This a shortened version of Making Risk Flow podcast, episode: “The Future of Underwriting Has Arrived: Examples from the Front Line | Jamie Wilson, Chris Varley and Salman Siddiqui". In this episode of Making Risk Flow, host Juan de Castro sits down with three friends in the underwriting and insurance space to discuss the importance of underwriters in the insurance industry and how technology can improve their efficiency and effectiveness. They mention the need to digitise data, automate processes, and provide decision support to underwriters. The importance of a customer-centric approach and the role of external data in understanding and evaluating risks is also on the table.
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. Recently, I had the pleasure of speaking at the panel titled ‘The Future of Underwriting Has Arrived: Examples from the Frontline’, during an event organised by an incredible team at InsTech and sponsored by Cytora. Together with three other fantastic panellists from Allianz, Hyperexponential and Moody's Analytics, as well as Matthew Grant, our moderator, we talked about the advancements in underwriting currently implemented by insurers. As you will hear in the episode, we covered a broad range of topics, from underwriting feedback loops to pricing transparency, decision support for underwriters, digital risk flows, and many other topics around data. And we often hear that the insurance industry is slow moving, but judging from the presentations, we are now at the most exciting time. And many carriers are undergoing successful digital transformations right before our eyes. I hope you enjoy this panel and the topics raised resonate with you and what's happening in your organisation.
Matthew Grant: Excellent. Right, we are going to ask each of you to do a brief introduction, please. You've all got really interesting backgrounds that are relevant to this. So, Chris, do you want to kick us off with your background?
Chris Varley: So I'm Chris Varley, I'm Chief Data Analytics Officer for Allianz UK. I've been with the company for about 18 months. Prior to that, 20-year career in the insurance industry, mostly in pricing, both within a large similar insurer and also five years in consulting.
Matthew Grant: Thanks. Salman.
Salman Siddiqui: My name is Salman Siddiqui. I help Moody's evolve its proposition on a number of different topics, primarily on the use of different data to help add actionable insight into the underwriting process. Maybe just a quick aside, I think a lot of people in the audience will be familiar with Moody's as a rating agency, and that's understandable. I'm representing the other part of Moody's, which is Moody's Analytics, which is a data and analytics company partnering with insurance companies. You might know some of our acquisitions like RMS or Bureau van Dijk. Just wanted to put that out there a little bit.
Matthew Grant: Right, and Juan, perhaps you could introduce yourself as well.
Juan de Castro: So, I'm Juan. I'm the COO at Cytora. So I oversee our product and engineering, customer deployments and sales teams. Joined Cytora three years ago from Hiscox, where I was their COO in their UK business. And before that, my background is in technology and management consulting.
Matthew Grant: Excellent. Jamie, finally, you.
Jamie Wilson: Yeah, so Jamie Wilson, Head of Pricing and Innovation, Hyperexponential. So I'm a pricing actuary by background. Started my career in consultancy, moved in-house in the London market, working in specialty commercial markets in the UK, US. Heading up pricing teams, heading up a predictive analytics team, and generally loving data and transformation.
Matthew Grant: Great. So we get two insurers or three insurers for a price of one tonight. Juan, want to kick off with you first of all. So there's been a lot of investment in analytics for improving underwriting over the last 10 years. What's happened to the money that's been spent on that? Has that improved the efficiency and effectiveness of what underwriters are doing?
Juan de Castro: Yeah, so probably just to be polite, I would say probably mixed results. I think that is the way of saying probably a lot of it has not driven much value. I think there's been two phases in the last 10 years. So 10 years ago until probably four or five years ago, a lot of investment was pin-point solutions. So it was about the policy admin system re-platforming or things like that, which ultimately drove really little benefits to underwriters. It didn't change their life much. And actually, the business value was reasonably limited. In the last five years, the focus has completely shifted to underwriting workflows. So platforms that really drive the end-to-end or not just a single platform, but the components that create a completely a more digitised end-to-end workflow and really drive a better experience for underwriters. And I think that is what we are seeing in the last few years, a real uplift in productivity and better reselection and in better broker service.
Matthew Grant: Yes, it's really interesting. So it's gone from being like a technology-led solution to, I guess, a workflow and underwriter-led solution. And is that why you talk about this next generation operating model?
