Why the London Market is ready for intelligent automation

Robin: Welcome everybody to this week’s InsTech Podcast. My guest today is Elizabeth Wooliston. Welcome, Elizabeth. 

Elizabeth: Thank you very much. 

Robin: Now, you’ve got the title Chief of Markets at Artificial, and you’ve been in the job for six months or so. Are you enjoying it? 

Elizabeth: Yes, I am. Now, I’m sure you’re expecting to hear that answer, but the reason I’m enjoying it so much is it sits at the intersection of everything I find the most interesting, which is the complex market dynamics, which I’ve been learning for the last 30 years, and the relationships that I’ve been building up over those decades. [01:00] 

And finally, the technology is catching up with what the market actually wants and needs. So you bring all those together, it’s a really fun place to be. The team are highly motivated to deliver world-class technology, and then David and Johnny as the co-CEOs, founders, they bring so much energy to the team that it’s a really infectious place to be. It’s great fun. 

Robin: So let’s talk about these markets, and you say the tech is finally catching up. But is it? I’ve been talking about this for 25 years, and you must have some proof. Give me some evidence. 

Elizabeth: I think you and I, since we’ve known each other, Robin, we’ve been used to all the sceptics, and I would say that I have definitely been one over the years because we’ve seen Blueprint 2, we’ve seen large transformation programmes which have stalled and been enormously expensive and not delivered what they’d said they were going to do. Vendors promising revolution and often delivering incremental improvement. So the track record has rational scepticism, but it feels to us that the pressure points are different now. [02:00]  

So we actually see Blueprint 2 being wound down not as a defeat — it’s actually created a forcing function. Lloyd’s themselves have said that technology being deployed by market participants had advanced markedly since Blueprint 2 was even conceived. So rather than waiting for this top-down infrastructure overhaul, individual firms are now moving independently and can move faster because of that. And then you add on top of that agentic AI capabilities — we tend to separate those two — that they can reason across unstructured data, which has always been one of the biggest challenges the insurance industry has had, and can help navigate complex placement logic. Things that weren’t available five years ago are now available across the market, and are, in some instances, reasonably easy to use. So we are definitely seeing a step change. 

Robin: I think what’s driving that — so this is slightly a leading question, because it seems to me that the brokers, [03:00] particularly the big wholesale brokers, have always been the gatekeepers to this. And to some extent, progress is as fast or slow as they want to make it. And you can see palpably that they are getting mobilised. Is that the single biggest reason why we’re actually seeing a bit of momentum now? 

Elizabeth: I can see how the optics are like that, and I do think the larger brokers are a significant part of it, and you use the word gatekeepers. That’s what I would use as well. You control the submission flow. You control the pace of market change. But what’s different is for the bigger brokers, it’s not a digital ambition anymore. It’s a digital strategy. They all have digital strategies. They have delivery roadmaps that they’re working to, so it’s not abstract anymore, and they’re building infrastructure that changes how the risk is presented and compared and placed, which is enormously exciting. So we’ve got the distribution being highly digitised in certainly the larger brokers. But I wouldn’t want to give the brokers all the credit. [04:00] 

We’re also seeing the carriers pulling from the other side. We’re going into a softer market. Where do underwriters find more margin? Often in ironing out friction. There’s a genuine broker demand for carriers to respond faster, but not at the detriment of risk appetite and great underwriting. It’s just being a little bit faster and a little bit smarter about how you do it. The genuine carrier demand needs the tools to do it. 

Robin: But just to drill down to that a little bit, because I think there are always progressive carriers, and they’ve been investing in this stuff for some time and have capabilities that are superior to the herd, as it were. But isn’t what’s changed that the herd now has to move? What I mean by that is these carriers are incredibly dependent on those big brokers for a lot of their distribution, and where they are prepared to show what their digital strategy is, you haven’t got much choice if you want to see that business, [05:00] have you, than to align over the long term with what your broker partners want to do? 

