Exploring AI Adoption, a Board Level Conversation | Jeffrey Hayman

by Juan de Castro, COO Cytora

This a shortened version of Making Risk Flow podcast, episode: “Exploring AI Adoption, a Board Level Conversation | Jeffrey Hayman". In this podcast episode, Juan de Castro interviews Jeffrey Hayman, a CEO, Chairman, board member, and entrepreneur in the global insurance industry. Hayman has an impressive track record in executive roles at carriers like AIG and, more recently, in non-executive roles, including his current position as Chairman of the Board at R&Q Insurance Holdings Ltd. They discuss various topics, including the state of the market, the board-level conversations about AI adoption,  and the use of external capital in the insurance industry. Jeffery also mentions his involvement in investing in and advising promising startups and shares his experience and insights for executives leaving large organizations to set up their own ventures.

Listen to the full episode here

Juan de Castro: Hello, my name is Juan de Castro and you're listening to Making Risk Flow. Every episode I sit down with my industry-leading guests to demystify digital risk flows, share practical knowledge and help you use them to unlock scalability in commercial insurance. Welcome to another episode of Season 4 of Making Risk Flow. Today I'm talking to Jeff Hayman, who's an investor, advisor, entrepreneur in the global insurance industry, with quite an impressive track record, both across executive roles at carriers like AIG, but more recently with non-executive roles, including his current role as non-executive Chairman of R&Q Insurance Holdings. In this podcast, we'll touch on the state of the market, how AI is discussed at the board level and many other exciting topics. But Jeff, it's such a pleasure to have you on the podcast. Really thank you for joining me.

Jeffrey Hayman: Thanks for inviting me.

Juan de Castro: I've given a very, very short overview of your background that spans many, many years, but can you give us a more detailed summary?

Jeffrey Hayman: So I got a many, many instead of just many years. The industry has given me all the grey hair you see. It’s been 41 years. I spent 15 years just in the US domestic market with my first employer, which was Travelers. I spent 15 years with AIG, 12 of them living in Tokyo. In the end, I spent a couple of years as the CEO of AIG's global consumer insurance business. I had spent six years before that as the CEO of AIG's property casualty business in the Far East, which was their largest business outside the United States at the time. Having left the executive ranks, I embarked on a non-executive career. I have spent five years on the Board at Zurich Insurance Group. And as you mentioned, now I'm chairman of R&Q, which is much smaller compared to Zurich or Travelers or AIG but is still a global specialty insurance company. Bermuda based London Stock Exchange listed and operating in North America, UK and continental Europe. But I do other things as well. I invest in and advise promising startups and my wife and I formed a small family foundation and I sit on the board of another local charity. So we try to be involved in philanthropy. I like to say I'm on a couple of boards that pay me so that I can afford to be on the boards that cost me.

Juan de Castro: This is the reason I wanted to have this chat with you, Jeff. So you have this huge executive experience with non-executive roles but very close to innovation and to what's going on in the market. One of the topics I would love to get your thoughts on is in the last few years, five, ten years, there's been a quite significant capital injection from third parties into, well across the market, right? From MGA to distribution to setting up full carriers. What are your views on where that external capital is coming from and being invested in?

Jeffrey Hayman: I certainly see a significant investment in private equity-backed enterprises, not just in the United States, but in the UK and Europe as well. A lot of them are in distribution. And I think a lot of people think of roll-up plays where you get a broker who rolls up other brokers and then takes out expense, providing common administrative services, has a boost in profit margin, ends up with a bigger profile as it relates to negotiations with carriers and all of that stuff. But there've been other types of investment and distribution as well. I think, people in big companies, I don't want to use any company names today, but big global brokers can sometimes, as I did, in a big insurance company, get tired of working for the big global insurance company with all the bureaucracy and overhead and all that kind of stuff. And they want to do something more entrepreneurial. And private equity is backed up those kinds of companies. I'll use one as an example, a friend of mine is an advisor to McGill and Partners, which was a startup backed by Warburg Pincus, I think it was. So there's lots of entrepreneurial and innovative talent being backed by private equity and distribution, but also in the MGA space. I mean, MGA's are doing things from D&O, right? And professional indemnity, all the way down to pet insurance and everything in between. You know, it's not just non-standard auto insurance or CAT-exposed property. It's a whole bunch of folks with market niches and specific expertise, right? And the investment in MGA’s has helped facilitate a further disintermediation or re-intermediation of the market. It used to be an MGA, I'll simplify it, it used to have the choice of basically just going to one of a number of large insurers. You're going to go to AIG, you're going to go to Zurich, you're going to Travelers or Chubb and you're going to want their paper to manage your program. And one of two things would then happen, right? If the underwriter didn't like the program, they'd pull their paper and leave the MGA’s scrambling to try to find capacity to keep their business alive. Or if they did like it, they'd want to take an ownership position in it and eventually force the founder out and put their own people in, right? And so a whole new industry, some of which is private equity backed as well, has come along in the fronting or hybrid fronting carrier space, which is one of the businesses that R&Q does now, program fronting or hybrid fronting. Where they support MGA's and place the risk with a panel of high quality rated reinsurance capacity, keeping a small amount of net and operating for a fee. And that's a much safer pair of hands for innovative MGA's. And it's a great choice compared to just going to the big company with the big balance sheet. A lot of this has been facilitated by private equity money. A lot of value has been created and a lot of innovation has been released.

