InsTech - insurance & innovation

AddressCloud expands geospatial risk data with Kestor Capital investment and AI integration

28 min Listen to the original Episode page
InsurTechGeospatial DataArtificial IntelligencePrivate EquityProperty UnderwritingRisk Management

Summary

Mark Varley and Peter Thompson discuss AddressCloud's growth, highlighting a 2025 investment from Kestor Capital and Thompson's appointment as Non-Executive Chair. The speakers emphasize the company's focus on pre-bind accumulation management and sub-half-second data delivery via pre-caching architecture. They address emerging risks like wildfire in Europe and the cautious but active adoption of AI by traditional insurers. The episode outlines AddressCloud's shift to providing data directly into customer data lakes using the Model Context Protocol (MCP) to support GenAI tools. Finally, they discuss the roadmap for global expansion, particularly into the US, and the importance of validating AI outputs to prevent quality degradation.

Episode sections

  1. 0:28 – 2:59

    Company Background and Kestor Capital Investment

    Mark Varley founded AddressCloud in 2015. In 2025, Kestor Capital invested in AddressCloud as part of their Kestor Capital Three fund. Peter Thompson joined as Non-Executive Chair alongside this investment. The core value proposition remains providing geospatial risk data to insurers, though the delivery methods are evolving.

  2. 2:59 – 4:38

    Peter Thompson's Role and Motivation

    Peter Thompson, with a background in insurance and distribution (including BGL Group), joined as Chair after being impressed by the team and proposition. He was introduced to Kestor Capital by Mark Varley, which is unusual in private equity. He views the role as involving significant work on talent and investor support, distinct from traditional corporate chair roles.

  3. 4:38 – 6:54

    Private Equity vs Venture Capital and Scaling Goals

    Mark Varley explains that AddressCloud was profitable with 18 staff before taking investment. The Kestor Capital investment was taken for skills, professionalization, and scaling, not just capital. The management team and staff became shareholders. Varley states there are no regrets after one year.

  4. 6:54 – 8:49

    Emerging Risks: Wildfire and Accumulation Management

    Varley highlights emerging perils like wildfire, noting significant events in the UK. AddressCloud partners with a Polish firm for a wildfire model covering Europe/UK. A major investment area is pre-bind accumulation management, allowing insurers to assess specific peril exposure (e.g., flood) at both individual property and aggregate portfolio levels.

  5. 8:49 – 11:10

    Data Speed and Pre-caching Architecture

    AddressCloud addresses speed in two ways: onboarding new data and consuming it. They use a 'pre-cache' architecture, pre-calculating peril scores against building footprints. This allows them to return a full stack of peril scores in under half a second, crucial for price comparison websites and fast underwriting journeys.

  6. 11:10 – 16:15

    AI Adoption and Guardrails in Insurance

    Peter Thompson notes AI is changing rapidly. Traditional insurers are cautious due to cyber risks but are piloting AI. He cites an example where an AI team delivered a product 10x faster with higher quality than a traditional team for a restrictive task. Conversely, he warns against letting LLMs 'mark their own homework,' citing a case where an LLM rewrote quality hurdles to pass QA.

  7. 16:15 – 22:11

    AI-Ready Interfaces and Model Context Protocol (MCP)

    AddressCloud is moving beyond APIs and UIs to provide data directly into customer data lakes (Snowflake/Databricks). They are adopting the Model Context Protocol (MCP) to allow GenAI tools to access proprietary data securely. This allows them to preload context (e.g., UK-wide flood context, non-standard property flags) into the AI interface, improving decision quality.

  8. 22:11 – 28:26

    Future Roadmap: Global Coverage and Analytics

    AddressCloud's mission is to locate and describe every building globally. They are expanding to the US with a new office and lead. Future focus includes faster data refresh rates (closer to real-time) and providing complementary analytics to help customers understand ROI and value of their data usage, leveraging aggregated quotation data.

Claims worth verifying

Kestor Capital invested in AddressCloud as the sixth investment in the Kestor Capital Three fund.

Unidentified speaker (speaker_0) ·0:28

How to verify: Check Kestor Capital's official website or press releases for announcements regarding Kestor Capital Three portfolio companies and investment rounds. For recent events, verify 2026-08-15 through 2026-11-15. Use another period only when the speaker explicitly names it.

Mark Varley founded AddressCloud in 2015.

Unidentified speaker (speaker_0) ·0:28

How to verify: Check AddressCloud's official website 'About' page or LinkedIn profile of Mark Varley for founding date. For recent events, verify 2026-08-15 through 2026-11-15. Use another period only when the speaker explicitly names it.

AddressCloud had 18 staff and healthy revenues before the Kestor Capital investment.

Mark Varley ·4:57

How to verify: This is a specific financial/operational claim. Verify via AddressCloud's investor relations materials or press releases detailing the Kestor Capital investment round. For recent events, verify 2026-08-15 through 2026-11-15. Use another period only when the speaker explicitly names it.

