly. Yeah I think it's quite likely. You know out of your standard general liability I'd think back to you know I don't know whenever the image of the telephone was at 1895 was it in 1901 but you can imagine the liability under the author of those days saying these newfangled telephones are an absolute nightmare. In recent years before we was pretty clear whether we were liable not but at least in be a commercial liability. Everything had to be done. I had to send you a letter, you had to sign it and send it back and it was pretty clear what we'd signed. You know you've got my cup of your signature on it and I've got your cup with my signature and everyone can see what it was and we can argue about what those words mean but then it came to a moment suddenly it was people starting to do business on the telephone got forbid. You know people started buying insurance over the telephone didn't they? They took until about 1980s to start doing that but it came to a point where telephone was just embedded it is not excluded from general liability or from depression and identity because it's just the way we are. I mean AI is already becoming embedded and almost everything. Do you think you can exclude it? You know if you're standard general liability or clause says what I will ensure you for all the liabilities arising out of your normal activities as a company whatever it is that operates for clause at the beginning and then you have all the exclusions of the clarifications behind it but if AI is just what we do you know it goes wrong then we liable. Yeah no I think the guardrails will go in I think wording will become clear and I think to your point we are seeing AI permeate so much of society that it is going to be very difficult in some cases to distinguish between what is just a function of society versus what is AI specific and per the wording of your contract is delegated out but I think actually what we're seeing at the moment mark is that one of the biggest drivers of loss is not actually something specifically going wrong It's that people's expectations have been mismanaged about what the outcome should be. And so as an industry, we are just going to have to probably see some of those bigger losses come through and have the court litigate some of those guardrails about what is explicitly delegated AI that should be carved out from policies versus what is just part of everyday life now as we're evolving. Obviously, when you have very AI-specific companies say, "Hey, I'll come and do some AI-related thing that will improve your business," it's almost that the insurance could be a wrap around that company. Absolutely. You're making a lot of claims about this AI and we're going to help. And I think also what we will see in the courts is, let's take a chatbot output that gets taken to court over a claim that wasn't articulated properly. Is it the AI manifest as the issue? Is it the person that trained the AI model in the first place? Is it the issue? Is it the sentiment that was put into the modeling process over time that's the issue? Are you going to be able to get into the modeling cycle closely enough to see where the deviation could have come from and was that done by a third party? It's going to be a very interesting road ahead. But I think ultimately, it's great that the industry has been so forthcoming about this issue. I mean, nobody's hiding from this. And I think the fact that we can write 120 page reports about it is indicative of the fact that there's a lot going on. Back in the old days, I remember I was doing a liability business and you know, had electromagnetic fields. If you're doing the liabilities of a big utility electricity, this is a big deal. And there were plenty of underwriters back in the 90s who would say, "Mark, let's not talk about this. Just a minute we talk about it, we've got a problem. Can we just keep it silent, please?" Now, the industry has been very forthcoming. And I think, again, silent cyber has probably been a great, recent enough event for people to realize that there's actually no value in shining away from this. Let's talk about it. Let's experiment. Let's be pragmatic as a community, as an industry. And I think a lot of the innovation, somewhat perversely, that is being talked about in the liability case, might also be the same sets of tools to help us solve the problem. I think AI is going to come to its own rescue. When you go to an underwriter, there's always a moment in an underwriting conversation. Particularly when you're trying to do something entrepreneurial, there's a moment when they often say, "Well, this is really not insurable. This is not the sort of thing we want to get involved in because this is an enterprise risk. This is kind of an entrepreneurial risk that the venture capitalist should be taking this risk on our behalf. It's not really something that insurer should be doing." So, again, when you get to the point where you're almost trying to say, "I'm validating this AI and this process of methodology, do you think insurers would be comfortable with that?" Because you're getting very close to the enterprise risk of that AI business itself. Yeah. I mean, I don't think that's unique to AI. And I think every industry really goes through that situation. But I think that venture capital is there to stand up the businesses, the businesses are there to, if they're on the liability detection side to play a role where they're trying to help the industry move forward. If they're in the production side, almost inherent of the nature of AI, without very close caption, modeling techniques, it will start to evolve in a way that requires a lot of scrutiny. But I do think that insurance has got to play a role in this because there are going to be losses and there are going to be losses that are going to be very complicated. And I think for our society to be able to continue experimenting and getting the best out of AI, insurance has got to come along as part of that journey. I mean, I can't imagine a situation where we as a society more broadly could be contemplating autonomous vehicles or delegated workflows or generated chats without the safety net of somebody's managing the financial downside. It would be a very different exploration into what is potentially the most important technological innovation ever. So it has to