InsTech - insurance & innovation
00:21:08 · Matthew Grant
[00:21:08] S1: I'm glad you mentioned that point about pain points because... I mean, no surprise, it sort of goes back to Amrit's background when he was an actuary, he'd also, can feature engineer, software engineer, so he could write code, he said, I understand from him, there's some things he wanted to do, couldn't do it when he was working for an insurance company and built a company to solve it. Actually, interestingly, we've got a webinar coming up with one of the large US insurers and they're asking them how they deploy AI, and they look for where people are complaining. When they get a complaint from somebody about they can't do something, they go, Right, okay, can we deploy an AI solution here to solve that problem? What I think is fascinating is at one level is we're sort of hearing about companies still struggling to articulate the large investments in AI, but actually in the companies themselves, I mean, I think you're a great example of this, people finding solutions to actually just create that efficiency at the operational level. But then I'm sure you're also finding ways to accelerate both the development of what you're offering to your clients with the hx tools themselves? Is there anything there you just want to draw attention to that is tapping into large language models or agentic AI or other ways of just improving that experience?
00:22:20 · Richard Gunn
[00:22:20] S2: Yeah, I'll probably try and tie the two together. What we find interesting is you have the approach where you're looking for specific pain points and you solve for them leveraging AI, and that kind of brings people on the journey, has some great wins, and generates an immediate ROI, or hopefully, depending on what kind of report you read. But then there's also a question that we ask ourselves is, where is there an opportunity to redraw the traditional boundaries of different roles or different capabilities as a result of the evolving capabilities of AI? So this is something we ask ourselves a lot in the context of actuaries, underwriters, and IT, like even pre Gen AI, a key area that we look to add value was to bring together those key stakeholders around a risk and allow them to huddle and iterate on a risk in a way that's very hard to do and you haven't got the right kind of fluid level of technology that sits in between. And so it's so important to do in the Northern London market, in particular, as well as kind of broadly across the commercial B and C space. But we see it outside of insurance as well. If I think about our product org, very traditionally, you have three key stakeholders huddling around the product development, if you will. You have the product team. You'll have the design team who'll focus more on the UI and UX, and you have the engineering team who actually build it, and then we'll subsequently support it. And we're seeing with the evolution of AI and kind of enterprise grade vibe coding, if you will, or enabling you to push something through to semi production much faster than you could beforehand, those three roles consolidate into what's being sometimes referred to as a builder. And that's also exciting because you can start to think around, maybe we can have one of our customer success managers who works very closely with one of our clients. And maybe you're one of our clients, and you say, like, no. Richard, I've got some challenges with my experience rating and the triangles and how they currently work. I've seen it work in this way more effectively.
00:24:00 · Richard Gunn
And if you and I could mock that up and then I could push that straight through to a builder who could actually flow that through to the code, can suddenly see that the feedback and the iterative nature of working closely with the clients and at the cold face and how you feed that into your product could really accelerate. So it's a challenge I'm asking myself on the go to market side too. Like, should we be seeing the advent of the first full stack AE account executive who can support on the customer success side as well as the technical engineering side? Or actually, are we just gonna see kind of the opposite of that point of view would be the key buyers of specialization. You'll able to see much more productivity for each of the stakeholders. So where you have solutions engineer, could they support five clients rather than two clients or 10 clients rather than five clients? And could an account executive, because you've got agent doing a lot of the administrative work, Could they carry a much bigger quota, or could they support more clients as well? So there's a lot of kind of testing and learning on that, lots of very healthy challenge from VCs around benchmarks that they're seeing, around productivity at the same time as speed. And, yeah, we're excited both in terms of how we can hopefully put that in the hands of our clients with some other agenting capabilities that we've been releasing, also internally, how can we use it all the way from products and engineering through the go to market.
00:25:13 · Matthew Grant
[00:25:13] S1: Frisch, before we leave that, can you just talk a bit more about vibe coding? Because you mentioned it in passing and my sense is that is gonna be a major part of what every person needs to think about in their career going forward. I'll let you explain what vibe coding is, but it could be as a concept and I think you touched on it in a very interesting sort of career development point of view. People score themselves either to people that can handle technology and they're going to stick their hands in and play around with it and those that can't. I think what you're saying is that, I'm trying put words into mouth, but the technology is not gonna totally replace everybody. I mean, you talk about a digital or bionic AE, maybe it will, not an underwriting for sure, but you kind of have to have your bionic arm and your tools that are gonna help you. Yeah, it'd be helpful just to talk a bit more about vibe coding. Again, if there's anything specific you can recommend to anybody who has sort of heard about it or wants to try it, is not quite sure where to start, that would be fantastic as well.
00:26:12 · Richard Gunn
[00:26:12] S2: Yeah. There's been an undulation of reporting around web coding in the news. Unsurprisingly, many of the the kind of podcasts that I listen to, turns out they're actually funded by some of the the the big kind of providers in the space. So I think, like, the most recent way before recording this was about whether or vibe coding was dying, and then I think it's Lovable reported that 5x their users over the course of the last two to three months. And so I won't try and kind of predict that. What the reflections I would have is, like, vibe coding is ill defined in terms of how it's being used, but as I understand or at least as I use the terminology, it's about... Often about allowing a nontechnical user or someone who historically wouldn't have jumped into the zeros and ones of code to to be able to... A natural language or in a low code, no code interface, be able to very quickly spin up either... And the Vibe coding definition will probably more likely lend itself to either mock up a quick kind of proof of value or kind of illustration of what it could look like all the way through to... And I think where we will certainly see it resolving to is actually production grade websites which you can use. And, like, a lot of the work and the development that's going on now is around actually how can you make it the code which is written by agents versus written by humans, either at the hands of a sophisticated engineer or a nontechnical user. How can they actually be greater in terms of the security and their performance than if they were coded by a human? That's where it starts to get exciting because it goes from, like, the quick wins now are... That illustration I gave beforehand, can you have a customer success manager who's inherently gonna be less technical than an engineer working with a customer to mock up what an iteration of your product could look like and then feed it back into that builder role who does have engineering capabilities towards actually that customer success manager mocking it up, running it through some programmatic testing in partnership with an agent, and then and then actually committing that for review. And I think there'll be, like, a spectrum of use cases there.
00:28:04 · Matthew Grant
[00:28:04] S1: So Richard, you covered a lot in there, lots of useful angles. What's going be happening in the next twelve months? What should we come back and talk about with you or one of your colleagues in November of 'twenty six?