AddressCloud has active customers in the US, Canada, and mainland Europe.

Mark Varley ·2:37

How to verify: Check AddressCloud's website for case studies or customer logos, or press releases detailing international expansion. For recent events, verify 2026-08-15 through 2026-11-15. Use another period only when the speaker explicitly names it.

AddressCloud partners with a Polish firm for a wildfire model covering Europe and the UK.

Mark Varley ·7:35

How to verify: Check AddressCloud's partner page or press releases for details on the Polish wildfire model partner. For recent events, verify 2026-08-15 through 2026-11-15. Use another period only when the speaker explicitly names it.

AddressCloud can return a full stack of peril scores in less than half a second.

Mark Varley ·9:39

How to verify: Check AddressCloud's technical documentation or API performance benchmarks published on their website. For recent events, verify 2026-08-15 through 2026-11-15. Use another period only when the speaker explicitly names it.

AddressCloud has migrated its infrastructure to Databricks.

Mark Varley ·17:15

How to verify: Check Databricks' customer case studies or press releases for mentions of AddressCloud as a migrated customer. For recent events, verify 2026-08-15 through 2026-11-15. Use another period only when the speaker explicitly names it.

JBA created a global flood model for the UK at 5-meter resolution.

Mark Varley ·17:35

How to verify: Check JBA's official website or product documentation for their UK flood model specifications and resolution. For recent events, verify 2026-08-15 through 2026-11-15. Use another period only when the speaker explicitly names it.

The British Empire was the largest empire in history by landmass.

Unidentified speaker (speaker_2) ·26:00

How to verify: Verify via historical records or reputable encyclopedias (e.g., Britannica) comparing land area of historical empires. For recent events, verify 2026-08-15 through 2026-11-15. Use another period only when the speaker explicitly names it.

Vern Harnish is the author of the 'Scaling Up' methodology.

Mark Varley ·26:37

How to verify: Check the book 'Scaling Up' by Verne Harnish (note: speaker says Vern, book says Verne) for authorship. For recent events, verify 2026-08-15 through 2026-11-15. Use another period only when the speaker explicitly names it.

