Verisk Analytics, Inc. ($VRSK)

Good afternoon, everyone. Thanks so much for joining us today. My name is Kelsey Zhu. I'm the information services analyst at Autonomous. With me on stage today, I'm very pleased to welcome Lee Shavel, the CEO of Verisk. Thanks so much for joining us today. Lee. Really appreciate it.

Thanks for having us, Kelsey. It's great to be here.

So AI has probably been the hottest topic within the information services sector this year, and I think that's a good place to start. And I know Lee, you recently launched your MCP server. So maybe talk us through a little bit more about the rationale behind that launch. What data is included in the MCP server? How do you protect your data? How do you protect your customers' data and the pricing and the target audience for this MCP?

So we're just going to dive right in. So let me start, and I will get to that. Kelsey has shared that we may have some generalists in the room. And so I just want to provide a very brief overview. Verisk started out as a utility to the U.S. property and casualty insurance industry. And we were there to serve a role in collecting data, collecting regulatory filings, managing those on the part of the industry. And it gave us access to an extraordinary wealth of data, particularly loss costs and the ability to standardize forms that enabled us to help the industry understand loss experience across individual product lines and also to develop an intermediary relationship with regulators. And that as we began to apply more data science and analytics approaches to the business, we were able to build on that and acquire and develop other businesses such as our catastrophic risk modeling business that models hurricanes, severe convective storms, wildfires, floods, we also built claims oriented businesses that capture potential fraud risk in claims and allows carriers to estimate losses from claims that are filed with them, all of which serve the industry and importantly, in a classic utility fashion, we have the ability to invest in a data set, invest in an analytic or a technology that we can rapidly scale across the industry. And the industry benefits from that because they get the value without all individually thousands of insurance -- individual insurance companies without having to make that investment. And so that dynamic is still kind of critical to what we do and recognized by our carrier clients for the value that we can bring to the industry. And so with that, we have been working with AI in classic forms for at least a decade. We've developed procedures and policies to make certain that we are deploying AI with responsibility, with protections for the data, for protections for privacy and generative AI has come along in the last few years, and certainly captured the kind of collective imagination. And we have been working to deploy generative AI and large language models into our data sets. So this is not something that is new to us. We have developed products that support our clients but we've recognized that with the frontier AI model companies that there is a moment where many of our clients are interested in utilizing those models in their work. And so we were connected to [ Ampropic ] as they were developing their insurance strategy, and they recognize that we were a unique and valuable player in the insurance markets, and they said we would really love to develop 2 use cases or connectors to your business. And we went through in the course of about 4 weeks, over a dozen potential applications, some of which may follow these initial 2, but we came up with 2 connectors that connected the cloud models to an underwriting data set for our forms, rules and loss costs that allows the model to interrogate the data, and it's important to understand that this doesn't give access to the underlying data, but it allows the model to query that data in the same way that a subscriber, one of our client subscribers might gather data on loss costs by -- on a state-by-state basis or a business line by business line basis and accelerate their ability to pull that data, understand the analysis to inform their underwriting decisions. In addition, we have also developed a connector use case that connects the cloud model to our restoration analytic, which allows a restoration professional to interrogate our materials and our labor costs to evaluate if I need to do a restoration on a kitchen or on a garage, what might be involved, what level of quality am I looking for so that it can come up with an estimate for that renovation or that repair very quickly. So these are natural language interfaces that much more efficiently gather and connect that data set. That we have seen in both cases, demonstrated at a micro level, the ability to get that information faster to get to a more connected analytics associated with it, and we believe will create value to the industry as a whole. Now importantly, there's nothing exclusive about this. We have clients, some of which are -- have cloud licenses, some of which have chat GPT licenses co-pilot licenses. And so we do think this is a broad opportunity for us to show how our data sets enable and deliver much more tangible value than simply buying a generalized cloud license. And a lot of our clients have had the mandate to do that, and they have told us, look, we -- this is interesting, but we really need to demonstrate where there is economic value, and that means tying it and interacting with the data sets and with the workflows that are out there. So we're excited about what we accomplished in a short period of time. It builds upon the knowledge that we've developed in AI, and we believe it's just the start for, I think, what we can do for the industry and our classic utility role.