Elizabeth: I think in short, yes, but the carriers who are doing this well are framing it as an opportunity rather than a compliance exercise. It’s easy to generalise, but let’s talk about the ones who depend heavily on the major brokers for their flow. They know that if they can’t receive submissions in a structured form, respond with appetite quickly, and participate in smart follow decisions at scale, they may lose access to that type of risk. But the follow market’s a really good example, I think. So historically, following a lead was largely administrative. You took the rate and the terms set by the lead, and moved on. Smart Follow changes all of that, and you can algorithmically assess whether a risk fits your book at speed and at scale. And the way that we build it, you can adjust your appetite all the time. So it becomes a genuine [06:00] underwriting discipline rather than a processing exercise. And the carriers who are ready on that will have a competitive edge. That said, some carriers are more exposed than others. Those with strong direct relationships or highly differentiated appetites or particular niches may have more time. But those who are genuinely commodity followers of large-volume classes, I do think face some pressure to move faster. 

Robin: Where are you seeing most of the focus? You talked about Smart Follow. That’s clearly been underway for a few years now. How about facilities? 

Elizabeth: A lot of the focus has been on facilities. I think that’s fair to say. We see them as a beachhead rather than a destination. It makes sense that they go first. But what we’re seeing at Artificial is that the appetite to expand beyond facilities is accelerating, particularly for open market placements where the prize is bigger, but the complexity is so much higher. The [07:00] Smart Follow on open market is genuinely hard. It’s genuinely hard to codify that appetite for us, but we’ve been doing it for a long time. We understand how it works, and dealing with slips, bespoke wordings, complex risk data, unstructured documents — it’s a difficult thing to do, understanding the mechanics of insurance underneath that. So we are building some of that for certain carriers. But you’re absolutely right, it starts with facilities. But ignore open market at your peril. The other piece that we’re seeing is midterm adjustments and the renewal cycle. So the original placement gets an awful lot of attention, but actually a policy is a living, breathing thing for 12 months. If it’s political risk or something like that, it’s way longer. And there’s an enormous volume of repetitive, data-intensive work that’s really ripe for automation. So we are doing a lot of work with carriers on that because we want the brokers and carriers looking after that [08:00] relationship, talking about the risk, not doing the sort of manual, boring, repetitive tasks that we’ve been doing for decades. 

Robin: Doesn’t this leave some carriers, perhaps quite a lot of carriers, with a lot of work to do in a quite short period of time? And I’ve made this observation before. I think Artificial has been in this space three or four years — you’re very knowledgeable, perhaps longer. There are not that many vendors who are able to get a deep understanding across these various smart follow and open market. And then some have got money, some have not got money. Are you seeing any sense in which people think they can do this themselves? The old build-versus-buy argument. What are you seeing there?   

Elizabeth: Yeah, we have an enormous amount of conversations across the market, and we see different models across all of them. The first thing I’d say, it doesn’t have to be an either/or. There can be a very happy combination of build versus buy. [09:00] The Artificial model is actually that we spin something up for you, whatever that might look like, but we then try to put the power back in their hands, and we train individuals within organisations to configure the platform themselves, which is a very happy path, I think. So they’re not raising tickets and waiting for however long to get a change made. They can actually do it themselves. We can build it externally with a lot of help from the company that we’re partnering with, and then when we get to a place where we’ve all decided we’re ready to go, we would hand the tools back to them. Build-or-buy decisions in specialty insurance have a particular complexity that gets underestimated.

And you mentioned five years — the domain knowledge required to understand what correct looks like, slip structures, MRC (Market Reform Contract) standards, reinsurance constructs, cedent data quality, it takes years to accumulate. And when you start building on general-purpose large language models, you’re essentially [10:00] building that domain layer from scratch on top of foundational models that weren’t really trained for specialty insurance. So we spend a lot of time with users, the underwriters themselves, making sure that what we build responds to what they need day to day, and not what the LLM thinks is right. The real risk of building isn’t technical failure, it’s opportunity cost, I think. When you’re 18 months into building something that already exists in a mature form elsewhere — i.e. Artificial — the market’s moving, and we can spin something up in 10 to 12 weeks across a business and prove that out then across the rest of the business. If you’re starting from scratch, if we are doing a real build-versus-buy comparison, I would caution because you can lose an awful lot of time getting to where we already are, and we can come in and deploy things much, much faster. 