play a role alongside the venture. Yeah. And your report has already identified problems because there was an interesting detail on there that AI is causing losses that are half of them are not being insured and what sort of losses are these? Well, I think a lot of them probably at the moment fall into those traditional, is it DNO, is it not, is it ENO, is it not? And I think because we're not at that stage of maturity, it probably is falling into those traditional losses. I think as well, particularly in the states where a lot of the litigation is happening, I think it comes back to the point I made earlier, which is your expectation of what the outcome ought to be and is therefore not covered because you never explicitly said in the first place what you thought your expectations were. But because we don't have a hugely mature AI liability specific class of business that everybody has access to, all the other claims are having to pick up the slack, essentially. Yeah. So it seems inevitable, of course, because regulations way behind, I mean, it's bound to be things are moving so fast we're talking about in the last couple of years. I mean, I started asking people about AI only a couple of years ago. And now I can't not talk every conversation. Every conversation. And we're talking to the point where, you know, big up, any news, even front page of a news moving faster than us. We're talking about, you know, potential cyber liabilities. We're talking about today with the Claude Mythos being withheld because it could undermine the internet as we know it. Absolutely fascinating. So, litigation is just bound to happen in the absence of very clear new laws and experience moving ahead so much faster than regulation and the legislation. Would you say, litigation is just absolutely inevitable? I think so, I think particularly as it's burgeoning and evolving, I think a lot of what we will later agree on as sort of hard and fast rules will probably have come out of the courtroom. So I think it's inevitable. The way of attacking this, and again, you have to read this report, it's absolutely amazing. There's so much in this, but you get to the point where you're starting to highlight a pathway to solving this problem and you highlight some of those businesses which have reminded me that I must get them on the show now. Absolutely. So I really want to talk to them. But one of the methods, of course, how do we evaluate the efficacy of an AI or say, is this AI better than the other AI? You know, these are two risks that have been presented to me by broker. I'm an underwriter, which wonder why I think is better than how do I go about creating a methodology by which I measure, I model, I say, this one is better than that one and this one should have a lower premium than that one. Yeah, I think it's the same as any pricing discussion, it's understanding what the long tail impact might be of it going awry, but also there is a huge amount of court data that will give you an indication of what things typically end up costing. But these are always within the context of physical losses or casualty losses. And so I think that the actual indemnification part of it is probably still quite well defined. I think where the real headache is is a to what extent has the AI model been trained and left to its own devices in that sort of a genetic way or to what extent have things been determined that people don't really understand. But without having the ability to scrutinize that as the underwriter, I think people are probably still taking most of their indications from what's the worst case scenario. Let's say there's a manufacturing plant using AI to build things that makes blow. It's understanding what the amplification nexus is or there being AI versus not being AI. And I think again, it's going to be a series of real big losses that help an underwriter ultimately understand how to start thinking about this risk. There's a great section in your report about the ways that AI has been evaluated to date. What struck me was with say 98% of AI's can do what a PhD student used to be able to do 10 years ago. And then it's getting more and more advanced. And obviously these are sort of tests set by a human to say, right, AI, if you can do this, then you're done well. You pass in sort of a turing test to all of you want to call these things. You wouldn't need to be very imaginative to think that there's only been a couple of years. You know, we've already get into the point of most PhD students are not going to be clever enough to set a new test for the AI. So what point do we have to start using AI to test AI? Oh, which case are we getting to metaphysical question here? Which AI do you trust? I'm a firm believer that humans are never going away. I think that ultimately if we're not consuming something that's a net positive, then what's the point? And I think even with the most intelligent AI tools, most of them are still, if they're text AI organization, they're still scraping what they want to be raw material, which is inevitably have come from human beings. And you're not going to necessarily any good at predicting the future because when you go to sort of Nazim Nithlas Talab, you know, the old black swan, fantastic, everyone has to read the black swan, you know, there's a chart of relative happiness of a turkey ever its lifetime. And it goes from sort of first of January, and then it goes up and up and up and up and up. Suddenly it gets to 24th of December. And it's, oh, because it's Christmas. And it goes back straight to zero. So it's about putting things in the right context. Yeah. And at the same time, I suppose you could have the AI that was doing an amazing job of predicting hotel occupancy. Uh-huh. Fat way, we have some parametric insurance products around 2019. And suddenly we had the pandemic and suddenly, oh, God, it's just exactly like the turkey is now had its neck run and it's now in the oven. And everything that happened before is now all over. So one thing these humans we've got on our side is that we are better adapting to new data. We've got experiences, right? And even that example, I mean, the chances of a global pandemic happening are small but not zero. But if the model doesn't have the new answer ability to factor in the unlikely, but possible, it's never going to show up in its