In this episode of Making Risk Flow, host Juan de Castro speaks with Johan Slabbert, an Independent Senior Advisor for Insurance Advisory Partners, to delve into the transformative impact of AI and data analytics on the insurance industry. With over 30 years of international experience, Johan has held key executive roles, including CFO for Latin America and the Caribbean at AIG and CEO positions at Augustus Specialty, Chaucer Group, MS Amlin, and MSIG Holdings. Johan’s global journey has undoubtedly enriched his expertise across diverse insurance markets.
Together, Juan and Johan explore the transformative role of AI in underwriting, particularly in managing the increasingly complex landscape of data integration. They discuss the balance between traditional data sources and new digital data streams and how AI-driven automation is unlocking more dynamic approaches to risk assessment. Johan also sheds light on the shift from simple system upgrades to broader, strategic transformations within insurance, emphasising how AI can support portfolio-level decision-making while reshaping the role of the underwriter.
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. Johan, thank you so much for joining me today. It's such a pleasure to have you on the podcast. Let's start with a brief introduction and background of yourself.
Johan Slabbert: Certainly, Juan. Thanks for having me on the podcast. I've been in the insurance industry for about 30 years now. I started on the finance side with AIG in South Africa, where I was regional CFO for the Africa region. I was Chief Investment Officer for the Africa region and also ran the risk finance portfolio for a period of time. Following a few years in South Africa, I was transferred to the UK, where I spent three years as a senior vice president on the finance side. Following that, I was transferred to Japan for six and a half years, where we all lived through the wonderful financial crisis, including the AIG crisis. My responsibility there was the Far East. I was regional president, and I was on the board of not just the property and casualty business, but I was on the board of the holding company, securities, trading, real estate, and a number of the supplemental entities AIG had there. Following the financial crisis, I was transferred back to London, working on the IPO, which is then called Chartis, which was the property and casualty business of AIG, trying to separate ourselves from AIG, the holding company. Basically, it was not concluded on, but a year later, they asked me to go to Latin America, which is one of the regions I've not been involved in up to that point in time. So I spent two years on the Latin America side before they asked me to move to New York to run the emerging markets and growth economies, out of New York. I left AIG and joined the Hanover Group, Hanover Group, Worcester, Massachusetts. They immediately asked me to relocate back to London to run the Chaucer Syndicate, which I then became CEO of Chaucer, which is the Lloyd Syndicate that has now subsequently been sold to China Re. At that point in time, I was approached by Leon Black from Apollo Global Asset Management to set up a new company in the US, which I did go and do. Unfortunately, two years into that, we had COVID and a number of other issues that emerged. So I sold my equity back to Leon and came back to the UK to run Amlin. So I did that for a number of years. And then Mitsui asked me to go back to the US, lots of travelling, but back to the US to run their North American business, which I did until March this year.
Juan de Castro: That is quite impressive. I don't think you've missed many continents, Africa, Asia, Europe, and the Americas, it’s fantastic. I think you've got a great global view of the insurance industry. You've had executive roles either on the finance side or as the chief exec in a number of different insurers. So in the era of data and AI, how do you see the transformation of these insurers? Are they adapting? Do you see any type of shift in the way they think about how they leverage technology and data?