Robin: Carrying on that theme, assume that they go with [11:00] Artificial as a choice or other vendors — foundationally, what do they mean? You must see various degrees of quality and maturity in the infrastructure, but what’s the minimum platform that you can work with in terms of data and tech? 

Elizabeth: So I’d break it down into three things. First of all is the data structure. That’s got to be right at the beginning for all the downstream processes to be right, and to really grease the wheels of the risk going through the system. So we know that data arrives in every conceivable format: PDF, spreadsheet, handwritten annotations, broker-specific templates. And before you can really build those intelligent workflows going through, you need to take that data and put it into a structured representation, and then you can put AI onto it to reason. So that means investing in data normalisation, in good ingestion technology, [12:00] not as a parallel workstream, but as the foundational step. And then the second is a domain-appropriate data model. Generic insurance data models don’t work for specialty insurance. We’ve seen it — it doesn’t work. You need something that understands the semantic difference between a line of business, an attachment point and an aggregate retention, and that can understand and represent reinsurance constructs correctly, that can handle delegated authority structures. The reason I bring all of these out is that’s where Brossa, our domain-specific language, comes into its own. It does all that critical work. It’s not a nice-to-have. It’s the difference between AI trying to process a reinsurance slip and something that genuinely understands that piece of data and then knows what to do with it, because it’s not a standalone piece of data. It has to go somewhere, and you need to deeply [13:00] understand what it is — is it correct — before handing that off. And the third one is really governance. So the firms moving the fastest aren’t the ones with the least governance. They’re the ones where governance is built in. It’s not bolted on. Clarity in the foundations gets you clarity in the outputs. 

Robin: Looking in, as I do, one thing I observe is that there’s a lot of interest, particularly on the agentic side of things — a lot of point solutions, a lot of different solutions solving problems. And then because there’s enthusiasm, because people want to encourage that enthusiasm, they proliferate, and the next thing you know, you’ve got lots and lots of point solutions. At some point, if you are going to be the kind of overarching platform capability of choice, you then walk into that, don’t you? Are people starting now to grapple with not just what it is they need to do from a broker point of view, but the fact that they’ve got a proliferation of point solutions and all [14:00] this now needs some kind of overarching, smart, well-governed infrastructure? That sort of feels like where we are right now. 

Elizabeth: Yeah. I think foundational gaps will remain. I think what I love about the AI conversation is it’s really accelerating the tech conversation for us across the market because you’ve got the Claudes and you’ve got the Geminis, which are really easy to use and just make technology feel really accessible again. So that’s really exciting, and I think the enthusiasm for it is absolutely justified — these tools, we couldn’t do this only 18 months ago, and now we’re able to do incredible things. Our AI arm are retraining themselves every three months because the technology’s moving so fast. But enthusiasm without the core architecture is just going to create a really fragmented landscape. So I was with a big broker this week who was telling me, “You don’t seem to have much competition in this space, Elizabeth.” And I think that’s because tech [15:00] partners who can credibly anchor really large investments are rare, because the bar has become higher due to AI and because of the failed projects that have come before us, and you need really deep domain expertise. So I keep coming back to that. You need a track record of production deployments at scale, and you need a foundational layer, not just a UI layer. So what’s mostly being marketed as agentic AI for insurance is a wrapper with some insurance-specific prompting, and that’s great, but if it hasn’t got the foundation, it’s not going to come with the right answers. We come at it slightly differently to some of the AI-only vendors who, to your point, are individual point-solution people, which can be great for a point solution, but not if you’re looking to deploy enterprise-wide technology. 

Robin: Changing tack slightly, tell me about your own AI lab. You launched AG Labs earlier in the year. What’s it do, and what are your plans for it? 