Juan de Castro: And why do you think some of those MGA's are set up outside the large corporates? Where you look in the world, what you were describing could be also be done. Like a large company could spin off an MGA too, right? So, is it because it's less capital intensive and can be done as a venture?

Jeffrey Hayman: So I think there's a combination of things, you know, it's very difficult, at least from my experience, I don't know the whole industry. I spent 15 years at AIG and 15 years at Travelers. I don't think my board experience at Zurich qualifies me to know what was going on under the covers everywhere in Zurich, but it's very hard in a big company. To have small company entrepreneurial things going on because of the governance and oversight and the layers of management and all of that. So I think people want to be truly entrepreneurial and truly own what they're doing. There's a big element to that, right? And then the fact that each of the things that I talked about in terms of being areas of private equity, investment, and innovation are capital light. There are private equity firms and other investors who invest in balance sheet heavy, but you know, insurance enterprises as well. But these things, you know, the hybrid fronting carrier offloads 90% to 95% of the risk to a reinsurance panel and therefore has a smaller balance sheet and less capital. It's a more capital-light business. The MGA's are a fee business. The brokers are a fee business. And, you know, underwriters sometimes struggle through the insurance cycle to post a reasonable return on equity. Whereas these feed businesses or these, you know, EBIT earnings for interest in taxes businesses tend to post great profit margins. So it's attractive from an investment perspective and folks feel ownership and control of their own direction.

Juan de Castro: They also have like a clear exit strategy, right? So they can become a feeder for the all those roll up gains later on, right? 

Jeffrey Hayman: For an MGA, an MGA can grow, can expand beyond its initial core program into other related programs, can buy other MGA's or can become a full stack insurer over time as well.

Juan de Castro: Do you think the legacy technology is also a consideration? So obviously when you're in a large corporation, you're tied to legacy, not just process, but also technology, architecture, et cetera. And when you set something up from scratch to some extent you're more nimble too, right?

Jeffrey Hayman: That depends on the nature of the company and the organisation of the company. Suppose it's a large company that dictates what technology has to be used by the business from top to bottom. And puts everything through a lot of hurdles to make decisions, it can be difficult to do something new. But, you know, even big companies are successful in launching new things with new technology, right? I can't even tell you how many years ago, so 23, 24 years ago at AIG, we launched a new direct-to-consumer business in Japan in a country where they'd been in business since the 1940s so they had a lot of very old legacy technology. We were able to front end with that, with what was then modern technology, as long as we were then able to feed all the end result through the old backend, right? So you can front end things and you can do things that are a little easier to accomplish, a little more customer friendly in terms of the interactions that you can generate through hand-helds or, you know, back in those, we were just trying to get quotes done over the internet. It's a lot more complicated and a lot more customer-friendly now. But I think the concept is still there. As long as the new operation has some leeway to use new things, use new technologies. Yes, they still have the overhead of having to be able to pass that information through the old back end, but you can still get things done.

Juan de Castro: This might be a bit of a tricky question, but you're talking about executives eventually leaving large organisations, setting up some of those new brokers or MGA's, what advice would you give?

Jeffrey Hayman: I probably wouldn't give them any advice because my recent attempt to be entrepreneurial wasn't particularly successful. One thing I left out of my history is I took about two years off between leaving the Zurich board and joining as Chairman of R&Q. I was leading a SPAC, a special purpose acquisition company, attempting to get an IPO done and then try to bring one of these probably PE backed privately held companies to the public markets and it flopped. The capital market situation was just impossible. There was no way to get a SPAC IPO done and we eventually gave up.

Juan de Castro: Was it because of the environment or was it hard to do regardless of the investment environment?