Juan de Castro: Exactly. It's been more focused on how we make the life of the underwriter or the experience of the underwriter more effective, which is ultimately what drives the value. I talk a lot about, or we at Cytora talk a lot about making commercial lines a flow business, which is actually very aligned with Allianz's vision and strategy. This is something which is not new, right? If you think about it, this is a concept that was deployed in manufacturing companies a hundred years ago, moving from workers having to touch every item of work to really creating processes and workflows where items feel through the factory and human intervention is only deployed where required. And this is the same concept in underwriting. It's just taking us a hundred more years to get there, right? But it is about how can risks flow through an insurer. And humans will always, especially for mid market and large, you always require underwriters. But you only deploy that capacity on decision-radium risks that contain all the information and underwrite risks.
Matthew Grant: Okay, well, it might have taken 100 years to get here, but the theme today is that the future is here. So hopefully now we've got here, we're going to be moving a bit faster going forward. Jamie, talk a little bit about Hyperexponential and how does that fit into this overall flow that sort of underlines the importance of the partnership that we were talking about earlier.
Jamie Wilson: I think, kind of, we've got a very, very similar vision with Cytora. Hyperexponential, this love of kind of automating things, automating data ingestion, reducing rekeying and thinking about, okay, well, how do we improve the underwriter’s life? So kind of we throw this idea around that Hyperexponential quite a lot of pricing decision intelligence. A lot of pricing platforms are out there and they're there purely to try and produce a number. I'm going to get a spreadsheet, I'm going to put on this platform, it's going to kick out a technical price, the underwriter is going to move on. And I think Hyperexponential, we think about, okay, well, how do we automate as much as possible? But then how do we move pricing to higher orders of insight rather than just producing a technical price? Can my pricing platform, can it tell me about my portfolio? Can it benchmark this risk against my portfolio? Can it ingest more external data? Those types of things so that effectively the underwriter, yes, they're given a technical price and that hopefully that technical price improves over time, but they're also given significant amounts of other real insights about that individual risk, about its impact on the portfolio. And I think that's what we really drive towards Hyperexponential with our platform and it's partnerships like Cytora that kind of make for like really exciting opportunities there as well.
Matthew Grant: Well, exciting and also multipliers. I was talking to Amrit, your CEO and founder, our co-founder, and his great concept, which was if you think that technology makes a change and that gets people to change, it needs to be about 10 times the improvement. If you take two organisations that are building technology at 10 times the improvement is a multiplier. So the hope is that Cytora plus Hyperexponential is 100 times. I guess you had in Moody's, so I'm going to maybe get to a 1,000 times if the maths is right. Anyway, Chris, let's talk a little bit about the world of underwriting. And I guess just want to pick up on Juan’s point earlier, in your experience, do you feel that our underwriters generally, you could talk a little bit about Allianz specifically, but are they moving fast enough and embracing some of these changes now, having sort of seen a difference in the way the world is coming out with the more underwriting-led and workflow-led situations, as opposed to having technology thrust upon them?
Chris Varley: I think that the whole point around focusing underwriters' energies on where they add the most value, I think is something that we can all kind of agree on. I don't think there's anybody out there who doesn't feel that underwriters should be exercising their professional judgments and building relationships in the market rather than processing anything. I think there's a high degree of enthusiasm for that. But when you start to think about that more through that sort of decision support type lens, the decision support kind of starts before it gets anywhere near an underwriter in the first place. You're thinking about automated underwriting, decision making, rule-based intervention. Juan mentioned there about that sort of conveyor belt concept. Well, if we're going to have automated rules that push things off that conveyor belt and land them with an underwriter, the decision support sort of starts there, really. Why have you sent it to me as an underwriter? What rules have been triggered? What is it about this risk that means that we need an underwriter to be involved in it? Has the exposure changed? Is it the claims history? What is it that's driving that?
Matthew Grant: And so, let’s talk a bit about that because I think again, back to this point about disruption, we're a long way off being able to get rid of the underwriter. So the underwriter has a role to play. They've got access to data analytics. If you're looking at hiring great underwriters or coaching people to be great underwriters, how do you develop them or hire them so that they're actually accessing data that adds an additional level of value from what can be automated?