Elizabeth: Yeah, so Alexi heads up that team — he’s been with Artificial seven years, he knows us well. But we wanted to double down in that area because we wanted a working environment where we could test and develop and validate agentic AI capabilities for the London specialty market. So Alexi and his team go into, exactly to the last conversation, carriers where we’ve identified point problems that aren’t for Artificial, and they can go in and almost sandbox with them. You give us some data to work with, and we will give you some outputs. And they can do it in a couple of days, which is really exciting, because we know that something like an Artificial [17:00] platform will take 10 weeks or so to spin up. So to be able to put something in people’s hands in days is really exciting. We’re working on all sorts of projects with carriers there at the moment, and one of the nice things is if they have got secure platforms that they’re very comfortable with, they can run the agentic AI on top of that. So what we’re trying to do is twofold really: test and learn with the market about what safe agentic AI looks like, but also look round corners, because in an ideal world, in a decade — dare I say that to you, Robin, as you’re always saying it’s never gonna happen, it’s never gonna happen — imagine a world where we do have more agent-to-agent transactions going on. We want to get the market comfortable that that can happen, but it starts with us with small steps, and that’s what AG Labs are doing.   

Robin: So you mentioned earlier, or I think I did, the levels of enthusiasm and the fact that underwriters and users are starting to see tooling that they really like. Does that [18:00] mean that the sort of cultural resistance which has been such a barrier for this is diluting? And then as part and parcel of the same thing, you talk about going in and working with the carriers — you can’t do that if the carriers haven’t got people who want to engage, and the people you need to engage with are the users much more so these days than the IT team. Tell me what’s happening there, because it seems to me they’ve got some level of enthusiasm, plus this very collaborative model for companies like you building out to meet the requirements of individual carriers. 

Elizabeth: Yeah. We don’t particularly like the word vendor. The way that we work is much more of a partnership. I know it’s a much-used word, but it really is — we’re in the trenches together, as I describe it. I think, to your point about cultural resistance, it’s usually not about the technology. It’s usually about the fear of connection being lost — with colleagues, with clients, and with their craft of underwriting, which is a [19:00] craft. And we come from a position of deep respect on all of that, particularly as many of us have been market practitioners. We understand the power of the relationship with the client, the power of the relationship with the underwriter or broker, so we don’t want to get in the way of that. That’s not our job. Depending on the type of project — for example, we’re doing a big deployment at the moment where they just want to get to the risk as fast as possible, and then they’ll do their job from there. So we ingest the risks, we sort them, we triage them, we know that it’s in the appetite, we know that it’s right in their sweet spot, we know that it’s from a broker they want to serve. So we do all of that in moments, and we put it right at the top of their task list. So they are getting back to the business that they want to get back to as soon as possible, not going through 20 other emails to get to that. So they don’t lose connection, they don’t lose their client focus — we’re just getting rid of noise for them in order to do their [20:00] business. So in my head, it’s protecting those connections, not replacing them. I don’t get much resistance around “this’ll never work.” It’s more, “I want to understand how it works before I commit.” So there’s definitely more belief that it will eventually deliver, which is great. The other thing — back to the LMG (London Market Group) report a little while ago — it was saying there are as many people over 50 as under 30 in the London market workforce. But that under-30 cohort is arriving with a completely different assumption about how work should be done versus the 50-plus, who’ve been very used to doing work in a certain way, and we built a 160 billion industry in London on it. It’s been successful in its own way. But the younger generation aren’t defending a workflow — they’re looking for the best one, and that’s the kind of thing we can help deliver, freeing up underwriter time to help train those younger people coming through on the craft of underwriting. 

Robin: Change of tack completely. [21:00] I think January or February, you announced a $45 million fundraise, which is pretty impressive in itself. You can’t get that kind of money unless you persuade investors that you’ve got something compelling and that there are plans for it. How is that money being spent? I know US expansion is on the list. 