Do your own research

  • Verify Kestor Capital's fund structure and investment thesis for AddressCloud via their public filings or press releases.
  • Check AddressCloud's customer list and geographic revenue split to assess reliance on UK/Ireland vs. North America/Europe.
  • Review the Polish firm partnership details and the specific methodology of their wildfire model for European markets.
  • Analyze the technical specifications of the 'pre-cache' architecture and its latency claims against industry benchmarks for API response times.
  • Evaluate the risks of AI adoption in insurance by reviewing regulatory guidance on LLM hallucinations and validation protocols.
  • Assess the competitive landscape for geospatial risk data providers in the US market, including established players and new entrants.
  • Investigate the Model Context Protocol (MCP) adoption rate among major data lake providers like Snowflake and Databricks.
  • Review AddressCloud's financial performance post-investment to determine if scaling goals are being met without compromising profitability.
Full transcriptexpand
0:08
Unidentified speaker (UNKNOWN): Mark,
0:08
Unidentified speaker (speaker_0): Peter, great to be talking to you today. Mark, we've spoken a few times before, so you've got to compete against yourself, see if you can get more downloads from the last time. And Peter, I believe it's your first podcast. So thrilled to be talking to you. Welcome. Thank you, Matthew.
0:21
Mark Varley: Thank you, Matthew. Great to be back again. I think it's been a few years or a couple of years since we did the last one. Lots changed since then. So looking look forward to digging in.
0:27
Unidentified speaker (UNKNOWN): Right.
0:28
Unidentified speaker (speaker_0): Well, just a bit of background first of all, then we'll catch up on what has indeed changed. So Mark, you founded AddressCloud back in 2015 and last year Kester announced that they had taken on an investment in DressCloud as the sixth investment in the fund Kestor Capital three. We've been delighted to have your support, Mark, as AdressCloud, for a number of years at Instac. I Actually, so said we did our first podcast back in twenty
0:47
Unidentified speaker (speaker_0): 2020. twenty, and you are based in wet what rainy south of France does now.
0:52
Mark Varley: Yeah, that's correct. It was lockdown podcast, I think, wasn't it?
0:55
Unidentified speaker (speaker_0): And Peter, you joined Address Cloud at Chairman alongside the investment, and you've got a significant industry experience. We'll talk a bit about that. Pleasure to have you along as well.
1:03
Unidentified speaker (UNKNOWN): Thanks,
1:04
Unidentified speaker (speaker_2): Matthew. Delighted to be on the podcast and background insurance through and through, both insurer side and distribution. Let me
1:11
Unidentified speaker (speaker_0): just read you how Kestor described Address AddressCloud Cloud at the time of the investment last year. and see if you've got anything to add to this. So AddressCloud is a fast-growing information and data leader with an exceptional reputation, providing insurers with geospatial risk solutions across their underwriting journey. Address cloud enables its customers to understand peril and climate risk in granular detail, resulting in
1:31
Unidentified speaker (speaker_0): faster and more informed underwriting decision. Sounds like the Address Cloud pad I know. How does that felt cut with what you're doing today?
1:37
Mark Varley: Yeah, absolutely. I think that core value proposition hasn't changed. It was around this time last year that we were kind of doing the rounds with the private equity companies, and the pitch to them remains the same as it is today. So it's still understanding where the property is, what its physical characteristics are, what its risk profile is like in terms of specifically around geographic risk and geographic perils. And
1:57
Mark Varley: then finally what its accumulation risk looks like. So what are the other properties that are insurance nearby? So that proposition remains unchanged. It's the way that we're delivering is changing. And what we're seeing is an emergence in the way that people want to consume the data. But the actual core proposition is exactly the same. One of the things that attracted KESTA to invest in us and some of the other companies that we
2:17
Mark Varley: spoke to was the breadth of where we were able to offer that service. So historically it's been UK and Irelands, but with the emergence and changes and improvements. in data and technology, we're now able to offer that service more broadly. We've got some a couple of active customers at the moment who are using our services in North America, so US and Canada,
2:37
Mark Varley: and then also in mainland Europe.
2:39
Unidentified speaker (speaker_0): And I suppose for those less familiar with a kind of P investment versus the venture capital one, this is very much a company that's looking to invest in you for growth on the basis of successful revenues already. If you've got a pretty healthy pipeline of things. coming out. Peter, for you, the role of chairman, for those who aren't close to what it means today to be a chairman of a scale up, they might not appreciate as quite how much work is
2:59
Unidentified speaker (speaker_0): involved. What was it for you that motivated you to take on the role of chair for Address Cloud?
3:05
Unidentified speaker (speaker_2): It was a really exciting opportunity. I met Mark and Tom Holdeness, key part of the business, quite a few months prior with a different inquiry, and I was just enthralled really by the power of the proposition and also just the enthusiasm and passion of Mark and Tom. So I was really taken aback. So it was before the sort of conversation with Kestra had taken place. And
3:25
Unidentified speaker (speaker_2): I just said to Mark, Lovka, if I can help at all in your journey, I think it's such an exciting journey. If I can help, I'd love to, whether it's industry contacts, tax, whatever it may be. And it was Mark actually that introduced me to Kestor, Kestra, unusually. That's not typically how it works. in the private equity world. And having met Kestra and that was a key moment for me because that
3:44
Unidentified speaker (speaker_2): they're a very impressive P house. Their focus is really about businesses that have got either horizontal or vertical kind of integration capability. They are looking at businesses that solve very complex problems for their customers using data technology to do so. And yeah, I thought the combination and the
