Reinsurance's Digital Frontier: Data, AI and the Tech Wave | Emmanuel Clarke, BMS Group

by Juan de Castro, COO Cytora

In the latest episode of Making Risk Flow, host Juan de Castro speaks with Emmanuel Clarke. Emmanuel wears many hats, including serving as Chairman of the Board at the BMS Group and Compre Group, and as a Senior Advisor at McKinsey & Company. Industry insiders speak of Emmanuel as a living legend in the insurance sector. With over 30 years of experience in the industry and multiple enterprise leadership roles, Emmanuel is a performance-driven, strategic, and execution-focused leader with a proven track record for leading global and culturally diverse teams. Together, Juan and Emmanuel highlight the digitisation challenges that both insurance and reinsurance face, and why they are more similar than they immediately look on the surface. Plus, the duo also cover how AI and data could be used to evolve reinsurance workflows, the impact seasonality has on the quality of data, and why it’s crucial to digitise unstructured data as soon as it’s received.

Listen to the full episode here

Juan de Castro: Today, I'm joined by Emmanuel Clarke, who's the former CEO at PartnerRe, amongst many other roles we'll talk about in a minute. First of all, Emmanuel, thank you so much for joining me today. Let's start with a bit of your background. A couple of weeks ago, I was talking to somebody in the industry, and they referred to you as, “Oh yeah, Emmanuel, figurehead in the industry”. So I'm sure many of our listeners know who you are. But if you can start with an overview of what you've done in the past, that would be helpful.

Emmanuel Clarke: Yeah, in simple terms, it's 25 years in reinsurance. I started as a specialty underwriter. Actually, my first line of business was credit and charity. And I started with PartnerRe, and I spent 25 years at PartnerRe and moved up from being a specialty underwriter all the way to becoming CEO from 2015 to 2020. So great company, a great ride. And throughout the 25 years, I've probably seen every line of business, every type of transaction, every geography. So a fascinating experience. And since 2020, I've embarked on a more plural activity. And I have my fingers, my hands in all parts of the value chain of insurance from distribution all the way to legacy.

Juan de Castro: And then we want to get deeper into the reinsurance side of things in just a second. But before we do that, you mentioned you've built a diverse portfolio of non-exec rules. Tell me a bit more about that. They feel like they've got some commonalities across them, and it's a very interesting portfolio. So if you can say a bit more about why you chose those opportunities and what are you excited about?

Emmanuel Clarke: Yeah, so a lot of exciting things here. First of all, I'm one of the people who think insurance is exciting. A lot of people think insurance could be boring. I think it could be super exciting. There are a lot of great people in insurance, and I'm at a stage in my career where I can actually pick and choose the people I want to work with. I don't necessarily describe my role as non-exec, although it is actually true. It's non-executive, and I enjoy this, to be honest. But I describe it as bringing my experience, my network, my views to investors, executives, and boards, to help accelerate value creation all along the value chain of insurance. I do work with a number of different companies with very different forms. It could be from ventures all the way to large public companies. So very different forms of ownership, very different sizes of companies. But beyond the fact that it's always with great people that I actually enjoy working with, the common denominators are, yes, it's insurance. And that's all I do. The second one is I try to pick companies where there is a shared vision for progressive insurance. What I mean by this is how can we actually harness technology, new ways of distribution, data, to make insurance a better product for both the end customer and the end shareholder. Which brings me to the third common denominator, which is value creation. It's how can we actually enhance the way we do insurance to help create more value? So that's what I do. So it's all the way from distribution of insurance or reinsurance through technology, insurance, reinsurance, and legacy.

Juan de Castro: And including private equity type of companies.

Emmanuel Clarke: Yeah, they could be family office-owned, venture capital owned, PE-owned for two or three of them, and public company.

Juan de Castro: So one of the points you made is the concept of progressive insurance. Actually, it is a term I hadn't heard at least referred to in that way. But in the end, it's how can technology data or innovation in the industry drive a better outcome for both clients and insurers and reinsurers. And as you said, you really span from the broker all the way through the insurer and reinsurers. You have to cross the whole value chain. So when you think about it, what are the major opportunities in the market?