Got it. That's a super helpful summary. I guess one of the main investor debates is really in the future, what kind of role do you think large language models would play in the insurance industry? Are you thinking of them, the cloud, open AI of the world? Are you thinking of them as your incremental distribution channel? Are you thinking of them as potential competitors or maybe partners for codeveloping some of the models and analytic solution that you will introduce in the future?

Yes. Kelsey, I think those are 3 -- we think about those 3 alternatives. I think our view is that this is clearly a distribution channel for us and the data sets. And it is -- I think that view is bolstered by the fact that we have this happen repeatedly where there is a new technology, for instance, cloud data storage or SaaS computing applications where our data has been critical from an industry perspective and integrating it into those solutions. So as they were adopting cloud storage, the ability for them -- for our clients to deliver data through that technology and access it became a critical function. With many SaaS solutions, particularly policy administration systems, many of them rely on our underlying ratings data, and so that became an ability for us to support the value that they were creating with those technologies and to find new means for us to monetize the value of the industry as a whole. And so when we think about the value opportunity for us, I think we feel, based upon the micro experience that we've had and extrapolating to a macro view, if AI can improve underwriting efficiency and accuracy by 5% to 10% and claims efficiency and effectiveness by 5% to 10%. Across an industry with a massive expense base dedicated to those functions, that's a substantial amount of value. And our ability to participate and help facilitate that value is the opportunity that we'll pursue. So I think, first and foremost, a distribution channel. We've also seen that as our most sophisticated clients increase AI utilization, they want to use more of our data sets. So the number of products that we're able to sell as we've talked about at Investor Day, which has shown a steady progression and a correlation to the size and scale of our clients will benefit. But the third point is they, I think, will also be effective partners where we're able to bring together the AI element with the data element and the process element to create a true intelligence layer that facilitates overall industry functionality. And so I think those are the 2 primary channels. We don't view this as competitive to what we do. The work that we do to gather data sets to normalize them from an actuarial standpoint, to cleanse them to integrate them with other data sets. -- is a very unique proposition. We have contributory data sets that can't be accessed publicly that our clients are very sensitive to the use of that data. And so in many ways, we're as much a custodian as we are a user of those data sets on behalf of the industry and good AI requires high-quality data sets, and we're happy to continue to play that role.

Yes. Well said. I guess one of the main points you mentioned is AI implementation at client side does lead to increased data consumption for [ Verus ]. My other question on this is just in general, when you look at insurance carriers, what kind of AI investment strategy do they have? Like are they mostly focused on top line opportunities, cost-saving opportunities? Or what are you seeing across the board?

From our experience, they are overwhelmingly focused on efficiency and cost savings opportunities. I think that's the first priority. I think the second priority is portfolio optimization and risk selection applications, although I think that is a step behind or a next cycle element. And the third is customer interaction. And how do they -- certainly, the use of AI to facilitate more efficient customer interactions. But when you look at the use cases, that our clients are most interested in. It's how do I improve the speed and the productivity of our underwriting teams or our claims teams, first and foremost, and then kind of more interest in how do we use this technology to evaluate our overall risk portfolio preferences and management.

Got it. And when you think through that 5% to 10% productivity uplift whether in claims or underwriting, I guess, my thinking on this is like what are the regulators saying about the usage of AI in underwriting or claims processing, -- and how do you see the regulatory landscape will change in the next 5 years?