4:04
Unidentified speaker (speaker_2): power of Kester and their support and know-how, together with the enthusiasm and the solutions that Mark and Tom and the team have developed, I thought it was a really powerful combination. So when I was asked to take the chair role, absolutely I took it. And you're right, Matthew. It's a very different role to in a scale-up Up to one that you might imagine for a far more established, maybe
4:24
Unidentified speaker (speaker_2): corporate chairman role. It's quite involved. but in a good way, particularly looking at talent, helping Mark grow to become the CEO that he is, and making sure that the investors are really supporting the business in every way that they can.
4:38
Mark Varley: Just to echo what Peter was saying there. So, from our perspective, you mentioned Matthew around the difference between kind of private equity and venture capital. As you may recall, we've never taken any investment. So I started the business on my own. We grew it up to where we were last year. We were 18 staff, good healthy revenues, lots of loyal customers. We didn't take the
4:57
Mark Varley: investment for the money. We took the investment really for the skills and the help that we needed. We got to the business to the level that we could get it to, but we recognized that what got us to where we were is not going to get us where we want to go. And so for me, the management team, all of our staff, uh the point of the investment last year were all shareholders. as well. For us, it was around how
5:17
Mark Varley: can we professionalize, how can we scale up? And again, it's around ourselves and having access to people like Peter is something we would never have had the resources to be able to do before, or nor the network. And so for us, it was finding a partner who could help us do that. And actually the private equity in general, and then Kestov specifically, was just really attractive.
5:37
Mark Varley: A year later, absolutely no regrets, it's the best thing we've ever done.
5:41
Unidentified speaker (speaker_0): Fantastic. And I think Peter's been a bit modest about his background. So, Mark, perhaps you might also just tell us what it was about Peter's own background and experience that it was why you approached him.
5:50
Mark Varley: Well, Peter has got a phenomenal background. Last while I was at CEO of BGL Group, and we we spoke to a few prospective chairmen, and Peter just got it from day one. He had our had the sort of deck that we'd put together for the private equity companies who was able to go straight to the heart of the value proposition, completely got it, completely understood it. Obviously, has that underwriting background and the core target market is
6:10
Mark Varley: very much still underwriters and CUOs. And Peter comes from that background. The nice things about Peter is a very, very senior but also very humble guy. And I think that in terms of the values of us as a business, that where it resonated really, really well with us. Having someone with the gravitas. the experience to both challenge the value proposition and
6:30
Mark Varley: help push it forward together with that experience as an as a leader is just made him ideal candidate for the role.
6:37
Unidentified speaker (speaker_0): Obviously, data, physical risks, we're all aware of those they are dominating how insurers are thinking about how they assess risk, how they assess losses. But what are the two or three areas you'll see that are becoming most prominent in in the current year and should be thinking about in the years ahead? and how you're going to be helping insurers assess those?
6:54
Unidentified speaker (UNKNOWN): And
6:54
Mark Varley: there's a couple of areas. It's the breadth of the, as we mentioned, the emerging perils, so the climate change, we can see that absolutely right in now. And this year actually seeing even in the UK some of the significant wildfire events. So we think the that the way and the demand for better, more richer climate-related data.
7:15
Mark Varley: is something that is more relevant now than it's ever been. And so if we still see this as an area of demand. So for having an organization like us who can go out there, find out we've got a fantastic partner in Poland who provide us with a wildfire model that covers Europe and UK. There's been a lot of focus on US wildfire, but historically not alone not so much on UK in specifically
7:35
Mark Varley: or Europe in general. And so being able to connect our customers with the local latest emerging data and get that into the platform. But then also for me, an area that we've invested in heavily over the last 18 months is our accumulation management capability and in particular pre-bind accumulation. We don't see a lot of our customers being able to do by
7:55
Mark Varley: the pre-bind accumulation at all or doing it at a peril specific level. And so I think for me the combination of
8:03
Mark Varley: being able to do, so for example, flood specific. So understanding these are the properties that I insure, understanding the specific risk for each of those properties individually in terms of its flood, but then also understand an aggregate, this is my exposure within this flood catchment, for example. I think the sophistication of combining those two different views of risk is something that insurers
8:23
Mark Varley: don't do much or at all at the moment. And I think isn't a very powerful value proposition.
8:29
Unidentified speaker (UNKNOWN): Well,
8:29
Unidentified speaker (speaker_0): certainly, talking about wildfires, a few weeks ago, we all saw the terrible wildfire in Starbridge, 19 homes destroyed. Small by the kind of US context, but certainly brings home to us here in the UK the devastating effect the wildfires can have. So you can see how that's getting more and more important. And you talked a little bit about there about breadth. I'm and thinking also about speed. What's the balance you're seeing in terms of the
8:49
Unidentified speaker (speaker_0): accuracy that people need and the analytics? in the analytics and the speed of which they want to be able to get that information?
8:54
Unidentified speaker (UNKNOWN): I
8:54
Mark Varley: guess two elements of speed. So there's that's the speed of being able to onboard new data. And that's something that we've always
8:59
Unidentified speaker (UNKNOWN): you
9:00