Chris Varley: It's partly around the hiring decision in terms of people who feel comfortable and competent working with data and analytics. But at the same time, what you're trying to do with decision support analytics is you're trying to support the underwriter in the decision that they're trying to make, and in the conversation that they're having. They're trying to understand the risk, understand the customer, understand its exposure, and be able to converse with the broker in a way that resonates with the broker in terms of the conversation that's taking place. If you're able to provide insights that help the underwriter to make the decision that they know they need to make, in a lot of ways, much like when we all interact with data and analytics in our day to day lives, you know, we don't have to be data experts if the analytics are presented in a way which is coherent and makes sense based on the thing that you're trying to actually do, if that makes sense.
Matthew Grant: Okay, well, it's good to hear that they can have coherent discussions with the brokers as well. That's definitely something that's improving. Salman, that's a nice link into what you're doing with Moody's. So you are providing data, you're accessing other third-party data. What's been your experience about how you're helping insurance underwriters access this information? And I just kind of reinforce the point to Chris, which goes beyond what can be automated, but how does an underwriter add value by actually having to work with sometimes what can be quite messy data to make decisions as part of the underwriting process?
Salman Siddiqui: Yeah, I think this idea of messy data or disparate data, we do come across this a lot. And I think Juan mentioned this idea of rekeying data many, many times over. So if you take a step back and we think about the way insurance companies capture data at the moment, they're very good at capturing what has been insured and where it's been insured. Not so good at capturing who has been insured. If I look at policy admin systems across the market, the data within them is quite patchy of who's actually been insured, what sort of business are they doing, what sort of sector are they in. This customer lens is kind of missing a little bit. And the role of external data is multifaceted. In one situation, we can think about cleaning up that data at the point of entry. So not having to rekey that risk multiple times over, at the point of entry, you're using a validated, verified, curated data set to help validate the data at the point of entry effectively. And what that basically means is from there, you don't have to rekey that five times over. You've got a clean record. That record becomes your golden record for all things to do with that entity. So from that, you can unlock a number of different things, right? You have a consistent way of looking at your corporate exposures because historically, insurance companies would look at their exposures from a physical risk perspective and say, this is where they're located. This is the NACAD exposure of my portfolio. But you can also look at this from a corporate perspective. You may have insured Wilkinson's, for example, five times over. And not realise it, actually, right? You may have insured five different subsidiaries of Wilkinson's and again, not realise it. That corporate level data allows you to look at those exposures at a corporate level. And that leads into automation, that leads into pricing and the augmentation of external data alongside traditional risk metrics. So just talking to Jamie earlier today, you have your traditional risk metrics, which guide you towards what the right technical price is, which can be augmented by non-traditional data that are also a predictor of loss. One of the clients that we work with, for example, they use our credit ratings because their view is that a strong credit rating is a well-governed company that has good risk management, good controls in place and therefore should have a lower propensity to claim. That's one way of thinking about non-traditional risk metrics to augment the traditional risk metrics. Cyber scores, they're not just limited to cyber insurance. Cyber risk scores can predict or can be a good way of thinking about the propensity to claim on D&O insurance, on credit insurance and financial lines, other lines of business which are not directly linked to cyber, as you would think about it. But again, ESG is the other example.
Matthew Grant: Juan, I just wanted to come back to you for a second. You talk about decision ready risks for the underwriter, which to my mind sounds like what I do on late at night. I need to get some food to go and put in the oven. Is that a good analogy? My ready meal as healthy as a ready meal or is it how do you describe it?
Juan de Castro: Could be a good analogy. It actually builds on what the three other panellists were talking about. Decision-ready risks are defined by almost three factors. One is it contains digitised information. So it contains all the information required to evaluate the risk ready to be looked at. The second one, it's been evaluated based on the distribution strategy, the underwriting strategy. And the third one is it's converted from pure data insight into that decision support that Chris was talking about. And that is how actually you drive an acceleration in the underwriting workflow is you provide the underwriter with the insights. So by the time the risk gets to the underwriter, it's been evaluated. There are these three red flags. Please look at these three things. This is how it compares to similar risks. It's actually helping the underwriter make a decision.