Emmanuel Clarke: I think the major opportunities are, the way I think about this is threefold. One is all that has to do with efficiency. I mean, if you look at the money that you spend on your insurance policy, whichever it is, and what gets into the pocket of the ultimate risk holder, there's a significant portion of that premium that goes into friction costs. And I think the industry has tried to lower these costs over time, but I think there's still an enormous amount of potential on how we actually make the insurance product ultimately a more efficient one. Everybody talks about the protection gap in insurance, right? With large populations not being insured and not being enough insured. And I think one of the key levers of expanding that or reducing that protection gap is actually making the insurance product ultimately cheaper. And if we can actually reduce some of these friction costs, we'll ultimately make the product more affordable. So one of the things is efficiency, clearly, through automation mostly because there are still a lot of manual tasks that are involved through the insurance process. 

The second thing is service. I think we've talked a lot about insurtech and then the rise and I don't know if it's a fault, but a lot of what insurtech has actually brought to the industry is improving service, improving client experience, making user experience more friendly, reducing response time. There are a lot of things that we can also improve here. And I have to say that from a personalised perspective, I think the industry has actually done quite a good job at digitising large parts of the workflow. So that whenever you and I have a claim, you get responses from your insurance company in a much faster way today than you used to. But here again, I think there's also further potential to make service better. 

And the third one for me is just profitability. You can actually enhance the underwriting, augment the underwriting. We always tend to think that humans can actually make better decisions than machines, but actually, a lot of the tests, whenever you pilot things, you actually realise that not only machines can make as good decisions, but sometimes better decisions. And for me, these are the three levels where innovation can really break through: efficiency, service, and profitability.

Juan de Castro: Just touching on the three of them. So on the efficiency one, we've been talking, so the industry has in some buckets, like a 40% expense ratio, right? Which is, I think this is kind of what you were referring to about the friction cost, which ultimately results in products being more expensive than they should be and the potential under reinsurance. How ambitious do you think the industry is around this? Because often we talk about how to shave one percentage point out of the expense ratio, but should we be thinking, how do you move that 40 to 20 to 10, right?

Emmanuel Clarke: Yes. I think what you're referring to is should the industry think about this in incremental terms or in more transformative terms? I wish the industry could actually take a less incremental view on this, but to be realistic, one of the forces that actually inhibit some of that ambition is just the existence of very large, old, legacy systems with a lot of insurers. And that's been a drag through innovation because a lot of the budgets and a lot of the resources in most of the largest insurance companies in the world still go into cleaning up, improving, bringing together, and integrating legacy systems. But to your point, I think in personal lines there have been significant gains. I think most of a large portion of the personal lines workflow has been digitised already and has actually brought a lot of efficiencies. But as you move up the chain from personalised to commercial to reinsurance, I think this is where further potential gains lie.

Juan de Castro: Is a strategy then think big, so think how do you actually work from a 40 to 20 expense ratio type of thing, but then get it there incrementally? Would you favour that approach?

Emmanuel Clarke: So first of all, you're picking one of the three and we need to work with the three there. I think there is a need for a vision of where we can actually get to as an industry. And then the question is, how long does it take to get there? What are the steps to actually get there? Also, I think that we're entering a decade where I believe there's got to be more differences generated between those who really embark, and harness innovation and technology earlier than some of the others. And this is going to create a gap in terms of creating these efficiencies and creating more competitiveness in the market. So I don't think you can speak on behalf of the whole industry, but I think some of the winners of tomorrow will probably be bolder in how quickly they can get to a significantly different efficiency model.