Mark Varley: know prided ourselves on is to be able to bring new data products to the market very, very quickly. But then the speed of actually being able to consume that data. So being able to consume that Assume within a an existing underwriting journey. Obviously with home insurance we're talking specifically home insurance here. Most home insurers will be selling through price comparison websites, which Peter knows a lot about, having
9:19
Mark Varley: been BGL being the parent company of compare the market for many years, a big customer of ours. But you know, when you're on a price comparison website, you need to be able to respond very, very quickly. So again, one of the reasons for us creating address cloud and specifically the way that we've architected it is that everything's pre-cache. So we basically pre-cache all of those perils
9:39
Mark Varley: against every building footprint. And that allows us to be able to respond extremely quickly within you know sub half a second, we can give a full stack of perils and adding in an additional attribute is fairly trivial for us. So again, I think two aspects being able to say a customer saying to us, actually, we really care about wildfire. Can you add that in?
9:59
Mark Varley: And by the way, can you bring it in so it doesn't affect the speed to quotes? Are the things that Address Cloud can absolutely power insurers to do?
10:08
Unidentified speaker (UNKNOWN): And
10:08
Unidentified speaker (speaker_0): when you mention pre cache there, that's the technical term cache as in C A C H E, not the money one, in terms of actually the preloading the data so it doesn't delay the runtime. Have I have I got that correct?
10:20
Mark Varley: Yeah, absolutely. Fundamentally, what Address Cloud is a big stack of data. We're bringing in, we've got a team full of geographic experts or geo. nerds as I like to call myself included. And what we're doing is we're ingesting all of this very highly complex data and then ultimately cookie cutting that and then linking that back to a building footprint, a LAN parcel, an address point.
10:39
Mark Varley: And because we do that all in advance, we're able to store that data and serve it up very, very quickly.
10:45
Unidentified speaker (UNKNOWN): And
10:45
Unidentified speaker (speaker_0): so Peter, as we're talking about as the technology and certainly playing both in your role and as you see things now. because it's moving fast. What's your sort of experience been and your impression of how senior leaders and insurance companies now are thinking about understanding technology and this balance we've all been aware of how fast or how slow to go with AI. What you're seeing, and maybe even some of your own experience about how to
11:06
Unidentified speaker (speaker_0): assess whether you build versus buy versus just wait and see what's happening?
11:10
Unidentified speaker (speaker_2): Yeah, it's a really dynamic situation at the moment, Matthew. It's fair to
11:13
Unidentified speaker (UNKNOWN): say.
11:14
Unidentified speaker (speaker_2): And I think AI itself is changing so quickly, so in terms of its application. And the insurance industry, well, I think we're in a number of different camps here. On the one hand, you've got some fairly traditional businesses who are aware of the advent of AI and the potential risks as well as opportunities it can bring to a business. I mean, cyber in
11:34
Unidentified speaker (speaker_2): particular, is something that everybody's particularly concerned about and want to make sure that any vulnerability vulnerabilities are are covered where they can be. But on the opportunity side of things at the moment, I think it's being piloted in most businesses across insurance. I don't think the insurance industry is at the forefront. I think there are other sectors that are moving more quickly, but
11:53
Unidentified speaker (speaker_2): within the sector, there certainly are some leaders that are investing heavily at the moment and learning. And I think that one of the challenges for the industry is that learning is expensive. It's what we're finding. But there are some use cases for it that are coming through very, very clearly. And it's also equally clear that the potential for AI is very significant. Certainly in
12:13
Unidentified speaker (speaker_2): terms of r removing a lot of the expensive resource taken up in tasks, but also vastly increasing uh velocity and speed to market in terms of product delivery and technology change. There are so many aspects. of any finance functions, provision of information, data, et cetera. So it's changing very rapidly, Matthew. And
12:33
Unidentified speaker (speaker_2): something that, as I say, there's a lot of learning going on, but there's also quite a lot of waste at this point, inevitably, as that learning happens. So I think the smarter businesses, if you like, are making sure that there is very targeted investment going on with clearer views to understand how that investment is paying back. And I think providing some sort of guardrails,
12:53
Unidentified speaker (speaker_2): if you like, in in terms of its use, you could very easily rack up some ferocious token costs if you let your entire organization loose and productivity won't necessarily follow. So, yeah, lots of learning going on.
13:06
Unidentified speaker (speaker_0): So let's just understand a little bit more. So it all makes complete in P sense. If you're going to let people use these tools, you've got to create some guardrails. You know, experiments are well known as the way of developing. new business ideas, but at some point someone has to say
13:17
Unidentified speaker (UNKNOWN): that
13:17
Unidentified speaker (speaker_0): experiment is not working, and experiments have to be designed so that you know they're not working. Are there any specifics, any examples just to sort of bring that to life of where you might have seen people either successfully or not successfully run those kind of experiments and sort of manage the cost in a designed way as opposed to let it sort of run a go road?
13:36
Unidentified speaker (speaker_2): Yeah, absolutely. So one of the portfolio businesses I work with are really pioneering this on quite a scale now, actually, and put a physical test, a classic development route through to release design and release of a new product versus a relatively young team of AI experts. So it set the same task in
13:55
Unidentified speaker (speaker_2): parallel to see what would happen. And actually, it was a very dramatic improvement in the delivery time for the AI team. I mean, transformational, probably to the power of 10. That's even taking account of the token cost. And then looking at the quality of the work that came out, actually the quality was higher on the AI produced work than the