Matthew Grant: Okay. Well, I'm going to come back to your ready meal in a moment because I want to just understand that whole workflow from the point of which it comes into the insurer and then goes through and works on different people. But Salman, I want to come back to a point you made, which I could completely agree with, we spend quite a lot of time talking about insurance and brokers and occasionally we forget who the real client is. And one of the comments you've made, and I think possibly given your coming at this from outside of insurance is like, what about the customer? So can you just expand on that a little bit by what do you mean? And what does that mean in terms of how underwriters can get better at underwriting?
Salman Siddiqui: Yeah, so I think this is the relationship management part of underwriting. There's a risk management part of it. There's a relationship management part of it. And when I look at other industries, they've moved very quickly from a product-centric approach to a customer-centric approach. Insurance is catching up. But I think there's a consistent theme of insurance catching up with other industries here. Where insurance companies have historically looked at their portfolio and said, oh, I sell D&O, I sell property, I sell casualty, to saying, actually, I underwrite the retail sector. I underwrite manufacturing or I underwrite telecoms. These are sectors that you sell into. These are clients that you sell into. And kind of having the data that you need, building a customer 360, understanding the client's needs better versus sending a product, there's a nuance to that. In that situation, the external data plays a very big role in helping you understand where the customer is and where they're going, it allows you to better understand what they need and deliver the right sort of insurance products to them and then partner with them, right? Insurance is not just about risk transfer, it is about risk partnership at the end of the day. And this customer lens is absolutely essential to go back to adding value in the value chain if I think from that perspective.
Matthew Grant: So Chris, just on that, Salman said we need to look at the customer as an entire entity. You've got a background in insurance, I might come to you Jamie and ask you about your experience. I mean, it sounds easy to say, but in practice for a large global insurance organisation, how, I wouldn't say easy, but is there a journey towards being able to get to that single point of view or your golden view of the client?
Chris Varley: Firstly, I think it's a great point you make. I think the whole sort of orientation of the industry around product and line of business is one that's very well ingrained. And so breaking that is difficult. I think the examples I would draw on help answer your question, very often insurers will have products that are defined in a relatively rigid way, perhaps that have commonalities amongst them, perhaps in terms of property exposures existing within an industry segment or within a property focused account or a real estate account or whatever that happens to be. And ultimately the ability to be able to provide decision support to an underwriter and to be able to harness the power of all that data, kind of relies on some common data foundations. Do you define some insured in the same way? Do you define your industry segmentation in the same way? Because that's kind of the fundamental piece to being able to not just be able to draw on the wider experience of just that product, but also every other property in your cumulative experience either across your country or territory or even the whole world. And I think ultimately that is a very difficult thing to unpick, but at the same time, what we're doing is we're providing coverage for well-defined entities, whether they be properties or they be vehicles or they be employees. These are relatively well understood concepts in general society. And so the insurance industry can make a bit of a meal out of that. And I think that ultimately you have to go back to first principles, what is it that we're insuring and who's the customer that we're serving? Which I think you made the great point very well.
Matthew Grant: So Juan I wanna come to you and then Jamie, I want to learn a bit more about Hyperexponential fits into this. So when you look at what you're offering and improving this flow and what bit of the pie you're trying to solve for? You're focused on commercial risks I know. And then to Salman’s point, how broad can you go in trying to link up different exposure? Because that must be an important part for an underwriter when a risk is coming in is who else is that organisation looking at.
Juan de Castro: You're right. So we focus on for the time being exclusively on commercial insurance. Think about like everything that needs to happen before an underwriter starts making a pricing or an underwriting decision. And if you think about it, what do you need to do? You need to digitise the information that comes from the broker, often in emails with attachments, et cetera. You need to understand who the client is. So stop working with a client name as a string that says Starbucks LTD, or whatever, and start working with unique IDs that allow you to match subsidiaries, et cetera. You need to start pulling all the external data. And often when we talk about the external data, people think about like, this is like innovative data sets, quite often it's just the information underwriters are using today, but they're just manually pulling. That's what we do. And some of our competitors do, pull external data. And then one of the most complex steps is then how do you make sense of that? So, often the broker will have provided a business description, but the Allianz system expects a NAICs code. You start then having to map and do some inference from the old information you have collected into what insurers are expecting. This is what I refer to as the digitisation step. Only when you've done that, then you can start saying, okay, let's evaluate. Is it a vanilla risk that could be straight to process? Is it a complex one? Who should it go to? Is it within appetite? Is it a high priority risk because it's in a segment where we want to grow? Does it have anything special that requires special attention? That's when you get into the evaluation and decision support.