Johan Slabbert: Yeah, look, I would say the success or the lifeline of both a CEO and a CFO is the quality of the information they receive to be able to make decisions for the future. And bear in mind, a lot of the traditional data is historical data based on performance, based on activities that have already taken place. I think where we are as an industry, we're seeing a lot of transformation taking place. And I use that very lightly because it's not a transformation in technology. It's a transformation in strategy. And that is by better utilising data. And I'm not talking about just the historical financial reporting, portfolio management, risk management type of data. I'm talking about new sources of data. We now live in a very digital world. And a lot of the data that is being produced is not necessarily being used. Insurance companies can, to a very large extent, tap into those new sources of data, combine it with their existing data, to get a very different view on risk. So a simple example, and I'll maybe use two just to elaborate on the particular point, but I'll use an existing company. Historically, commercial buildings have just been based on historical renewal data, nuances, depending on whether it's an office block or other type of commercial industry. But having companies that manage elevators, you get a different risk because that elevator, now actually produces data, that data can be utilised by third parties to determine foot traffic, determine hours of operation, determine a whole bunch of factors that have historically just been a piece of paper Q&A filled in by the insured. So that data can enhance that particular risk. So you get a better view. I think a more look at some of the consumer lines examples, historically, a motor insurance policy is pretty vanilla. You use it for work or you use it for personal use. And that's how they would underwrite it as a personal use plus you use it for work or you just have it for personal use. If you look at an Uber driver, an Uber driver now has multiple data sources for you to look at to distinguish the four different categorizations of the usage of that vehicle and its independent verification rather than the driver giving you what he should give you to get a lower premium. So, if you look at an Uber driver as the owner, he parks the vehicle on his premises either inside a gate or outside a gate. And that's a GPS can tell you that whether it's factual or not. The second categorization is when that Uber driver uses that vehicle on the weekends for his family. It's not a commercial use. It's a private use. But your liability potentially is different because it's not just the driver and it's not just stationary. The third aspect to that is fee for providing a ride. So the app has been switched on. You now have a new source of data and he's going from point A to point B working for the Uber company. But there are no passengers. And then the last categorization is when the ride starts, your liability changes again because now you have passengers with you. That is a differentiation in a single risk. Historically, look at it as a single vehicle, one driver type of scenario. Now you can look at a very different element and you can vary the actual liability in each of those four categories or the actual value of damage to the vehicle. So it's very interesting that data can be used to supplement your traditional information in a very, very different view of historical risk.
Juan de Castro: So you mentioned at the high level three types of data. So historical data, which is the data that every insurer holds in their systems, the data you receive from the broker or from the client, and then you're talking about external data that you can get from, in this case, in your example, from the Uber app. I think each of those comes with its own challenges. But do you think underwriters, most likely, use most of this data today when analysing a risk? They look at the documentation provided by the broker and they might go to a few external data providers and they might look into their policy admin system to see similar risks or existing exposure to that risk. I think what you're trying to get to is it's less about how you look at a risk, because probably a lot of that data is being used today, but it's how do you combine all that data into a streamlined underwriting process that doesn't require the underwriter to be looking at, 100 of pages and internal systems and 10 external websites.
Johan Slabbert: I think that's a very important part of the underwriting process. I think it's how it's utilised. And we can talk a bit about AI in terms of the process. Using third-party data typically is for validation. Does he have a driver's licence? So you go to a third-party website, you do that. So there are many examples of where you use third-party data. But when you have feeds from third-party coming into your environment, that allows you to look at a different data model that gives you a different view on that risk. I think that's a step forward. And I think that's where the future is. I will mention, though, that there is a downside to leading this. And the downside to that is, if I use the Uber example, and I actually rate every one of those four buckets separately and aggregate them, I'm more than likely going to come up with a more expensive price, which means that Uber driver is going to go somewhere else. He's going to have a loss, and it's going to be way more than the expected outcome of that particular underwriter because he's using a traditional rating methodology, where if you do it right, you're going to probably struggle to compete in certain buckets, but you'll do really well in your own portfolio where, you have a low-risk portfolio, but you won't be able to grow until everybody's using the same mechanism. It's a bit like telematics.
Juan de Castro: But even though you would struggle to grow, wouldn't you be selecting the right risks?
Johan Slabbert: You absolutely will. And by theory, 50% of good drivers, 50% are not, because the average is somewhere in that. So yes, but you will be selecting the right risks. You'll be pricing them appropriately. And maybe that's the right approach to take. Let somebody else take the bad risks and let them learn that the rates should be higher, or they should be looked at the risk differently.
Juan de Castro: So should we jump into the role of AI and perhaps keeping the theme of these resources of data, historic data, broker-provided data, and external data? Do you see a specific area across these three where AI is being most impactful?