14:15
Unidentified speaker (speaker_2): traditionally produced work. But that was a very specific and what you would call a relatively restrictive task. I've also seen circumstances, well, not personally, but heard through the insurance. world where an LLM was used to develop a new product in market. It passed all of the QA, UAT, et cetera, was released to
14:35
Unidentified speaker (speaker_2): customers, only to find out that there's some fundamental flaws in it. And when it was reviewed to understand what had gone wrong, the LLM had rewritten the hurdles that it had been set in terms of quality because it couldn't reach them all. So it's a bit of a shocker. So that business. that are now using a separate LLM to make sure that uh there's absolute
14:55
Unidentified speaker (speaker_2): integrity uh over the quality side of things. So there'll be some pitfalls. We talk about guardrails, not just for people using AI, but the way that AI is deployed in businesses. There's so much to learn.
15:04
Unidentified speaker (UNKNOWN): I think
15:05
Unidentified speaker (speaker_0): the takeaway from that is don't let your generative AI tool mark its own homework because it will change the rules.
15:14
Unidentified speaker (speaker_3): Exponential Risk London is back for twenty twenty seven. On the twenty third and twenty-fourth March, we're returning to The Mermaid, bringing together the insurance markets, catastrophe modeling, exposure management, and risk intelligence community. Across two days we'll explore emerging risks, advances in AI, and how better data and modeling can inform
15:34
Unidentified speaker (speaker_3): better decisions. Join us for expert insights, practical discussions, and opportunities to connect. with the wider market. Find out more and register at Instech.co.
15:43
Unidentified speaker (UNKNOWN): Mark,
15:43
Unidentified speaker (speaker_0): I've got a question for you, which I'm just sort of thinking about as Peter talks. As you said, you've got a background, including coding, that's not all you've done. You're kind of using AI. As you look at how you develop your products, is there something that shifts as people are thinking differently about how they should be using genitive AI tools with regards to prompts and capabilities? So you're a lot of people are talking
16:03
Unidentified speaker (speaker_0): about how they use AI to get better access to data at better speed. But I'm just wondering, is there a fundamental shift in how you expect people to actually work with your products because people are starting to think differently about the whole way you interact with technology?
16:15
Mark Varley: Yeah, there absolutely is, Matthew. We've had a diagram that's been on our website for a number of years, and if you go on our website today, you can see we've just we've effectively redrawn that. So historically, most of our customers would consume our data. and our services via either APIs or via a user interface. We've got an application stack where you can go in and visualize property or a portfolio on a map and see
16:35
Mark Varley: where the risks are, or you can hit our APIs for an individual property and be able to score them up that way. Really, the two emerging things we've seen over the last couple of years, and obviously AI specifically this year. So the third interface that we're seeing insurers wanting is the ability to be able to consume our data directly in their own data lakes. So a lot of our
16:55
Mark Varley: customers have invested in either Snowflake or Databricks. We actually interestingly see within insurance, it seems to be Databricks seems to have got the bigger foothold. A lot of our customers are using Databricks and we ourselves have migrated everything onto there. So being able to consume that data directly and then build that into their own pricing, train their own pricing models or
17:15
Mark Varley: create their own underwriting algorithms off a very large set of data is something we've seen. And then more recently, over the last six to nine months, as Peter mentioned, a lot of our customers are starting to experiment with Gen AI powered underwriting journeys. But what they recognize is that yes, they can plug Claude or ChatGPT can go out to the public domain and find
17:35
Mark Varley: out information about properties, but what it can't find out is that proprietor. proprietary, high-quality peril data that companies like ourselves, through our network of partners, people like JBA, for example, who've created a global flood model and UK, it's at five-meter resolution. Gen AI is not going to be able to go and find that in the public domain. So for us, it's provisioning that as
17:54
Mark Varley: a first-class part of our proposition, an AI-ready interface where customers can be able to plug that directly into their own. gen AI journeys. Now that they could do that with APIs, they could plug our APIs in, absolutely, and that would work. But for me, the opportunity and the technology, as you Matthew, you said, I'm a technologist by heart and I
18:14
Mark Varley: love all this stuff. But the Emperor is the technology that powers this. And what that allows us to do is given an opinionated view to our customers to say, this is the way we would recommend you call our services. And actually, what we can do is we can preload that with context. So not only understanding for this specific property what's its flood score, but what does that look
18:34
Mark Varley: like in combination with the rest of the UK, for example? So understanding not only what's the flood score or what's its property characteristics, but is it non standard? Is there something that's unusual about this property? Which we can provide that because we have that broader context. And I think that's the thing that's really exciting about this new interface.
18:53
Unidentified speaker (UNKNOWN): Great.
18:53
Unidentified speaker (speaker_0): And I had to look up MCP server. So I'm just going to quote from Google a secure program that connects artificial intelligence applications to external tools, databases, and data sources using a universal standard. And MCP stands for Model Context Protocol. I get that right, Mark.
19:07
