00:16:10 · Amrit Santhirasenan
[00:16:10] S2: Yeah. You know, we're always looking to close those gaps because I think if you look at the models in the world, been built over the last twenty years, they just haven't changed very much. Right? You know, notwithstanding the fantastic advancements in the calibration rate making benchmarking space of actuarial work, The GLM side of things and obviously more advanced machine learning models now. GLMs have been wonderful for twenty years, but they're not state of the art anymore. You know, there's so much more that can be done now in the analytical world. And so for me, I think when I think about this question, I think it's like actually less important about the model is what information can we surface? What information can we surface to the people making the decisions that they just don't have right now? Your example is such a good one, right? Because, you know, you've all you will have always known that that data is out there, but it just wasn't practical for you to read 100 reports. And the same thing applies for every single risk that... Pretty much every single risk that an underwriter looks at today. Right? There's a world of third party data. For the biggest ones, if you're insuring the international group or you're insuring Zencouran, the biggest risks in the world. Yeah, you know what, it's extremely worthwhile for the actuaries and underwriting assistants and the underwriters to go and do as much research as they can to get that information. But you can't do that on every single risk, right? You can't do that. Whereas... So now I think, know, to answer the question about the gaps that the models have. It's not necessarily even about the models. It's about like, what's the information that you put into them? I think our models actually, the for the most part, you know, it's not too difficult to design a model, you know, that actually is very capable, but for the data that it doesn't have. And plugging that gap is just just a key part of what we do. You know, we talk about data, we talk about insights, you know, so the mixture of facts, the mixture of judgments and then the decisions that come out of that. Just the amount of data and insights that you can put into a model before you make decisions. So it should be a hundredfold.
00:18:06 · Sabine VanderLinden
[00:18:06] S1: It should be a hundredfold. And the quality of data is actually critical. But one thing which is happening, and as you are speaking, I mean, thinking about my own experience as a business owner, right? I go to loads to get my business insurance as we do in the world we are in. And, you know, once upon a time, I remember it would take maybe two weeks to get a quote back. And today, you know, in the summer, my broker reach out and within, I would say twenty four hours, in two hours, it's Sabine here, it's a new quote. Are you happy with it? And it is done, right? So fast quote to buying cycles are critical for us to differentiate. Probably when you have good risk, the one you are seeing here. What competitive, I would say strategies, how does agentic AI start helping carrier meet the needs of their brokers so that they can continuously differentiate and actually grow as well alongside the carriers AI survey?
00:19:08 · Amrit Santhirasenan
[00:19:08] S2: Yeah. Yeah. It's a great question. And, you know, it's been so interesting because actually, ironically, the last twenty years, the focus has been on what's been our bread and butter, right, the pricing and underwriting models and algorithms. Actually, what we haven't been able to do is do the very quick prioritization, the triage, right, the analysis of the fundamentals of a risk to work out whether you can say yes or no, or maybe quickly. I think one of the things that agentic capabilities have brought is the ability to do this, right? It's the ability to dramatically increase the quality of service, because for a very large amount of risks, you know quickly, right, whether you're in or out. Then the ones in the middle are the ones that takes the time and effort, right? The ones where you're not sure. I think we're seeing, know, we have a triage offering, right, that's being added to our suite of capabilities. And what we've seen really clearly is that using AI now, AI has unlocked particularly even, you know, before it was agentic. Right? But the generative AI models have unlocked the ability to pull data so quickly out of the information required for underwriting that you can put a very quick red, amber, green status on these risks. You know, again, several orders of magnitude faster than ever before. You know, I think when I talk to our clients, absolutely, they are obsessed about profitability, particularly in today's market. And that's, I think, a differentiating factor of the next generation technology providers will be not just focused on the administrative side, but the actual optimisation of profit, which is something we think about very deeply. But they also care about service. So I think what we're seeing right now, and your example is a very pertinent one, is that the technology hasn't been able around to actually provide that service feedback loop as quickly as we can see. And, you know, it's amazing, isn't it? You were saying two weeks and now a day, you know, I think we've got even in the specialty space, we'll get even faster, right? Think it'll be hours soon. Yeah, right. Mean, cyber
00:21:03 · Sabine VanderLinden
[00:21:03] S1: and all those new risks that we are seeing, right? Know, SME, cyber, I think the technology is there. And so now it's about the underwriter feeling comfortable to have enough data to actually win the risk. They are comfortable to win faster. And I think the term faster is a critical one right now. So another piece of research I was reading while getting ready for our conversation today, Amrit, is the multi agent ecosystem. So McKinsey has put some research out describing multi agent systems as those systems being able to handle intake, risk profile, pricing, and compliance together. Can you describe how the Pypex Financial or Hx sees these agents working together and where the human underwriter comes into play?
00:21:58 · Amrit Santhirasenan
[00:21:58] S2: Yeah. Yeah. And, you know, I think the analogy of, you know, the human jobs to be done is the framework that we think about very deeply, right, when we're trying to build this technology for our customers. And while step changes are absolutely possible, the first place you start is on what's happening today. So, you know, when you look at our technology stack, we have an ingestion agent, we have an actuarial agent, an we underwriting agent. And in the same way that those jobs to be done were done by different humans in the past, they're done by different agents with different skills today. So, know, it's not too different in my mind from the way work was done before. Collaboratives, you know, we use this fancy phrase swarm now, right? We don't talk about humans in swarms, but we talk about agents in swarms, right? A large group of different people, digital people, digital or otherwise, with skill sets collaborating very deeply is how you get the best results. You know, Hx has always been really obsessed about this feedback loop between the underwriter and actuary. And, you know, my time in insurance for the last twenty years doing it myself, I could see where the feedback loop was very tight between the underwriter and actuary and got incredible results. And now we're able to replicate that with a group of agents that work 20 fourseven. They don't get tired. They don't lose their patience. Their performance doesn't degrade over time as long as you've got enough power for them. I think our take on this is very clearly, we have our infrastructure, the core Renew platform, the Hx platform, which is the framework that people build, the decision making tools that they've been doing for the last seven years in Hx. But that just now forms a substrate, right? It's the kind of core information that all the agents who want to work on the different jobs to be done can refer to. So if you've built in a very powerful rater in Renew, our ingestion agent can say, Well, I know what the fields are that I need to fill out. So I'm going to go out with that information. I'm going go and capture the information. Our agent can say, Okay, this is great. This model has these type of parameters. So I need to go and actually do the analysis on the data sets to make sure I can fill out that I can calibrate those parameters.
00:24:04 · Amrit Santhirasenan, Sabine VanderLinden
And the underwriting agent can say, Okay, these are the most important factors in this model. So I'm going go and do some research on those factors. And so to me, it's like that human in the loop and the human in the loop side, I think sits really tightly across that because these agents aren't going to be able to do everything themselves. In the same way when I was a junior actuary, I had to go and get some help when I wasn't sure what to do. The humans are going to be there. So it's not too different from what we've done in the past just with a supercharged set of capabilities from these agents and the humans doing the appropriate guidance, judgment, conducting the orchestra, if you will. [00:24:42] S1: Well, you just mentioned something that I actually heard over the past three weeks, working with various conferences and hosting webinars, which affect not only claims, but also underwriting is the fact that we are now recruiting younger underwriters. And so we need to find a way to build those teams. So using the technology as a teammate to enable the more experienced underwriters to train easily the younger underwriter. What is your experience with this? Because our industry is suffering from a talent gap. And so new underwriters are coming to market. I've heard it from US to Europe. All insurers have told me that over the past three weeks. How do we leverage the technology to help us there?