Juan de Castro: You touched on efficiency, service, and profitability and how to augment underwriting. And then the three of them are very linked to each other. So when you think probably about augmenting the underwriter, it's definitely about risk selection, but it's also about how you reduce the amount of effort it takes to underwrite a risk. So there's an efficiency angle as well as a risk selection angle. And I'm not sure if you saw this analysis from the World Economic Forum, where they did an analysis across all roles, across all industries. So this was not insurance specific to what were the roles with the highest potential of being impacted by artificial intelligence, and underwriters was the number one. I mean, the number two was whatever, architects. It was from completely another industry, the number one, which was underwriters, which is first of all, very fascinating that the current Economic Forum knows what underwriters are.

Emmanuel Clarke: It's actually surprising.

Juan de Castro: Yeah, exactly. That is exciting in itself. But the point I was going to make, which actually is something I really agree with, would love to hear your thoughts on it, they differentiated what the impact of technology or AI was going to be in those roles. And they've differentiated automation from augmentation. And they said 100% of underwriters' roles will be impacted by augmentation, but very little by automation. So it sounds like that is aligned with what you're saying. It's like, do you think the actual opportunity is more around augmenting the role of the underwriter and really making the underwriting process more efficient, better risk selection? Would you agree with that?

Emmanuel Clarke: That's an interesting thing. I actually think it's going to be both because in everything underwriters do, I mean, let's take personal lines out because it is already heavily digitised and automated. So focus on commercial and reinsurance. In both these segments, you're going to find some smaller transactions, mid-sized transactions, and some large, complex transactions. As an industry, I think every company is looking to actually shift resources from demanding the labour-intensive ones in terms of volume to where it should actually be more labour, where labour should actually be spent on and deployed, which are the high-value transactions. And so how do you get there? You get there by automating some of the smaller end-of-the-segment transactions and augmenting some of the larger transactions. You see what I mean? And actually, a lot of insurance or reinsurance companies have been trying to do this over the past few years. I have found a more efficient fast-track process for smaller trees, for reinsurance, and for packet data. We'll talk about this later. Or for mid-sized commercial risk. But overall, the purpose should be spending the resources on where it actually moves the deep needle most. And so on smaller transactions, that's where actually there's a space for automated underwriting so that you can actually free up resources and invest in the right resources for some of the larger, more complex, more value-add transactions.

Juan de Castro: I think that resonates very much. That is also how I see it. You were talking earlier about the more forward-thinking insurers, the ones taking decisive action, and often the way they describe their vision is, let's take an underwriting process. Early on in the process, be able to triage which ones are the simpler risks, and which are more homogeneous, you can consider straight-through processing, and only involve an underwriter, as you said, where it moves the needle in those more complex cases that really require that judgement, and then enable that underwriter to be most effective at underwriting or analysing those risks. That is all about making the role of the underwriter more exciting and impactful and better risk selection. But it also goes back to your efficiency point, right? Because you should be able to then write twice as much business with the same underwriting team, right?

Emmanuel Clarke: 100%. And all these three levers actually connect to growth. Because at the moment, I know a lot of companies who are not in a position today to handle all the incoming pipeline of deals, simply because they just don't have the manpower to do this. And particularly in some areas of the business where there's some more seasonality. And so you're down to saying no to smaller transactions and so on. You just don't have the time to process them or they're not economical to transact. If you do put in place all the, what we talk about, the automation, you can free up the time, more time for the right transactions. You can give some quicker responses to the brokers or to clients in terms of some of the smaller transactions. And that all helps you actually transact more business in the end.

Juan de Castro: As we said, let's dive deep into reinsurance, because I always feel like when you hear people talk about the future of the industry or disruption, often reinsurance feels like an oversight or an afterthought. It's rare that you see people talking about innovation in reinsurance or even insurtechs trying to drive value in reinsurance. And I always thought, why? Is it harder? What are your thoughts on why that's been less top of mind?

