Scouting for Growth
00:49:36 · Amrit Santhirasenan
[00:49:36] S2: Yeah. So maybe I made this joke last time, but I always say the robots will come for the actuaries first. So I say that as an actuary myself, I think, you know, it's been very interesting to me over time. You know, I was a London market actuary for twelve years. Right? I spent time out of the box with my team being shouted at by brokers at the coalface. And I will say this, right? I think this... What we're seeing, and we're seeing it inside technology companies as well, we're seeing very clearly that technical skills, what do they call them, verifiable domains, right? Technical skills where there exists a solution, a closed form or at least, you know, definitive solution to a problem. These jobs are being taken and commoditized much more quickly than getting a broker to bring you some business, than networking, than convincing people to like you. Right? This is a hard thing and it is not straightforward. And I've seen some extremely good underwriters, you know, be amazing at this learning how to say no, right, while maintaining the relationship. You know, the thing I would say is I don't believe that there's a fundamental like epistemic or capacity based ceiling on the work that anyone has to do, including actuaries for the record. Right. I think we all I don't know anyone in my current job or my past job as an actuary who goes, do know what? I've just got a finite to do list and I wish I could get it through more quickly. So I do really believe that we're not gonna have replacement. I think all that happens... There will be transformation. Yeah. There will be transformation of roles. There will be elimination of old upsets of work, and that will have that will make things difficult. I'm not glossing over this. I think there will be some roles which don't exist in the future. But in terms of the aggregate capacity and requirements for every human inside the company, there's only going to be more work. You know, it's become a cliche, this Jevons paradox, which is the one that goes around. And so for our listeners who aren't so into the AI world, Jevons paradox refers to the way when you make a certain thing cheaper, right, the aggregate demand for it goes up and the amount of consumption of it goes up more than the offsetting reduction in price.
00:51:45 · Amrit Santhirasenan, Sabine VanderLinden, Sabine VanderLinden
You know, think we're going to see that very, very clearly in the insurance sector, not least because insurance the insurance gap is extremely large, Right? Like, you know, there's trillions of dollars. [00:51:57] S0: Yes. 1,800,000,000,000, Sam Hurts. It's not small. [00:52:00] S2: That's it. So $2,000,000,000,000 of cover that's yet to be borne. We've got plenty to do. And you know, the idea that we can go after that. Guess will we do it with the same skills we had before? No, no way. But can we now have more capacity to go and do it? And what is that? What does that mean for society, for global growth, for global risk taking, right, for our sector, for our market? It's only positive in the aggregate. It will be tough for some people. There's no doubt about that. I'm not dismissing that. [00:52:24] S1: Yeah. Absolutely. But at the same time, how do we address some of the biggest organizational barriers? Right? We still have a lot [00:52:32] S0: of legacy systems. We still have a lot of cultural resistance. [00:52:36] S1: What have you seen in those organizations you are working with that they have been able to do better potentially than the others that need still to go through that transformation?
00:52:47 · Amrit Santhirasenan
[00:52:47] S2: Yeah. It's an excellent question. The one I'm thinking about hugely, because I am the AI maximalist in our company, and, you know, I talk to our investors and they go, my goodness, like you are one of the most obsessed founders about this in our whole portfolio. Right? And I think believe this is so important. So I'll say a couple of things. I think the first thing is to start with leadership. I'm a massive believer in this. Anyone who's worked for or with Hx will know that like I judge people by what they do, not what they say. So I spend my weekends building n8n pipelines, automated pipelines. I'm a massive fan of Windsurf, Cursor, Claude Code. Right? Every possible tool you can think of, I use it myself. My chief of staff and I on Sunday mornings at 06:30, we review the evals that we've built for our QBR pack together and we test and learn and tune them together. And I'm like, no one does X faster or better than the boss, right? I'm a big believer of that. No one turns up to work earlier. No one works harder. No one is more AI native. So I think it has to start with leadership. I think that's the most important thing, right? And like some of the most incredible companies that we've had the fortune to work with, you know, Convex, you know, we've worked with for many years. I remember talking with Paul Brand about how he'd set up the Expensify system himself. Right? Sorry, Paul. If people watch this, they don't know about this anecdote. Right? Paul is technical. He he he gets it. And, you know, when you have someone like that, that that trickles across the company. So that's the first thing I would say. The second thing I would say is you have to... And people have heard me say this. I say this pretty much every single presentation when people ask me this question. You have to optimize for change, not for results. But that's really hard for people to digest. Right? Really hard for people to digest, particularly very large enterprises like insurance companies. It's much more important to optimize to be able to deliver chunks of delivery, right? Chunks of value iterate it and change it than it is to set yourself some extrapolated over a lossily long timeline dreamy project. And I think that's something that is really hard. I find this hard for the record. Right? Like, these visions of what an incredible AI native system can look like.