Matthew Grant: Great. So, Jamie, so Cytora has brought it in. They've done one of the difficult things, which is tidying it all up. Sounds like some of it's still a bit manual, but they're helping the underwriters to get more efficient, helping Chris's underwriters get more efficient. What does Hyperexponential do now?
Jamie Wilson: Yeah. So if you're thinking about that partnership, so kind of as Juan was saying, risk comes in, email, whatever it is, whatever distribution channel. It gets digitised by Cytora. You know, the underwriter hasn't touched it at that point. Risk appetite rules applied. Underwriters haven't touched at that point. They're then presented with, you know, here are all the submissions that I could be looking at. They effectively from the Cytora platform can then say take me to pricing, you know, they've made their decision, I want to proceed on this risk, this is the one that I think I can add value on by quoting. Take me to pricing, click that button, opens up the HX renew front end, effectively that kind of the pricing, the heart of the pricing engine. And all of that data has already been pre-populated. So, you know, you've instantly removed all of that rekeying. So underwriters not entering any details about the properties, not entering any details about the policy. And at this point, all they've done is click a couple of buttons, you know, it's a couple of minutes in the pricing platforms pre-populated. What we then do, again additional external data enrichment thinking, it's a classic thing, you know, in pricing. What data can I bring in to augment my pricing decision? Click another button. How does this compare to the other risks I've priced this month or the rest of my portfolio? What's the marginal impact to my portfolio? The underwriter effectively at that point is getting all of that rich detail about the risk to understand, okay, what price point am I going to put on this? What is the optimum insurance structure to optimise the risk transfer and profitability? And so within that, we've, in the demo workflow that Juan and I have been working on, have then added in sanction screening at the click of a button in HX renewed because why not? Have it in one place. Don't make the underwriter go somewhere else. All that data is there. Click a button, get through your sanction screening. Embedded peer review. Click another button. Peer review is done. Then finally, the underwriter has got to the end of that process. They click, finally, quote. Quote document. Generated, like all pre-populated, and then goes back to Cytora. And I love it because for me, I call it the 10-click underwriting process. That's what underwriting in my mind should be. It shouldn't be, okay, I'm now going to this website. I'm now going to that website. I'm now going to pull down this data. I'm now going to go through this. Like, put it in one place, automate it, give the underwriters everything they need to understand the risk properly and let them, rather than clicking and going around and typing, spend time digesting the risk, pricing it appropriately and making the decision to move forward.
Matthew Grant: Oh, sorry, Jamie, that's a great link to the next question for Chris. So you've been going through a transformation program at Allianz. Does that concept of, that Jamie talked about the 10 clicks for the underwriter per submission I suppose we would assume it is, is that sort of the tiny thing you're thinking about as that transformation program and how is it going generally?
Chris Varley: Absolutely. I mean, I think there's a lot of focus around the sort of efficiency aspect of that. And then, you know, obviously absolutely can subscribe to that vision if you like. The important bit in the middle of all of that is to take a lot of the process away from the underwriter so that they can focus that professional judgement on where they have that experience. And that is all about that decision support aspect for me in terms of specifically, I mean, Jamie mentioned about the sort of pricing angle of it, which I think is absolutely critical. For me, pricing transparency is decision support 101. If you can't present a risk to an underwriter in a way that says, this is the price that we've attached to this risk and giving transparency in terms of which coverages have which price attached to them and why, how that breaks up between the exposure pricing and the experience pricing, what the expense allocation looks like, what's in there for CAT exposure, how that interacts with your reinsurance, that level of transparency is critical. But then also the level of transparency that's also about why. It's all well and good to say, this is the parameters that we put in this model, but why? Where's that come from? What's the collective experience that's led to that? How are you sort of almost harnessing the collective knowledge of your entire underwriting organisation and presenting it to an underwriter in a way that they can understand? That's kind of the foundation of it for me.