Jeffrey Hayman: Yeah, I wasn't the only one who had a difficult time in that particular environment. In fact, we counted ourselves lucky. We could have gotten the IPO done if we'd started a few months earlier, but we would have been in a market where it was very difficult to get a deal done afterwards as well. But I say that somewhat in jest. I did leave the safety of the board of directors of one of the largest financial institutions in the world to try something entrepreneurial. I had an idea of what I wanted to accomplish. I had people co-investing with me who had complementary skills. We thought we had a great team. But you know, I reached out and talked to somebody who had done it. The co-founder and until recently CEO of the company Vault, which is a high net worth. Personal lines insurer recently launched in the US. Was a guy I worked with for years at AIG. And so I picked up the phone and called him and said, you've done the jump from the safety of the big company paycheck to doing something entrepreneurial. And I got advice from him, you know, his advice, I'll boil it down to: If you know what you're doing, just do it. But you need to have a business plan. You need to have backers and partners. You can't do everything yourself. You also need to be willing to take the risk. And not every new venture succeeds. You have to be able to go through some iterations until you get something right. I only did one iteration and then I went back to the safety of working on a board so I might not have what it takes to be an entrepreneur, but that would be my advice to others.

Juan de Castro: That is still quite good advice as well. Be ready to take the risk, be conscious about the backers and the partners to make it successful and work with somebody who's done it before. Still, obviously, success is not guaranteed, but at least you're maximising your chances of being successful. So that's quite useful. So one of the things I mentioned in the opening was this topic of how is artificial intelligence being discussed at the board level, right? Many of us are working in AI startups live and breathe it every day, but actually don't have that much exposure to how it's been discussed at the board level. So give us a bit of an insider perspective of what's the board's perspective on AI.

Jeffrey Hayman: So you can imagine the boards of big insurance companies aren't packed with a bunch of super tech literate people. Most of us are getting pretty old, and may not be up on everything that's happening. Have a lot of executive experience and hopefully some good judgement, but we really have to rely on someone to tell us what's happening in specialised areas. Not directly AI, but I'll draw an analogy to vehicle autonomy. Cause I think it may go through a similar path. So we spent a lot of time in my final years as an executive and in my years on the board, like what is vehicle autonomy going to do to motor insurance? How quickly is it going to happen? And who's going to bear the risk for liability and whatnot. And how are we going to do deals and what's going to happen to basically a line of business that makes up 60, 70% of the property casually market globally. So we were interested in it. Didn't understand the technology, but we would have experts briefing us. And we thought, certainly I thought, in 2014, 2015, that by now, there would be a significant penetration of autonomous vehicles on the road, and it hasn't happened. The technology has advanced significantly and it hasn't happened because regulators and other humans are uncomfortable with machines making value decisions on prioritising the occupants of the vehicle over pedestrians outside, for example, things like that. So ethical decisions made by machines have slowed that down. Similarly, AI, in my discussions, insurance regulators, at least as it relates to policies sold to individuals, are very protective of individuals. There are lots of markets that want to make sure products that are sold are adequately explained. And the reason why people are charged, what they're charged is adequately explained. And that the end result is a suitable product and the outcomes are suitable for the insureds and all of that. You know, a black box making decisions is not congruent in that situation. So we've used AI to highlight things, but in a way that a human can understand and explain to customers. In areas like risk segmentation and rating plans, in areas like fraud detection, those types of things. So I think the presence of regulations, at least as it relates to consumers, is slowing this down a little bit. But then in other areas like the proliferation of chatbots and production of communications and all those types of things. I play around with ChatGPT and I think it already writes better than a number of people that I've worked with. But it also makes stuff, I don't know if you played around with it. So we have to be careful with it. But I don't understand it, I'm sure as well as you do. And so someone like me, in his mid-60s, who's at the end of his career and is just trying to give good governance and oversight to an organisation, is reliant on the experts to help us understand what's going on, what the risks are. And so what's important is to make sure that the board continues to get development and training on emerging technology and issues so that it can adequately govern. Not just to constrain the company when it's going too far, but not to constrain the company when it has an opportunity.

Juan de Castro: So you touched on the point of like training the board on the details and what's happening in this space. But I found particularly interesting your point about board being less comfortable with AI or technology making decisions on behalf of a human. So is your hypothesis that the use of AI is going to be more focused on how to enable humans to make better or faster decisions?