Elizabeth: Yeah. So Series B is mostly about product-market fit, and if we were to go to Series C, we’d have to show growth and expansion. It’s going well, and it’s certainly different to steady state, although I don’t think Artificial’s ever been in steady state. But we’re spending the money quite intentionally in three areas. We’ve talked about AG Labs. We are building out the US market presence, which is an obvious way to go, and we’ve appointed Eric Joost in that role, who’s well-known across the industry. We’ve also just planted a flag in Europe with Christopher Lohmann building out the team there. So we are a lot more international than we were six months ago. I think [22:00] just pausing on the US opportunity, it’s really significant. We know that the excess and surplus (E&S) market is large and growing. I was with some folks from Elayne when I was in New York, and they were talking about the increase in size in the E&S market just in that state. And it’s got a lot of the structural inefficiencies that London has as well, around complex risk, manual placement processing, data quality. So the advantage that we bring to the US, we’ve often already solved for in London, so it feels like a natural place to go for US carriers, MGAs and brokers who are facing the same challenges. 

Robin: How big are you these days? Give us an idea of the size of Artificial.  

Elizabeth: When I joined in January, Robin, we were 67 people. I remember that number quite specifically because of the whole six-seven thing. And then we’ve gone to 131 in five or six months, so kudos to our recruitment department and everybody across Artificial who’s been interviewing and finding the best people. [23:00] We’ve got a target of about 170 headcount by the end of the year. But honestly, the metric I care about is not headcount or ARR. It’s about production coverage. We want to get as many nodes as possible for the transactions to go across the market digitally — how much live risk is flowing through our infrastructure. Last year we did about 40 billion, so that’s an indicator of whether we’re genuinely embedded in an organisation — how much they’re really using the platform and transacting risk through it and not just piloting around the edges. And that number’s moving meaningfully in the right direction, so that’s very exciting. 

Robin: I hear that and I think, have there been growing pains? You can’t grow that fast and take on that many people without the occasional moment of pain. Has there been? 

Elizabeth: Absolutely, yeah. The pressure on delivery is constant. We’ve got to where we are because we deliver great technology, and you can’t let that slip because you’re adding people quickly. The culture has to be strong enough to absorb [24:00] new starters faster than it can be diluted by them, is the sort of metric we use. The other growing pain is prioritisation. When you’ve got a strong market pull, you get more opportunities that you can pursue really well — where you say, “This is our ideal customer profile, it sits right in the heart, we want to do it.” But we’ve had more discipline around saying no or not yet to things that are genuinely interesting, but we just have to prioritise some work sometimes. So we’ve been saying no to some RFPs that we think we might not be able to deploy as quickly, or that might not sit in our suite, or that fundamentally somebody else might do better. So we’re getting much better at being explicit about where our focus is and where it isn’t, both internally and with clients. Growing pains are a sign of something good, and we talk about this a lot in the organisation — it’s a good problem to have. It’s whether we’re learning from everything we’re doing to go faster. The hiring point is a really good [25:00] one. We’ve been focused on, as all organisations, hiring the right people, and we look for three things: highly capable, low ego, and lead with kindness. If you’ve got those three things in people, you’re going to naturally get a culture that moves faster, where every conversation leads with kindness. You don’t get blame culture, and so on. So those are fundamental aspects of people’s character and knowledge base that we look for when hiring. The other great thing about Artificial is that we don’t burden the company with unnecessary bureaucracy. We’re trying to keep that to an absolute minimum so that we can get out of people’s way and let them do their jobs. 

Robin: And I wanted a job at Artificial. Afraid I’d be ruled out on the ego. That would’ve done for me. 

Elizabeth: You’ve got to hit all three, Robin. 

Robin: Yeah, you’ve got to hit all three. Look, it’s really good to see you and to catch up.[26:00] Thank you for the update on Artificial. As you know, we’ve been very supportive of you. We think you’ve got a fabulous opportunity given the market dynamics you’ve spoken about, and the expertise that you’ve built up over the years. So good luck with that. Please send my best to David and to Johnny, and come back and join us again once your US expansion has been cracked. 

Elizabeth: Will do. I’ll bring Eric along as well — he can talk to you about it. 

Robin: Marvellous. Thanks for joining me.