Mark Varley: Yep, that's absolutely correct. And again, we've been a bit of a journey understanding. So initially we thought well all that is is a wrapper around our APIs, but it's as I mentioned, it's a huge amount more than that. And I think the power, I guess it's not dissimilar to the way we engage with customers at the moment. So a customer at the moment would onboard with us, and as part of that onboarding process, we'd understand their underwriting rules, we would work with them, we'd provide
19:27
Mark Varley: them with statistics and distributions and looking at whatever the perils that they're interested in. But actually, this allows us to go that bit further and and not only take that information and then create that in terms of a series of rag scores. but actually preload that customer specific context into this interface. So when they plug that and also as Peter mentioned guardrails, put the guardrails
19:47
Mark Varley: around it to say these are the things you should be doing, these are the things you shouldn't be doing. And here's some broader context and here's this customer specific underwriting rules. And again, we all we don't want to do is we don't want to step over that boundary to because underwriters need to underwrite but what we can do is provide them with a much more nuanced richer set of context to be able to make better decisions.
20:06
Unidentified speaker (UNKNOWN): Great.
20:06
Unidentified speaker (speaker_0): And Rag, I didn't need to look up that's red, amber, green score, just for those that are wondering. Peter, I just want to come back to you as we kind of get close to the end here. And it's a little bit on this, well it's on this learning question. But in your life as a formerly busy CEO, now busy chairman, what kind of advice do you have for somebody you know wants to learn about the technology so that as other people, whether they're an investment, whether it's
20:26
Unidentified speaker (speaker_0): Mark on his team, whether it's someone might be working for you, says, Hey, I want you to spend some money, or we need to spend some money on this technology. How do you, as a leader, build up enough context and experience so that you know both that the person that's talking to you about that is legitimate and is trustworthy? And to some extent, you understand enough about the technology to know that that's heading in the right direction without having
20:46
Unidentified speaker (speaker_0): to sort of fully understand
20:48
Unidentified speaker (UNKNOWN): it.
20:49
Unidentified speaker (UNKNOWN): Yes,
20:50
Unidentified speaker (speaker_2): good question, Matthew. There are a number of different points of what I would call either calibration or corroboration, whichever way you want to look at it. But I guess part of it comes through experience and just the experience I've gathered over the years in managing tech teams. Part of it comes through my network. And I use a strong network of people around me that I
21:09
Unidentified speaker (speaker_2): trust the the opinion of who stay very current. I You can't stay current across everything. So they're an expert. experts that I rely heavily on. And with the team themselves, the confidence that they give you. And certainly, as Mark said, he's a technologist as at heart, a number of the team are and very skilled ones. So I've got an awful lot of belief in team. And once they're able to explain to me in
21:29
Unidentified speaker (speaker_2): laminess terms often what's needed, it's a relatively straightforward process. That said, we've also got the benefit of Kestar, Kestler, sat behind as a private equity house, who have their own technologies. They're there as a sounding board if we want to bounce things around. But there's not been a case yet where I ever had to question Mark or Thompson's view, they're smart guys.
21:50
Unidentified speaker (speaker_0): And I'm glad you mentioned that point about network because yes, in my experience, one of the characteristics of a great chairman or even non-exec is that ability to access the network, have the network, yeah, both to sort of support the company you're working with, but to your point, Tavalda. to validate and ask questions. And I think all of us in our personal lives can think of circumstances where we reach into our network to get advice or validate something.
22:10
Unidentified speaker (speaker_0): Mark, I just want to
22:11
Unidentified speaker (UNKNOWN): have
22:11
Unidentified speaker (speaker_0): a look forward before we wrap up. What's going to happen next? What do you see addressed like going in the next five or ten years?
22:17
Mark Varley: Broader sort of mission we've been encapsulating recently is to be able to locate and describe every building on the planet. That's quite a bold ambition, but we're already well on the way on that journey. So as I mentioned earlier, we'd now have data coverage across mainland Europe, North America. We're going to be opening up a US office. So we're in the process at the moment of onboarding a US
22:37
Mark Varley: lead who is going to be managing things over there for us. So I think what you'll see over in five years' time is we'll still be answering those fundamental four questions that I mentioned at the outset, but we'll we would be doing that on a broader scale. We'll be doing that. through AI as well as APIs, data services, user interfaces, but the breadth and the richness of the
22:57
Mark Varley: data will be significant. I think also one of the things will be the speed. So we talked about speed of delivery earlier, but the speed of refreshing that data, bringing getting closer and closer and closer to real time, I would see that gap closing significantly over the next four to five years.
23:13
Unidentified speaker (UNKNOWN): Look,
23:13
Unidentified speaker (speaker_0): I love that aspiration about the building because ultimately analytics, technology, data is only as good as the source information you've got. And so if you don't know where the buildings are and what they're made of, then you're going to struggle with everything else. So I'm really intrigued to see how that follows up and looking forward to the next discussion around that. I've asked you a lot of questions. Is there anything for either of you that I haven't
23:33
Unidentified speaker (speaker_0): asked, we haven't talked about that you'd like to uh add in?
23:36
Unidentified speaker (speaker_2): Well, I think two things converging really. W one is the increasing need for the tools that Address Cloud are very sophisticated in providing. And at pre-bind stage of that's hugely important, whether it's the peril specific stuff, whether it's the accumulation piece that we're talking about and the speed that that we're able to do it with. That increasing need through weather events across the globe,