Johan Slabbert: I think, first of all, many people use AI in the same context as system upgrades or automation of technology, which we're going to be very clear about, those things have existed for decades. Automation, AI is not, especially generative AI, is not just an automation of a process to give you an end product. So I think it's important to basically narrow down what you want to use AI for. And AI, obviously, we all talk about the language that you're going to use is going to be in an environment where you can have reliable data, I can almost certainly guarantee you that even the traditional corporations, the large multinationals, their data is archaic. They've probably gone through a number of systems transformations, not updated the data capability to bring the old data through to a new platform. They probably got it archived somewhere. So it's going to be incomplete. I guarantee you that I've looked at a lot of insurance policies where a standard code has been used, not a specific code, and it could be a postcode. It could be a number of different things. So even your own data is not necessarily the most accurate data. But then for AI to be able to operate in that environment, you've got to have enough data. You've got to rely on the third-party data. And for it to be generative, you've got to set the rules for it to narrow down the risks that match your appetite. And once you've got to that point, it should then generate a majority of, and again, I'm going to distinguish here between consumer lines products, SME, and then more complex commercial risks. I think everybody in the motor or homeowners is pretty much pricing on a portfolio basis, not looking at individual risks. Yes, they go through a questionnaire and it ticks a few boxes. Therefore, you apply that rate or therefore you apply a different rate. On the SME side, a lot of the small businesses have vanilla products. So being able to say, right, I want to sell X million of that vanilla product. Here are the rates. Here's the construction type. That's not AI. That's just an automation in a selection process. Being able to get to a point where, right, we've now narrowed down to risks that match our appetite for generative AI then to make decisioning around selecting deductibles, looking at your portfolio, not necessarily the policy holder. So that is the next step in terms of making more of those decisions that your underwriters would have done in the past. Now, I also want to say that when you get to the large commercial risks, AI can only go that far because to me, what I define as there's an art and a science to underwriting. The art remains the underwriter's understanding of a segment of a risk of the history of a reputation of an organisation and a gut feel for is this a good risk or not versus AI providing the science. And you have to have both. I don't think you're going to be displacing all commercial underwriters. You have so in a large aspect already done that on the consumer lines and you're halfway through sort of vanilla products on the SME side, but you won't be able to displace underwriters who really understand risks that have been writing that account for 10 years.
Juan de Castro: So you made an interesting point there, which is the obvious first benefit of deploying AI is driving some automation, but the ultimate benefit is about really supporting underwriters on selecting the deductible or really doing all the pre-underwriting work for the underwriter and potentially some of them fully underwriting automatically. But I think one of the things we see is when insurers are considering, I think this is where also your CFO background comes into play, I would love to hear that perspective when somebody struggles to build the business case around the latter. Totally, they say, okay, let's build a business case on the automation. So how much underwriting capacity is this going to free up? What's going to be the impact on the productivity of the underwriting teams? Because the second part, which is the improvement on risk selection, it's less tangible for a business case. So if you were back in your CFO, CEO role, how would you look at it?
Johan Slabbert: I would say I would love the opportunity for AI to do a lot of the check and balances during the underwriting process. So let's say we've allocated a certain amount of capital to a particular class of business, particular product or segment. An underwriter would not necessarily have built in until after he's bound the account, he adds it to his portfolio and looks at it. Ooh, I'm over that. Now I have to buy reinsurance because I've actually exceeded a particular geography or limit. Being able to, during the underwriting process, do a lot of that. Check capital utilisation, check portfolio balance, check limits, check reinsurance as you bind the policy. Is it covered in IEXIS? Things like that, that it could assist the underwriter in getting to a conclusion a lot quicker and then decide, do I like this risk or not? But for a CFO, it's a hard task because once you get to a quarter end, if you haven't had all those checks, you really look at the portfolio and go, ooh, we're writing more property than what we intended to. Now you've got to buy reinsurance. Now it has become inefficient. And the underwriter's already spent so much time trying to bind that account. Your operation's engaged in doing that. Your claims people are running. So it's very inefficient where if AI could do that. Throughout the process at different stages to check capital allocation, to check limits, to check reinsurance coverage, to check a number of things. It just makes the organisation, again, the risk selection becomes so much better and more efficient.
Juan de Castro: So going back to one of the points you made earlier, which is going from the systems transformation. So I think you mentioned the industry a few years ago was very focused on system transformation, and now it's more focused on strategic transformation. We touched on a number of points around it, but can you elaborate a bit more on that?
Johan Slabbert: Look, I think there are obviously different markets. And if I look at the current trend, it's really about strategy. I think systems transformation has become secondary to the strategy. I think as we've seen the volatility, and there are peaks and troughs to this, as we see the volatility in pricing, as the rates over the last couple of years have come up, it's now plateaued. The questions are now around the strategy. What do you do when rates come down? And that can only be done by looking at data, historical trends, looking at sources of information that typically have not been available to CFOs or CEOs. It is now available. You look at actions that are taken when rates come down. Do you lower your limits? Do you go to excess layers? Do you ride the wave longer than what you did in the past? Reduce your risks very quickly. So those types of trends, it's not necessarily on individual policies. But it's on the portfolio and the decisions CFOs and CEOs need to make, which is a timing aspect as to when rates change. So that's just an example. So it's a strategy transformation and has become more pertinent now. We're sitting in a very interesting market where rates have started to plateau and you're getting a bit of a bifurcation between property and casualty. So it's going to be interesting to see who responds, who manages it well. And there are examples. If you again go and have a look at the third quarter results, you'll see some people have beaten their expectations or estimates by 20, 30%. And there are others that are announcing losses or substantial reductions to their profitability because they had too much property. They haven't made that step yet.