00:54:55 · Amrit Santhirasenan, Sabine VanderLinden
I don't even know what the capabilities in GPT-6 are going to be yet. Right? I'll find out earlier than the average person because one of our investors is an investor with OpenAI, so I'll get a bit of an insight into it. But who knows, right, what that's gonna look like yet. So you have to optimize for change. You have to create a culture where people are more comfortable to experiment and throw away. Then I'd say the last thing, and I run the... Company this way, is you have to measure it. You cannot optimize what you do not measure. Right? So if you want to drive this change, right, you have to measure it. You have to pay it back to people. You have to show people, hold up a mirror and say, look, these are the things that we can do. And, you know, I'll say this, but I see this inside the transformations that we do. Once you can show some success and you can show value, it sorts itself out. [00:55:44] S1: Yeah. No. Absolutely. I have no doubt about it. So I would love for you now to give just a few recommendations, right? We are talking about insurance carriers. They are going through their AI journey. So when you start thinking about piloting agentic underwriting, where should they start? What would be your recommendation for them out there? Some of them will meet with you just... In just, I guess, a few weeks at at least Vegas.
00:56:12 · Amrit Santhirasenan
[00:56:12] S2: Yeah. Great question. So apart from come to us for all your agentic needs, which I've been not doing a due... My duty as founder say that. I'll say, you know, where do you start when it comes to agentic underwriting? I'm a huge believer that you start with the jobs to be done. Right? And I think this is an extremely powerful framework. Right? It's essential in technology, but it can... It applies incredibly deeply in insurance, right? Like there is a low hanging fruit workflow in every single underwriting team in the world that is absolutely automatable today. I'm willing to say I'm an actuary, so I rarely assign 100% or 0% probabilities to anything. But I'm going to go out there and say, I'm confident that 100% of teams, one workflow that is absolutely addressable by an agentic orchestrated end to end pipeline. If we even had 5% of the market realize those, right, 1% of the market realize those in the next year, that would be an incredible uplift. We're not aiming for 1%, right? My goal is for every single one of our clients to be a daily active or at least monthly active, weekly active users of our agents by the end of next year. Right. You know, even if a tiny subset of them were able to find one job to be done, that would be touch paper for their companies. Absolute touch paper.
00:57:37 · Sabine VanderLinden, Amrit Santhirasenan
[00:57:37] S1: And so when we start looking at the fact that, and you know, the research that the E1, Microsoft, McKinsey, all the top consultancy talk about that, the importance of modularity, flexibility, this flexible architecture, right, to avoid building tomorrow's legacy, how you make sure that everything you do is adaptable in the age of AI? [00:58:00] S2: Yeah, it's a hard question. And, you know, I think I'm a big believer that, I say this inside my company, but trying to avoid tomorrow's legacy is a recipe to get nothing done. Right? And so I would say pick the areas where they are reversible and they're testable. I probably wouldn't be looking to replace... Rip out and replace my core ledger system with an AI based system today. I probably wouldn't be doing that. Maybe there are some, maybe if I was selling that software, I wouldn't be saying this. But I wouldn't be looking to rip out like that. Those are systems that are fundamental and you get them wrong, you get some hallucinations there, you're in real trouble. But I would say to avoid tomorrow's legacy, pick an area where you can afford to have some risk. And I say this to our clients, like how many of your spreadsheets are perfect? Yeah, how many of them really are completely error free that result in a really wonderful user experience that aren't fragmented and messed up? So don't look tomorrow to fix all of those problems. I would say understand two way doors and the one way doors, and go through the two way doors as quickly as you can. Be careful of the one way doors, for sure.