predictions. And again, I come back to this point of if we're not ultimately seeing a benefit from it, then I don't think the use cases are going to last very long. Now, is there going to be disruption along the way? Absolutely. But there are jobs that exist now that didn't exist 50 years ago. It didn't exist 10 years ago. And I think that as a very bullish, optimistic person, AI is actually going to create more opportunities for us to be even more creative. creative and more human, ironically. On the creativity question, it's more specifically about cyber, but then I expect a large percentage of that 95% of all the funding is going to be something cyber related, or a lot of them will be anyway, it's at the say it's a large cohort, so it's probably worth talking about specifically. With AI now, a genticae in the hands of hackers, we've always got that eternal question of, you know, we've got nuclear power for good, your nuclear weapons are at a bad. We've got that asymmetry always with something like cyber, whereas you've got a thousand holes in your system, you plug 999, you thought you did a fantastic job, but of course the hacker in need one. So it's asymmetrical, it's not fair, they only need one to get in and make a loss. For defense, a genticae AI, you're sort of hamstrung, you're slightly bringing your knife to a gunfight, because you can't let it wildly loose on your system, because things might happen, it might shut your whole system down and then no one can get cash out of the cash machine and you're in deep trouble, whereas of course a hacker doesn't care, they can just let it loose. So there's that asymmetrical problem, they can do stuff that you can't. Do you think we're ever going to keep up? There's probably the question in the cyber world, particularly now, obviously it's become a massive cyber question. Yeah, I do, I mean, I don't have a free, I'm with the first Blade Runner movie. I always sort of think of that scene where the machine is being interrogated and they're looking for that one thing that lets the interrogator know, oh, they're not human. And I think there are technologies now that are allowing us to stay apace and I think ultimately it's going to come down to who wants this more. And I think the side of the defender usually wins, but the issues will continue to evolve, but we are already seeing AI being used to detect AI with things like cloned for detection, specifically with pictures of smashed up cars, right? It sometimes is quite hard to detect what is real and what is not. Because of course, yeah, fraudsters are now using AI to have fake crash pictures, but they can actually be trained to detect what is AI generated. Similarly, a sort of presentation a couple of weeks ago where somebody was doing an experiment on faked hail damage on US roofs and they were just using a hammer to sort of smash holes into the roofs and the AI tool could detect almost immediately what was fake and what was real. And so I think the kind of catch 22, like I say, will be about who's prepared to put more people at work on this particular project to kind of win the arms race, but with AI, the vector keeps changing, right? Every time you think you understand something, it has evolved by its nature, but I think that on balance it's a lot more positive to come out of this than those sorts of negatives. Will you say I'm very optimistic? I am a very optimistic person. Well, again, I keep going back to this report because it's so impressive. I think everyone listening must read it. And I just thank you for increasing the general IQ of the debate and the level of the debate it's fantastic. Is there anything that you'd like to talk about we haven't spoken about yet? Something sort of specifically, but you sort of started the conversation so accidentally when we were even revisiting some of those old technologies. And I would say to anybody that is particularly concerned with this right now, there was a period of time when we were all told the blockchain was going to completely transform the industry. No one talks about any kind of distributed ledger technology ever. And so I think there are lots of apocryphal tales in the text scrapyard that surrounds our industry. I do think AI is fundamentally different. I don't think it's a technology that sort of antiquates itself, but deployed in the right way. I think it makes our industry more interesting as ever been. And I'm personally really, really excited about all of the opportunities that it presents us with. So the idea of the enhanced underwriting, enhanced broker is a very sensible one to cling on to and not being sold apart. No, I think as long as the human expert in the loop is being celebrated and I think as long as the technology is doing what it does best, which is predictable, menial, wrote tasks that free up people's time to focus on growing businesses, being strategically more adept, enhancing relationships with people. Then I don't see how we can get this wrong. I think standing around and doing nothing is probably the worst thing we could do and we're not doing that. Let's suppose it's the way that we're going to reduce the protection gap. Yes. Because we're going to bring the cost down finally. Bring the cost down, but also technology by nature is transcendental right so we can get into pockets of demographics of people that have historically been blocked by physical barriers and the cost of an intermediary being less of an issue means that certain lower cost business is now profitable, which means that we can get coverage to those people. And I just see as a win for everybody I really do. So basically there's going to be more opportunity for every broken underwriter to see how they're not going to be able to make it in a very bullish person. Yeah, I think so. Excellent. Well, thanks so much. Thank you so much, Mark. Thank you. 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 program. 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. 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 podcast is produced in association with Advantage Go, enabling enterprise-scale underwriting through a single pane of glass. Voice of Insurance is produced by me, Mark Gagan. Music was written by Anna Gagan and produced by Carlos Gagan. Check out more podcasts and written comment pieces at www.thevoiceofinsurance.com.