Juan de Castro: And would you say CFOs and CEOs need to look at and revisit those areas as rates start to go down? I think there's a component around making this strategic decision. But there's a second component, which is how do you then cascade those decisions into the frontline? Because it's fantastic to have a strategic board level decision, but then how do you enforce that through your frontline so that they aren't, to some extent you're requesting them to change behaviour from how they behaved in a hard market, right?
Johan Slabbert: Yeah, look, it's not necessarily a massive step change. It's utilising all the tools that are currently available in terms of you change price, you change limits, you change deductibles. And those are the nuances that say I can reduce, obviously, utilisation of capital is another one, but I can change the participation without losing the account. Losing the account means you're going to struggle to get it back in a future state. And they're going to go, well, hold on. Now the rates are going up. You want me to come back? So it is really, it's about utilising different tools that are available to the underwriters, making them think about the risk differently. And you can look at it on a return on capital basis for individual accounts.
Juan de Castro: So we bring all of these together. If you fast forward one, two years, I don't think it's much longer, how does the role of the underwriter then evolve? So you're thinking now you're providing all this data to the underwriter, digitised data from the broker, digitised data from external data sources, you're providing suggestions or steering on deductibles and limits, et cetera.
Johan Slabbert: I would say it depends on the quality of your workbench. And I use the word workbench because that's what most people in the industry are referring to. And that's the platform available to the underwriter and what information or data is available to him on that platform, on that workbench. That will allow him to make the right decisions as the rates start coming down and the market starts slowing. If it is the old traditional, here's your paper application, or here's the standard renewal information, I don't think he's going to be in a position to decide if I change deductibles or change limits, what does it do to the capital and the risk portfolio further downstream? So it is really dependent on the quality and more importantly, the underwriter's understanding of the risk return in the organisation. Underwriters, and I'm not belittling underwriters at all, but they're very focused on their underwriting on an account-by-account basis as they should be, but not necessarily always aware of the consequences of each account. If you ran it individually through a risk model, what would that look like? We run portfolios through risk models and we look at the outcome. But if you run individual accounts through with the portfolio and then the portfolio without that account, you can see the impact of that single account. They're not necessarily that in tune with the financial mechanisms at the back end as to whether they should or shouldn't write that account. I think if your AI has the ability and has enough examples of looking at data saying, well, this is going to be a, isn't a great of, to the return on capital, if you increase deductible, therefore, you will have a smaller loss than what you would have had pre-underwriting. So the underwriters need to not just have those workbenches, but also get a better financial understanding. And there are a lot of them that do, and those are the ones generally who do better in the market than others.
Juan de Castro: So you really, I think if I play back what you said, really underwriters will spend less time just handling documents and extracting information from documents, like the more manual routine, probably lower value activities. Technology will also provide them with insights on the risk, and impact of writing that risk and your capital position, et cetera. So you really are shifting the role of the underwriter towards probably still underwriting the single account, but more forward looking, what is the impact on the portfolio and the rest of the business, right?
Johan Slabbert: Absolutely. And that's the COO's responsibility to give them that capability.
Juan de Castro: Definitely. Obviously, so now you've got also a kind of advisory roles and you have quite a broad perspective of the market. What would you tell the industry is in terms of alignment with the vision you just described and getting on board with these more strategic transformation. And in general, what type of carriers do you see are taking the lead on this journey?
Johan Slabbert: Yeah, Juan, so as you say, I am currently serving as a strategic senior advisor to Insurance Advisory Partners, IAP, who are engaged with mergers and acquisitions in the insurance sector. What we do see is a combination of a number of things, in particular, buyers of insurance businesses who are investing for two reasons. One, is to expand on their distribution or their capabilities. And quite often, the criteria that they include today is technological capabilities and digitization of the companies they're acquiring. Because many of the traditional companies want to quite often see that sort of capability working. So if they can buy a distributor or a broker or another smaller carrier, or whatever the case is, that have those type of capabilities, it's sort of a benefit because they can then deploy that capability into the market. Clearly, it's not the primary. For the large strategic buyers. However, I would say there's also a significant amount of investment in that technology space, in insurtechs, in technology suppliers and services around the insurance space. Some of the larger European companies, insurance companies, have funds that specifically focus on that. Some of the Japanese multinationals have funds specifically focusing on insurtech. With the intent of deploying that capability, the AI, the technology, the utilisation of that within their existing businesses. So instead of going out, finding a third party to deliver that service, they would rather go out and buy somebody who has already demonstrated that capability and then slowly deployed within their own organisations.