Jeffrey Hayman: So I think in the near term, I would say yes. And it's not so much I don't think I said, or if I did say, I didn't mean to say that the board is worried about black box making decision. I think regulators are worried about machines making decisions. And therefore the board is worried about it and wants to deploy the insight that can come from machine learning, AI, any other statistical analytical techniques that are hard to explain to the average person. But that they have to be deployed in a way that we can make a consumer understand that we have not made, or to make a regulator understand that we have not made an unfairly discriminatory decision.

Juan de Castro: This is often when the industry talks about the difference between an automated decision versus a human on the loop decision. An automated decision is where the machine is making the ultimate decision and there's no human supervision whatsoever versus a human on the loop decision is where the technology is really providing the insights and the recommendation to a human, but it's a human who's ultimately making the final decision. So is that the type of model you envision, at least in the near term, right?

Jeffrey Hayman: Yeah, I think that's fair. In other industries, there's a lot of debate over AI in creative industries right now, and there's strikes going on. You know, writers recognize I think that AI is going to be useful in the production of content, but they want to be the ones in the room prompting it. Having a dialogue, if you will, with AI about how to create something, as opposed to being replaced in the process, right? So the notion of working cooperatively is important to a lot of people in a lot of industry, customers sometimes for the protection of employment. What I've been focusing on more thinking about it is. As in the vehicle autonomy example regulation and concerns about ethical decisions by machines, I think has slowed it down. I can't be the only one who thought by the mid 2020s there'd be a lot of autonomous vehicles on the road, and there are not, right? Same forces, I think, will be at play for AI and how it's used in businesses, at least as it relates to impacts on consumers. Business to business is generally a different thing.

Juan de Castro: Which insurers we make all the time, right? Just to offer a cover or not, or to extend the policy or not. And one of the related to AI, obviously, is the topic of using broader sets of data. I don't know if that's externally available. We see a number of use-cases where underwriters are using external data. To inform some of the AI models and make some of the predictions. Which on one side can be extremely powerful, on the other side, we know it drives initial scepticism. What's your view on how insurers are making use of external data?

Jeffrey Hayman: I think my first real experience with that concept was being part of a team that brought the credit file into underwriting and pricing decisions while I was at Travelers in the early to mid 1990s. Now, the leader in the space at that time, I think, was Allstate, but we were doing our own work and not far behind. And you know, I remember we brought in the credit score, which was a score created just for credit purposes, not for insurance purposes and we drew a line, people with a 700 credit score are higher, and people with 699 are lower. Right? And then we did all the best business just to find all the best business that we thought from a traditional underwriting decision making process was best. And then not the worst just everything else, right? So the best versus not the best. And so you had kind of a four box grid. You had above 700, below 700 credit score, and you had the best business and not the best business. And the 700 versus lower 700 broke the loss ratios between 40 and 80. But the traditional underwriting broke it between 50 and 70. So it was not that much difference. But what I did was I gave it back to the actuaries and analysts and said, along with others, said the same thing, not just me. “Isn't the combination of these two things more powerful than just saying, you know, wow, credit is so much better than traditional underwriting. And in fact, it was. Because traditional underwriting could subsegment an initial credit-based segmentation and credit could subsegment an initial traditional underwriting basis”, right? So what that said to me, and it carried on through my career is more broader data sets are better. And in fact, instead of just using the score, we should use a score that's tuned to insurance, not just a credit. And then we move on from there, we should just use the credit file and do our own analytics, not use somebody else's score, right? So that's the progression of things. An example I wasn't involved in, but watch was progressive and it's promotion of telematics, you know, snapshot. It gave people incentives. It must've been an outrageously expensive endeavour for them to incent millions of customers to put this dongle in their dashboard to get all of this acceleration, deceleration, mileage, sharp turns, hard stops, all that kind of stuff. This proved incredibly insightful because they were an insurance company with a huge data set over which to lay that additional data and do all sorts of risk segmentation, right? So if you're a big company with all sorts of underwriting and pricing attributes and you add additional information to it rather than trying to replace it, there are a couple of examples of how that was very powerful. But it's really expensive. And a lot of companies are a lot more focused on controlling expenses because expenses are easier to control than losses, or claims payments. But you have to take some risks to do these things. If I could just add like one small story. I am invested in a startup that is using sensor technology to measure real-time, minute-by-minute human occupancy in buildings. So they're doing it for advertising attribution, like in big box stores or for billboards and what that are placed outside. But we saw insurance application to that, those of us who looked at it. It could be basically building telematics. So, we think that data source could be very useful in different types of occupancies, being predictive of liability losses, but perhaps other perils as well. But it's difficult to find someone who's willing to make the investment to build up the data for long enough to be able to then analyse it against losses. Because a lot of people, again, are focused on expenses. I think a lot of people in the industry are just very conservative by nature and hesitate to accept new things until they've been proven. But somebody has to prove it, right? So I continue to support this particular, you know, little startup and try to introduce it to folks. I think at some point, our optimism about its usability in the space will be vindicated. But it's somebody who has to be willing to take the risk. I've given you two historical big wins in bringing additional data sources in. There has to be more potential for that.