Matthew Grant: And that transparency goes two ways because the underwriter wants transparency in what's coming in, but they're also going to be assessed in order to some extent on the transparency of what they've done. So presumably you want to be able to go back and find out why the underwriter made a decision.
Chris Varley: But for me, that's a really positive thing because that's a learning loop. One of the really key ways that you can build credibility with underwriters in a pricing function, particularly at the corporate and specialty end of the market, is by being transparent about the fact that we don't have absolute certainty around the exact technical price for that. So if you're talking about a three figure average premium SME risk, yeah, okay, there's a lot of confidence around that. You can actually communicate that confidence in the technical price of the underwriter. But when you don't have that confidence and you're giving them decision support about similar risks in the portfolio and the coverage options and the deductibles and all of that kind of stuff, actually what your underwriter is then doing, the professional judgement they're applying, it's not just about capturing what they did from a kind of audit point of view. Did they do the right thing? It's about learning from them. What did they do and how does that mean that we can learn from that and then implement that to help improve our technical price going forward? I think that kind of transparency works in both directions. It's kind of about that honesty that says that some of these problems are really, really difficult to solve and you have to work very much in partnership and share information if you're going to get better.
Matthew Grant: Yeah, and that's why we still need human underwriters.
Chris Varley: The most complex risks, absolutely.
Matthew Grant: Yeah. Salman I want to come back to the point you make, because I think everyone's talked about this to some extent, that you mentioned about the golden record and the benefits of being able to connect across different lines of business. Is that something you're doing just now? Because I've had a number of people asking me personally, what systems are out there to be able to pull in across different lines of business? So are you there with Moody's or is that still part of the development journey you're on?
Salman Siddiqui: Part of it, we're there. Part of it is part of the development journey. In terms of the consistency of data, we're kind of mostly there. In terms of what systems and clouds and what we use, we're somewhat agnostic to that at the moment. But the idea really is that single data set can then be enriched with different data assets, either from Moody's or some external provider or your internal data to collectively create what we call the secret sauce that makes you different from somebody else or the guy down the road, et cetera. Within that data can be used, I think Jamie touched on, is that data can be used for section screening, right? It's the same data set. Right now, what we're seeing is sanctions, compliance teams are using their own data, underwriting teams are using their own data, the risk teams are using their own data, but really trying to get that coherence, call it a risk data lake or a risk lake sometimes. Having validated data in there and single golden ID allows you to do all these different functions a lot more efficiently. That's the way we think about how insurance companies are moving forward; a long way to go though.
Matthew Grant: Not 100 years, but some way to go. Juan, that's going to bring us close towards the end. I feel based on what we said, that we can justify the title from this session, the future of underwriting has arrived. It probably hasn't all arrived quite where we wanted to be and not everyone's got there. What have we missed or what's happening next? As you think about the future of the underwriting being here and available today.
Juan de Castro: Just some context, right? I think the reason we put together this tonight session and the title is, I think we are probably all a bit tired of just hearing talks about the vision and the theory of how things can be done. We wanted to just bring real examples and actually, we've got a number of sessions after this panel with what are insurers doing today. So Craig in the second row will talk about what are they doing at Inigo, right? And then we've got a number of clients here. That's the reason for the title and the spirit of this event. I think we've seen a huge shift to embrace this concept of a flow business of supporting underwriters with decision support, and how to embed better insights in the underwriting workflows in the last few years. So I think what's going to happen in the next few years, obviously, LLMs and GenAI are, I don't think they're changing any of our visions, but they are massively accelerating. Both from an insurer's perspective and from a provider's perspective. Probably we've made, in the last 12 months with LLMs, we've made three-fold as much progress as we thought we would have. So I think it's becoming more real faster than probably we all thought.
Matthew Grant: And LLMs, of course, large language models, part of the genitive AI, we're going to be seeing an example from INEOS shortly, just seeing how that works in practice. Juan, thank you very much for bringing this together. Jamie, Salman, Chris, really great to get all the insights. I feel we got through an awful lot in a very short space of time. So thank you all very much.
Juan de Castro: Thank you. 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 podcasts so you never miss an episode. To find out more about Cytora, visit cytora.com. Thanks for joining me. See you next time.