Emmanuel Clarke : That's a great question. And there are some facts, right? It's probably a part of the industry that is not as well known. It is smaller than in the grand scheme of things of insurance. There's less volume of transactions. The number of transactions is definitely smaller than in commercial risk and in personal lines. So there are some basic facts that explain why this may not have been the prime target for innovation or disruption. There are also some more questionable arguments like reinsurers, and I can talk about it because I've been in one. You always feel like your business is different. Everybody feels that their business is different. But as a reinsurer, you feel that your business is different because it's more about judgement, and relationships with your clients and your brokers than it is about your product being a commodity. So that may have been friction to innovation. Also, the fact that data, when it comes to reinsurers, it's not as granular and detailed as what actually the insurance companies have. And so data could be unstructured, and may not have the greatest quality. So then the question about, do you have enough pools of data that you can actually work and leverage to make automated decisions? And then the third potential reason is reinsurers are portfolio managers and they build portfolios of risks because they are the ultimate holders of risk. And so they may consider that you're not necessarily looking at each transaction on its own, but you're looking at a transaction against the portfolio. And that's why some of these underwriting decisions don't necessarily lend themselves to automation or augmented underwriting. But I do think there's absolutely no reason, personally, I think there's absolutely no reason why reinsurance couldn't leap forward in a meaningful way in terms of leveraging data, but also artificial intelligence to actually make the business better. Make the business better, again, on efficiency, on service to your clients and the brokers, and on improving combined ratios.

Juan de Castro: So it sounds like you do see an opportunity. Earlier you mentioned, I believe you were referring to the levels of commoditization where you were saying personalised, commercialised brands. But it's not that insurance is so much more complicated. Do you still see that there's an opportunity there?

Emmanuel Clarke: I definitely see there is opportunity there. I'm not sure about commoditization. Commoditization is a tough word, I think. Although there are elements of commoditization of the business, right? If you just look at what happened for Property Cat over the last 15 years, it's hard to argue that the business hasn't actually become more commoditized. But I actually think it's making the reinsurance product more closer to a financial instrument than to what it is today. It's getting rid of a lot of the manual tasks. I mean, it's amazing how much the workflow is still very heavily manually processed. Why? Because reinsurers get submissions from all over the world. And there are as many submission formats as there are seeding companies and brokers in the world. And so this involves an enormous amount of manual steps in the process that a lot of them I think could actually be simplified and augmented. And then when it comes to the underwriting decisions, I actually think there's no reason why that couldn't actually be, again, as we discussed, some of the small midsize trees of this business that fall under the well-known criteria. Actually, you should be able to send back a quote or send back an acceptance or a declination to your broker within a couple of days so that you actually spend the time on transactions that are of higher value. So yeah, to your question, I think there is definitely room for that.

Juan de Castro: So when you think about the major pain points in reinsurance, you've mentioned one, which is heavily manual, the number of different formats.

Emmanuel Clarke: I would say unstructured data. The quality of the data depends very much on what you get from the seat. And so it's not necessarily equal. So that's another pain point. And I'm trying to think objectively of the real pain points and not the excuses.

Juan de Castro: One of the things that feels different also is the seasonality of the business. Is that also a factor?

Emmanuel Clarke: This is definitely a factor that should actually promote more automation. Because in reinsurance, what you get is that, I think at the worldwide level, probably 65% of the business is transacted on the 1st of January. And then the other big dates are 1st of April, 1st of June, 1st of July. In Europe, I think it's greater than 80% of the business transacted on 1st January. So what it means is that reinsurers have a workforce that is actually calibrated to handle renewals at the end of the year. But even then, it's actually a stretch. And so if you find a way that all this peak season can actually be handled differently with a number of things that could be done either before or quicker. And underwriters being able to look at files that have been really prepared for by the tool, by whatever system you use, I actually think that should definitely help with the handling of the seasonality, which in itself could also help, back to your point earlier, look at more business. Some of the business that you're just not capable of handling at the time of the year. You should be able to increase the fishing nets just because you have more capacity to look at things thanks to technology.

Juan de Castro: And this type of business, which is very seasonal. I mean, even outside insurance, right? Other ones, almost that these should be looked at first when thinking about how to drive efficiency and augmentation. As you said, otherwise you need to staff your workforce for that peak throughout the whole year. So any way that you can mitigate those peaks, obviously will make the whole insurance company massive.

