InsTech - insurance & innovation
00:08:59 · Diya Sawhny
[00:08:59] S1: One example is surety. Right? And we think about credit risk and, again, we've been doing this for years. We have lots of information and expertise on large rated entities, but then also really deep expertise honed over years working with clients across many different financial services sectors and looking at small and medium sized businesses even. Models that we know are really good at assessing credit risk for whole swaths of not just the largest but small and medium size. So you combine that with the financial data that we were talking about earlier, the ownership data, litigation, regulatory, paints that full picture. So if you think about that surety line of business, we're starting to see some initial sort of really good uptake there. We had an opportunity with a US specialty insurer that was wanting to write more profitably without pulling back on growth. The book was made up of small and medium sized companies and the credit data there tends to be a little bit thinner and there's not always a public company type of signals that you're looking for. So we worked very closely with them. We looked at their North American book and we created sort of a test, if you will, against Moody's lost predictive risk signals, so combining that credit data, the formographic data, and company data into a single claim risk score that was calibrated against the insurer's own historic claims experience. And what we found was that removing the riskiest 10%, looking again using the claims experience that this particular company had, it actually reduced their loss ratio by over 50%. That's the kind of use case where we're really excited about bringing these broader capabilities that Moody's has but then making it very specific to the insurance industry.
00:11:04 · S2, Matthew Grant
[00:11:04] S2: Exponential risk in London returns on twenty third and 03/24/2027, bringing together the catastrophe modeling, exposure management, and wider insurance communities at the Mermaid in London. Across two days, we'll explore changing and interconnected risks, advances in AI, exposure intelligence, and how to turn better data and modeling into better decisions. We're delighted to have Moody's joining us as platinum sponsors of next year's conference, booked by November 6 to take advantage of early bird savings. Find out more at insect.co. [00:11:34] S0: And so now you've all these these incredible assets, different parts of the business, different parts of the client's business. How do you connect them all together? And I guess the clue is in the the concept you're talking about now, which is connected intelligence. But so how does that all work in practice?
00:11:49 · Diya Sawhny
[00:11:49] S1: This is something where we're thinking about and talking about a lot here at Moody's. There's a lot of data out there, but how do you make that more consumable, more relevant, more fit for purpose? That's what we're calling Moody's Connected Intelligence. So at the heart of it, you have a knowledge graph which is a way to organize all this information. It's not just a data store. It's connecting the entity with all of the attributes around it, right? That asset could be a company, it could be a property, it could be commercial property, it could be residential even in the future. Knowledge graphs have been around for a while, I think it's now these use cases are coming to life because we have the technology and AI enablement there which is allowing us to do this at scale. And we have tons of data on commercial properties, 90,000,000 commercial properties, things like valuation and tenancy and so on. So think about all of that data, the corporate data, the property data, physical risk, but then connecting it up. It's robust. It's curated. It's governed. Right? So you know you can trust that information when you're getting it from Moody's. I know there's lots of value propositions for lots of AI firms out there, but this is where we think we really win. Right? That data, to be decision useful, needs to be auditable. It needs to be verified. It needs to be up to date. So it needs to be governed, right, so that you can make use of it. So imagine where you wanted to use that data in an insurance context for Lloyd's regulatory purposes or for solvency too. Right? And the domain expertise, I think, is the other aspect of it. Right? Underwriting in an insurance context is related but different than underwriting for a bank. And that's where we think we have something unique to bring.
00:13:48 · Matthew Grant
[00:13:48] S0: Yeah. That point about regulation is so important, but also that point about AI as well. As you said, it's not enough just to be able to surface information. You've really got to be able to give confidence to the person that's using it and making different decisions. There's really good valid data behind that. Just going back to the knowledge graph. So is there any way you can sort of bring it to life?
00:14:04 · Diya Sawhny
[00:14:04] S1: Think about a scenario where every quote that begins with a slip, and you might get some information. So let's imagine it's ABC tools company on paper. You get some information. It's two entities. It's one supplier, four plants with exposure to two ports. Right? That's all the underwriter knows. Now imagine if you query that against Moody's knowledge graph. Right? What subsidiaries and the suppliers throw up in Moody's knowledge graph? And so the graph might come back with information. Again, you've queried those 600,000,000 entities and you go the next level deeper. And so, you know, while the slip only included two ports, right, we might find that this company has exposure to 11 ports. And then instead of the one supplier is actually suppliers that the underwriter suddenly finds himself, you know, looking at on that knowledge graph. And then where does that risk really concentrate? Right? What if out of those plans, five of them shipped through a single port. Right? A port with a history of hurricanes. That's where you can really hone in and query this sort of real time rather than historically, you know, that data would have been sitting in five different databases that you have to go search. And you have to know what you're looking for. Whereas here, you're getting that information served up to you in that initial instance and it's more dynamic. You can then decide. Right? Is this the type of risk that you wanna take? What's the appropriate level of pricing? And so on. And I think it could also create opportunity. Right? Things like supply chain risk, BI risk, etcetera. Things have been difficult to model with this kind of data. You're starting to surface information in different ways, making it more accessible and hopefully supporting the industry in terms of creating new products there. Another example, maybe it's a commercial property in some mid market in Florida. You have limited information on some of the tenants in the building. You're being given an opportunity to underwrite one slice of that building. Right? One tenant that you know about.
00:16:15 · Diya Sawhny, Matthew Grant
What if, you know, you had information on the five other tenants there and you had some information on the characteristics of the roof condition and the roof geometry rather than having to query different databases, you know, know exactly what you're looking for, having that information served up so that you can have more dynamic decision making information at your fingertips. [00:16:39] S0: You've referenced AI agents, of course, everyone is these days. What does that mean for Moody's?