00:30:44 · Amrit Santhirasenan, Sabine VanderLinden
We can read whole blocks of text and work out the intent of what that text means and then fill out related fields inside a Rater Even though it's not a one to one mapping, because we have all of that extra context. And it means we can do just much more rich, much more intricate things than you'd be able to do if you just said, I care about these five fields. That's the tiny subset of the information inside our system like ours. Is super. [00:31:11] S1: So Amrit, as you are actually talking about the detail of the system, another question that comes often to me is how do we ensure that everything we do is compliant by design? When we start looking at those agents and you actually mentioned all the capabilities an underwriting team, an actual team, a successful team needs to have when they start using agentic capability. Do we ensure we avoid bias? We avoid errors, right? In a world where a discrimination exists. Now we are going to systems helping us do the underwriting. So how do we avoid that, right? How do we ensure that the feedback loop is always there, always robust to make sure that we adapt at the same time to current market conditions?
00:32:08 · Amrit Santhirasenan
[00:32:08] S2: Yeah. It's a really good question and a really tough one. And I would, you know, the sort of thing that people like Sam Altman and Mark Zuckerberg are worrying very deeply about and that we think about hugely as well. So I'll say a couple of things. You know, we've always had this challenge. You know, we've been talking about these things for a really long time before the of rise of Gen AI. And I think what Gen AI has done is it scaled the ability and velocity with which we can do this work. It has also added much more obscurity into the mechanisms by which it works. There's a whole new emerging field in machine learning called mechanistic interpretability, which is actually being able to look inside these models and try and understand how they actually work, how the brains inside these gigantic matrix modification engines are actually working. And, you know, this field is incredibly new. There's loads of really fascinating podcast apps on this. You know, again, if you're ultra nerd, recommend listening to some of them. I think, though, that closer to home, there's two parts to this. This is the problem of user experience. Right? First and foremost, right? I think it's an incredibly... Like, we're very focused on this. It's incredibly important to us that the moment an agent expresses an opinion, it is unambiguously clear that the agent has done that and it's not human. And there needs to be an incredible degree of precision and auditability over every single judgment or decision or entry or any other interaction that an agent has with any system. Because the reality of the situation is that we are not going to solve in the same way that we were never going to solve bias perfectly before. We're not going to solve it perfectly now. But if we can audit it, we can understand it, then we can put the appropriate judgment feedback loops on. We can set systems up for spot checks. We can set systems up to hoist best decisions and monitor them, analyze them. There's this concept of adversarial modeling. You can have a separate judgment and bias agent looking over the the other agents, right, and actually supervising it. And I think we're going to see a whole ecosystem of this technology emerge that sits on top of the AIs that are just doing the work. I think some companies are very massive, very AI native, very AI first.
00:34:20 · Amrit Santhirasenan, Sabine VanderLinden
You know, was talking to one of our investors, Andreessen Horowitz, about the rise of the full stack AI native carrier. These are coming now and those companies will be actually probably a bit more permissive, little bit more risk seeking in their attitudes to use of the agents. And then we talk to some customers who are just like, no, no, no, no. Every time an agent does anything, I need to know about it. It needs to be logged and it needs to be auditable. So Our job is to make sure the framework exists, that people can choose their own adventure, they can choose their own degree of comfort, that we're able to service both ends of that spectrum. [00:34:56] S1: When you think about the underwriters, they want to spend more time delivering quality and they want to spend more time as well spending less time doing things which are mundane tasks. So how can customers using Hx redeploy their time better as underwriters to actually drive value for the business?
00:35:17 · Amrit Santhirasenan
[00:35:17] S2: Yeah. I mean, is the whole premise. It's a great question. It's the whole premise upon which our company exists, fits on, right, is that if you can shift the time from the mundane to the judge... To the differentiated, the impacts are very, very significant. Right? And, you know, you know, the studies from McKinsey talking about three to 6% improvements from loss ratio, you can take a conservative slice of that, which we do, and you still come up with a transformative impact to the companies. Bottom line. Right? So I think there's a couple of things there. I think this question of what do we do with the time we save is a cultural and kind of corporate, general corporate thing. You know, there's I always like thinking, you know, my job as CEO is to think about the company strategy. One of the key parts of company strategy is like, when you make an improvement to any system, where does the surplus accrue? Does the surplus accrue to the end users? Does it accrue to the company? Right? This is a really good example, right? Like, you know, if you free up the time for underwriters, you give them two hours back. What do they do? Do they spend more time on judgments? Do they spend more time on networking? They spend more time on TikTok? Right? Who knows what they will do? It'll be different for every individual. But I think one of the most important things that we do when we're thinking about the later stages of transformation with our customers is to understand what they want to do with that time. But sometimes that time will still be inside renewal, just be spent on other things, like maybe portfolio optimization, maybe more time on judgment, maybe more time reading the risk research and applying human in the loop. So my answer to that is a very typically actuarial answer. It depends. But it depends, I think, on the kind of the strategic direction that each client has with what they want their differentiator to be. The less cop out answer I would give you is that we see very clearly people wanted to spend more time on the risks that are borderline. Right? That's the thing that we want. The art of underwriting in a market that's turning is working out where to spend your time on, not necessarily on a per risk basis, but between the risks, right?
00:37:21 · Amrit Santhirasenan, Sabine VanderLinden
So going I have a much more high degree of confidence that this one I can definitely say yes to, and this one I can definitely say no to. But these, you know, they're 50% of the other risks where actually I'm not sure, I'm going spend much more time on the underwriting of those. And we think that just has a massive leverage, you know, over compounding renewal seasons, right? People spending more time on the risk that are the differentiators. [00:37:47] S1: Well, you are talking about renewal season and a lot of people were, I think, at the rendezvous just a few days ago. So how do you see dynamic risk assessment moving forward? Do you have examples as to where AI driven insight can actually improve decision making and what are your expectation again as we move to a world where agents are going to help us become better underwriters?
00:38:13 · Amrit Santhirasenan
[00:38:13] S2: Yes. Yeah. So I was a reinsurance pricing actuary for the first three years of my life. So this time of year is very familiar to me. I was embracing myself for very long days, long nights, weekends, evenings of underwriting and pricing work. So it brings back some nostalgia to think about that. So I'll say a couple of things here. I think, you know, what Hx doesn't do is pretend that it's like pulled the future out the universe. So I think we're at the very beginning of using AI for some of this portfolio optimization and insight generation. What I will say really clearly is it's no longer the case. When you started a renewal season in the past, you scrabbled around for absolutely ages trying to get all the risks from the last year, understand what your playbooks would be going into them. So I remember one of the renewal seasons I did myself, had 41 risks priced in like three weeks. It was excruciating. Was like, physics spending stuff, right? And I remember the time running up to pricing those, was like, I got 41 renewals. I need to go and get all the information from last year. And I need to at least have a set of opinions about which risks are going take my time, which risks I'm going to really want to be ready to fight for, which ones I'm not. Now, if you use a system like Renew, that's a database download. It's not running around copy off, copy off file, do not use, which folder is it, going to IT if there's an IT migration saying, sorry, can you give me access to the hard drive again? So I would say we are substantially better prepared in the era that we are right now than we've ever been. As for AI, well, what rich fodder that data is to ask an AI. So, again, thinking about our agents, thinking about our portfolio optimization systems, going forward, probably by the end of this year, you'll be able to ask an underwriting agent or certainly early into next year, you'll be able to ask the renewal, the Hx underwriting agent, which risks and things that you need to spend some time on, which rinky dinks or edge cases, as you said before, or strange things that we do last year during those renewals that we definitely don't want to get caught racing to the finish line at three in the morning again during this year.