Emmanuel Clarke: Absolutely.

Juan de Castro: So obviously you spent lots of years in reinsurance. And if you fast forward, whatever, 5, 10 years, and you have a more digitised reinsurance market, what does it look like? What are the things that should look differently in the future?

Emmanuel Clarke: I think you could actually envision a process where the submissions come in, they immediately get into the reinsurer system. The transaction format gets converted immediately into the reinsurer's format. You may actually have an AI agent looking at whether the submission is complete, anything missing, or whether the data is actually of good enough quality. In an automated way, you can go back to the broker and say, can we please have this? Can we please check that? And so on. And then once the submission is complete, standardised and so on, the data should flow through the experience, the lost data should flow through automatically the pricing tool of the reinsurer and come back with a quote, if it's a quote that you need to provide, or a profitability metric analysis as proprietary to the reinsurance company. And then there should be clearly some transactions where you should be able to send a declination to the broker right away. You know, this is not in our appetite. Sorry, you know, we're full on this, whatever it is. So that, and this should actually back to the broker fairly quickly. You don't need a lot of time for that. There's a bucket for transactions where it falls into the fast-track process. In other words, we know the risk, we've been on this, this is not a new business. This is renewal. We're comfortable with the changes. We've looked at the pricing in an automated way. And so here's the response back. And then there's a bucket for some of the newer, larger, more complex business where the transaction falls on the underwriter's desk. But that could take a day or two instead of two, three, four weeks right now. And so compare the underwriting decision and say, here's what we looked at. Here's the pricing. Here's what comes out of the pricing. Here are the changes from last year. You can even integrate there because I think we haven't talked about this, but from part of the underwriting decision, it's not just data and pricing. It's also legal contracts. And so there's an enormous amount of innovation in AI for legal documents. You can also prepare your other changes. And I'm not just talking about red lining some of the contracts here. It could actually be a new contract and you can have the AI analysis just telling you the things we need to change in the wording to make it more robust and solid. From our perspective, the underwriter has a decision and is almost prepared. It could almost be a recommendation. You can also parameterise all this to make sure that you're working towards a portfolio that you're trying to build as a reinsurer. So in other words, the system can actually say, yes, at the moment we have more appetite for the marine business, but we're full on agricultural business, for instance, or you could actually change the diet and just dynamic pricing, there's a lot of things that just can be prepared in an automated way so that the underwriting decisions get, again, faster and better. And so some of them being automated, some of them being just augmented. I think that's the ultimate vision. Sounds great, but I don't think it's very far away. I think it's definitely achievable.

Juan de Castro: Yeah, but you say if you compare this with some of the digitised work, we're saying in primary insurers, it looks very similar. So you start with a challenge of very unstructured data, right, that is needed to digitise up front. I think you're advocating for digitising the risk as the first step in the process to enable the rest of these steps downstream, right?

Emmanuel Clarke: Exactly. And on this point, Juan, you asked me earlier, what are some of the reasons why it hasn't happened yet? And I've seen this throughout my career. I think reinsurers were kind of paralysed by the fact that the data was so unstandardised across the industry. And there's been a number of projects within the industry, sponsored by the industry, to try to bring some standards and the definition. But this is impossible, first of all, to achieve. And actually, you just don't need this because AI can actually do this for you, for your company, in a much better way. You don't need to standardise the industry. You just need to convert what you get from many different formats from your clients into the one format that works for you. And that's going to be very efficient for your system. And I think the industry has spent an enormous amount of time spinning wheels on trying to find that standardisation when actually AI is actually bringing this to you. It reminds me of an historical analogy. In the 17th century, the countries in Europe were all investing in their warships and the Marine Army. And to build more ships, you had to plant more trees. And there was this prime minister in France called Colbert. He planted a massive amount of oak tree forests in the 17th century because he wanted to plant for the next three centuries in terms of having enough wood to build the ships. And actually what happened later was the Industrial Revolution brought all the steel there. I mean, it's great to have all these oak tree forests. But just what we're seeing right now is the equivalence of an Industrial Revolution for the financial services business. And so all what we've been trying to work on to try to bring all this standardisation, it's not relevant anymore because the tools are there to actually do all what you want and need for your business.