Juan de Castro: But to me, almost the key message from those two examples just described is very impactful. So when it works, it's really impactful to use the credit score to identify the better risk or the telematics in the cars. But it's often really hard to capture the data. And I think there are two categories into that. One is that, in some cases, you have to deploy the whole hardware infrastructure to capture the data which by itself obviously is expensive, but often data does not require physical devices. It's just data that is available that it's unknown whether it correlates with loss ratio or loss experience or not. Isn't what insurance should be thinking is, how do you create this flywheel of having a hypothesis of what data could correlate with loss behaviour, capture the data, correlate over time, correct it with loss behaviour and then accept or reject. So if insurers had that flow in a way that would be repeatable almost like every week, is that, do you think that would address some of the friction that prevents insurers from using that data currently?

Jeffrey Hayman: Yeah, I think it's the cost of getting credible results. I think that holds a lot of people back. So. If there is a way to introduce new data sources in a cost-effective way. And if the company or a third party who's assisting the company is set up in a way to take advantage of adding that data to the data set that's used for analytical purposes, for pricing, risk selection, underwriting, whatever you're talking about. That then I think lowers the hurdle, right? What's important, I think, is that the company that has a large portfolio of risk, immediately has an advantage over some startup that's got interesting technology, right? I mean, so there were other telematics examples of companies that were startups that said, we have a better version of telematics. But they didn't have a big portfolio to use in tuning how to use those telematics in risk selection and pricing, and they had a more difficult time. So the sweet spot is the additional data for a company that has a credible level of exposures and has the data set up in a way that it's easy to analyse segment slice and dice. And you don't really need years and years of exposure to get credible results. If you apply something to a large number of policies for a relatively short period of time and your actuaries will tell you if you're doing a split test, you know, how many claims you have to have in each cell in order to have credible results. It's only around a thousand claims, right? So they can get credible results relatively soon. I'd be an advocate for having a portion of your budget set aside to test the power of new data sources. I can only influence one company at a time now.

Juan de Castro: But also just going in a full circle to where we started this chat. We were talking about large corporates versus the newly formed MGA or underwriting companies. That is almost where the tension starts, which is like large companies will have those large portfolios and historical data, but potentially it will be harder for them to be agile enough to capture new data. Whereas those newly formed MGA's are already thinking about how to capture as much data as possible but will not have the historical portfolio data, right? So I think there's a balance there between the two and is there a solution for that?

Jeffrey Hayman: No, I mean, there is a learning curve necessary, right? So the new startup, if it's with someone who really understands the market and has in their head a good idea of how to select risk or price risk or manage claims so that there's no worse experience in claims than what you had in your previous organisation. There's still going to be a learning curve because back in the old days, I remember somebody who said, well, all we have to do is copy XYZ competitors' rates. Minus 5% and our results shouldn't be worse than say, “you know, whatever the rate level difference is” like, no, you don't, you can copy somebody's rates, but you can't copy how they select their risks by who they appoint to sell their policies or their geographic concentration. You don't end up getting the same basket of risks as the competitor you're copying. So when you're new, you can be really smart, but you can still be surprised. In the early days of your portfolio. But hopefully, these folks are smart, entrepreneurial, make adjustments quickly and can take advantage of the situation.

Juan de Castro: It has been a fantastic chat, Jeff. I really enjoyed how we've almost gone full circle in the conversation from our in-chief through private equity, through MGA’s, how AI is discussed at the board level to data and like driving innovation in both small and large companies. So, thank you again for joining me. I've really enjoyed the episode. Hopefully, I believe you told me this was your first podcast. Hopefully, you've enjoyed it still.

Jeffrey Hayman: It's not the first podcast I will have listened to, but it's the first one I've contributed to.

Juan de Castro: Exactly. Well, we enjoyed it. Thank you so much, Jeff. 

Jeffrey Hayman: Thanks, Juan. 

Juan de Castro: Making Risk Flow is brought to you by Cytora. If you enjoy this podcast, consider subscribing to Making Risk Flow in Apple Podcasts, Spotify, or wherever you get your podcast so you never miss an episode. To find out more about Cytora, visit cytora.com. Thanks for joining me. See you next time.