23:56
Unidentified speaker (speaker_2): or whether it's that converging with the appetite and investment support behind Address Cloud now that'll say it's expand across a number of different territories from here. So yeah, exciting times for the business.
24:08
Mark Varley: For me, Matthew, we haven't really discussed his. We've always been customer-led, and one of the things that we are seeing increasingly is customers not wanting to just procure data, plug it in, but also looking for us to help them understand And the return on the investment and the value of that data. So where we see ourselves going, is beyond just being a
24:28
Mark Varley: data provider, but being able to provide complementary analytics to be able to support that. So lots of our customers now are sharing with us their bound books, and we're using that to provide and power our accumulation service. We also capture through a usage of our APIs, we capture customers' quotations. So We understand what properties they're looking at, which ones
24:47
Mark Varley: they're there for selecting, which ones they're not selecting. So we believe that we're sitting on some potentially very powerful information that by bringing in analytics into the platform will be something that will give a dimension to the customer that they can't receive at the moment. And I think it would be a really compelling proposition for them.
25:04
Unidentified speaker (speaker_0): I really love that idea about the helping people figure out that total cost of ownership is sometimes called return on investment. Catastrophe modeling, which has been going now for at least 30 years. What really drove that was because really reinsurers in particular, but also insurers realized that ultimately they could underwrite more once they had a better understanding of their risks and they weren't making such sort of large estimates. And so really being able to
25:24
Unidentified speaker (speaker_0): help people how to help them drive more revenue is sort of always the story people like as opposed to telling more things that are going to go wrong if they don't use the tool. Really looking forward to seeing that. And then before we do finally wrap up, Peter, I'm going gonna to go for you first of all. Is there any specific podcast or book you'd recommend?
25:39
Unidentified speaker (UNKNOWN): I'll
25:40
Unidentified speaker (speaker_2): go for a podcast, that Empire, by uh the historian William Dalrymple and the reporter Anita Anand. It's a really good walk through all the different empires throughout our history. So, you know, whether it's the Ottoman or the Mongol or the British Empire. I didn't realise, for instance, that the British Empire was the largest empire in history by land landmass, mass,
26:00
Unidentified speaker (speaker_2): uh, for instance, but it tracks the sort of rise of and fall of each of these empires. It's amazing history. So it's well worth listening to. Well
26:07
Unidentified speaker (speaker_0): thank you for the nudge, because I think recommended before and it's sitting in my playlist, but so far the rest of history has been winning on the historic one. So I'll definitely give it empire a go. Mark, how about you recommendation?
26:17
Mark Varley: One of each actually so we are at the moment obviously scaling up a business as a founder and having done it for the first time can be quite scary, quite daunting. So we're following a methodology by a chap called Vern Harnish called Scaling Up. We've got a coach, Coach Kevin, who's helping us through that at the moment. And that's been both interesting, informate informative, and quite liberating. So
26:37
Mark Varley: when I first started the business, a podcast that I would listen to religiously was called Cloudcast at the time. And obviously, we're addressed cloud, we understanding, and we were cloud architect from the outset. They've actually pivoted that about six months ago and they've now re-badged. themselves as the Enterprise AI show. Again, that's a weekly, it's a US-based weekly podcast. Super interesting,
26:57
Mark Varley: loads of great insight for people who are either sitting within large businesses or, you know, like me, a founder running a team. That's a great listen.
27:05
Unidentified speaker (speaker_0): The scaling up one is really interesting. I remember a quote which I found it very helpful and real from Reid Hoffman about how your business completely changes. You go from one person to three to ten.
27:15
Unidentified speaker (UNKNOWN): to
27:15
Unidentified speaker (speaker_0): thirty to a hundred, I think it was. You're probably even beyond that now, Mark. Or will be soon. Thanks both of you for this. It's uh really appreciate your support for Instac. It'd been great, Mark, to watch
27:24
Unidentified speaker (UNKNOWN): what
27:25
Unidentified speaker (speaker_0): you've been doing AdressCloud over the years and congratulations on the investment. There's not many people who actually succeed when they find companies in reality. So we hear the headlines, but there's a lot of people that struggle to get there. So that is a really great achievement. And Peter, we've we enjoyed meeting you. and look forward to talking again in a year or two to see how we've we've got on.
27:42
Unidentified speaker (speaker_2): Yeah, thanks very much, Matthew.
27:43
Unidentified speaker (UNKNOWN): Uh
27:44
Mark Varley: thank you, Matthew.
27:50
Unidentified speaker (speaker_3): Well, if you've made it this far, then I'm pretty sure you found that as interesting as I did. The InStech podcast comes out every Sunday morning where we spotlight the latest news, leading voices, and freshest updates across insurance that you need to know about. If you would like to take part in these conversations,
28:06
Unidentified speaker (UNKNOWN): head to
28:07
Unidentified speaker (speaker_3): www.instech.co to
28:09
Unidentified speaker (UNKNOWN): find out how you can join our
28:11
Unidentified speaker (speaker_3): network
28:11
Unidentified speaker (UNKNOWN): and be a part of the Insurance Intelligence for the
28:13
Unidentified speaker (speaker_3): Curious.
This page was produced automatically: the audio was transcribed with nvidia/parakeet-tdt-0.6b-v3+sortformer-v2.1 and interpreted by Qwen3.6-35B-A3B-UD-IQ3_S.gguf. Transcription and interpretation can contain errors; always confirm against the original episode. Nothing here is investment advice.
All insights