Juan de Castro: That is actually, Emmanuel, music to my ears and brings me back memories. So when I started this podcast three years ago, the first episode was a solo episode of me sharing my views on the industry and why I wanted to do the podcast. And I called it the Industrial Revolution for Insurance. And the Industrial Revolution was hosted by new forms of fuel that enabled automation, etc. And to your point, data and AI is a new fuel for insurance. And I think you summarised it really well, which is we've been trying to solve this problem of unstructured data. You think about it in many different ways. So we tried to create standards that failed. Then each insurer and reinsurer tried to create their own broker portal to capture the data in their own way. However they overlooked the fact that brokers don't want to input data into 15 insurance portals. And some risk to go through placement platforms in mid-market. Like, Acturis in the UK, etc. But you need a way of really standardising or really being able to communicate effectively across the whole value chain between thousands of brokers and hundreds of thousands of insurers. And I think rather than trying to fix it in those ways, I think now technology is creating a new way of being able to take completely unstructured data from a number of different brokers and convert it into whatever data you need to power your internal processes.

Emmanuel Clarke: Exactly. I think one of these issues is standardisation. The second one, I referred to this earlier, was that people assessed the quality and the granularity of the data that was coming up to the reinsurers, which is not enough. And so reinsurers have all invested in data science teams, which is great. And, you know, this is the way to go. But sometimes with limited success, because you bump into, well, there's no data in there. If you have the right tools, AI-driven, to actually go and fetch the data and enrich your databases on an everyday basis, just to make your underwriting decision, your data is just going to be incrementally greater every day. And all this data that reinsurers are looking for, it's out there. You just need the right, you know, innovation, the right technology to actually go fetch it, store it, enrich it every day.

Juan de Castro: So without trying to put words in your mouth, Emmanuel. But to me, three takeaways from what you just said. I think one is to be able to drive the vision of the more efficient and high-performing insurance industries about how to digitise risks upfront so that you can abstract the complexity of different formats and be able to use that data. Second is how do you then deploy digital underwriting work fields that allow you to embed that external data or internal data or those internal data science models into the evaluation of risk? And then the third thing is how then are you able to triage the risks as they come in to enable what you were describing some of them will be declined, some of them will be straight through process, and some of them will have to go through an underwriter. Almost like, if you achieve those two things, it feels like then the reinsurance industry will be in a better place. But I think the thing that is very exciting is it feels very similar, both the pain points and the solutions to the primary insurance world, right? So I think we should just get going. Almost, I feel like there's no excuse.

Emmanuel Clarke: Yes, I agree. It's exactly that. It's no excuse. And I think I also mentioned portfolio management earlier. Similar to insurance and personal lines and commercial, there's a need to, people want to make sure that there's a lot of proprietary intelligence that's going to go into this. This was my reaction to commoditization. So in other words, I think reinsurers will continue to have different views about different risks because that depends on their capital structure, on their appetite, on how aggressive they want to go about cycle management, for instance. There's a lot of things that are proprietary in the way reinsurers will want to differentiate and outperform the rest of the pack. But all this should not be an excuse not to actually use technology and innovation. All this should actually get into the parameters on how you actually build and equip your technology to take all these factors into account so that you can actually drive to the best results. So it's definitely feasible.

Juan de Castro: Very exciting. Well, Emmanuel, as always, it's always a pleasure catching up. It's always a pleasure hearing your views on the industry. I always feel very energised after some of these discussions. So thank you so much for joining me today.

Emmanuel Clarke: Well, thank you, and similarly, you can feel that I'm actually quite passionate and excited about some of the things that we hope to change and will change.