Juan de Castro: And when you think about those type of insurers who have already made a leapfrog in terms of technical capabilities and deployed some of these, do you see mostly smaller players taking the lead on these? How does it change between the large global players and the smaller ones?
Johan Slabbert: The larger multinational insurance or global insurance companies, for them to do an acquisition, it's got to move the needle. Now, bear in mind, if you look at some of the very large organisations like Allianz, Mitsui, AIGs, Chubb, for them to make an acquisition, it's got to be equative and it's got to be meaningful because quite often transaction costs for some of these smaller transactions are astronomical. And you don't generally find a huge amount of technological advances or advantages in the companies they acquire. So I would say the middle and even the lower end of the market are the more strategic ones in terms of acquiring businesses today. They look at businesses either their size or smaller that substantially change the equation for them, and they bring in capability that they haven't previously had. So, I would say early startups get picked up if they're really good at what they're doing, get picked up pretty quickly by mid-size carriers or organisations. The larger ones, as I say, for them to do a strategic transaction, it needs to be big. If you're writing billions of dollars, you need to buy a business that can actually do anything from, I would say, 10% to 20% of premium that is accretive or growth within your portfolio. Otherwise, it's just not worth it for them.
Juan de Castro: It's too small, yeah. So that is from an M&A perspective. From an organic perspective did you see differences or similarities between how these global insurers are going through this strategic transformation in terms of the large players versus the tier two, tier three ones?
Johan Slabbert : Yeah, I think the large globals have all got a desire to get a good understanding and use cases for AI. I think go back to their data, as I mentioned before. They have very, very huge data depositories that have been basically stored because it came off a different platform. It's not usable in the current platform. So they are missing a big chunk of their capability. I think in terms of the larger carriers, they have to resolve their internal data before they actually merge anything with external data. If they can get to that stage or even maybe rightfully draw a line in the sand and say, we're only going to focus on data that's relevant, and that's when rates started coming back up. If you do that, then you may not know what to do when the rates start going back down. So it is a tricky one, but I do think they all have a desire to enhance their data. They all have a desire to have use cases, not necessarily currently applying AI, but use cases as to how can we do claims differently? How can we do operational things differently? How can we streamline the underwriting process? But I don't think anybody is, to a large extent, deployed AI for a product, for a segment of their business.
Juan de Castro: Yeah, and I see huge investments in this space. So I was doing an episode with Jillian, who's the chief innovation officer at Aon, a few weeks ago. And I think Aon is investing over a billion dollars in sorting out their internal data and infrastructure around this. But I think one thing I mentioned to you earlier, which has surprised me in the last, I would say, probably 18 to 24 months, is I would have expected smaller insurers to have been the ones taking the lead, on adapting GenAI and AI on some of these use cases. And what surprised me is actually, it's been the largest global, so the Allianz, the Chubb, the ones that have one, probably the strongest vision around how is this going to transform their business, but also the highest level of urgency to stop actually benefiting from it. Whereas some of the smaller mid-sized ones are still, in some cases, we hear things like, oh yeah, we've identified 400 use cases, where GenAI has the potential to drive impact and say, okay, and how many have you deployed? Well, none yet. And I think that sort of decisiveness and urgency is something we see more on the large players.
Johan Slabbert: I would, however, say that I think in terms of state of readiness, I think the smaller or midsize carriers are in a better state of readiness to utilise or deploy AI than what the larger carriers are. So, yes, they're going to have to do an acceleration of their internal data cleansing, enhancements, et cetera, to be able to get to utilise that data in an AI environment. But I think the smaller and mid-size ones go through it more frequently because it's not such a massive undertaking as it is for the larger multinational carriers.
Juan de Castro: Definitely. Johan, I've really enjoyed this chat. Hopefully, you have too. But thank you so much for joining me today.
Johan Slabbert: Again, Juan, thanks for having me. It was great to have a chat.