Yes. You're absolutely right. And I think it's important to say at the outset on the -- on any of the use of AI, the insurance industry is highly regulated. It's regulated in a very fragmented fashion on a state-by-state basis. So it's going to be complex. It's going to evolve. I can tell you from first-hand experience and talking to a number of state insurance commissioners who are our clients and that we work with and engagement that we've had with the NAIC. They're very much aware of concerned about AI. And I think while there is an orientation to find ways for the technology to provide value, they are certainly going to be focused primarily on protecting the consumer. And I think in some ways, the technology may be helpful in that regard in standardizing treatment from both an underwriting and from a claims standpoint. One challenge that the regulators have or are focused on is the fact that generative AI, by its nature, as kind of a predictive tool that is trained -- a model is trained to predict the next token in a response to a query is not -- it's not predictable, it's not auditable, you don't know what goes into that. And so in that context, I think the regulators are always going to be opposed to taking the human out of the loop because you don't know -- you're not going to have consistency and you're not going to know how that decision was based -- and it's a fundamental regulatory principle, there has to be a justification for pricing or how a claim is handled. I won't go into great detail, but it's been interesting to us that the regulators have been focused on a form of AI known as neuro symbolic, which is a more determinative AI application that limits or puts a constraint based upon domain knowledge to make certain that it is a more -- it's a more transparent and predictable and consistent outcome. We're actually working with a client who is deploying neuro-symbolic AI, and utilizing our data sets to pursue that. And it's something that we are having conversations with regulator to pursue. So that's the -- I think what we're experiencing from a regulatory environment. They realize that it will be an element, but they're going to want to make certain that it's predictable and transparent.

Got it. And just to clarify, is AI allowed to be used as part of underwriting decisions today? Or is It is not?

Well, it is allowed to be utilized to support decisions, but ultimately, the individual or the company is going to be responsible for whatever is submitted from a regulatory standpoint.

Got it. Very helpful. The other question I want to get your thoughts on is really this build versus buy question that I think all the carriers are trying to think through. So maybe there's a difference in terms of larger, more sophisticated customers who have more resources, how they're thinking about building new AI solution versus small to medium-sized carriers. And I was wondering from your dialogues with clients. Are they mostly looking at internally developed solutions? Are they looking to partner with Verisk? Are they looking at AI start-ups? Or do they even have any appetite to work with start-ups? Or what are you seeing in terms of the build versus buy?

So it really varies dramatically. We have the benefit of being able to see across a wide range of scale of players, some of the largest insurance companies, those whose brands are well known to smaller regional players. At the high end, I think there is a sense of responsibility to have ownership of the technology to explore it and determine how they can apply it with their data sets, but in almost all of those cases, the clients in that category have wanted a specific dialogue and conversation with us to talk about how we can integrate our data sets into their strategy. What are we doing that they might learn from or leverage. And there are, in many cases, where elements of -- how we're in that they feel they can leverage that, for instance, taking our ratings and building an agent that rates a policy for them to evaluate rather than them investing in an independent ratings element or a ratings agent associated with that. But they clearly want to drive that deployment of technology and evaluating it. We also have clients that have asked us to work with them in developing a technology platform we spoke on our call about a very well-recognized and respected insurance company that has asked us to build an a genic underwriting platform based upon our knowledge of the data sets, the regulatory needs and the technological dimension. So we're excited about that dimension. I mentioned the neuro-symbolic client that we're partnering with. But I think there is also a wide range of clients that realize that they may not have the resources or the expertise to develop their own customized installation. And in that case, some of the products that we've developed like our underwriting assistant or in some of our claims areas, our AI solutions for that enable or apply an AI technology may be more than sufficient to meet their needs. So I think we're trying to meet our clients where they are at all levels of sophistication, and we have the flexibility to be able to do it. We also -- I think it's probably not recognized, we have an intense level of engagement with our clients functional level. And so we are truly a partner in understanding what they want to accomplish with their data strategy and their technology strategy, and it's a very regular part of our dialogue to understand what they're doing and what they're considering. You also asked Kelsey about their inclination to work with AI start-up. And I -- our experience has been -- they're often interested in what may be accomplished at some of those levels, but there is a general reluctance to scale off of a small emerging company because of the uncertainty around where that company is going to end up or what they're going to do. And I think that's been a challenge for a lot of the startups. We have found in instances where if there's a good technology that's been proven, we often have the ability to monetize that if it's an acquisition to accelerate that and to build on our strengths. But in a lot of cases, it's very difficult for them to achieve the scale necessary to convince a large insurance company to adopt them.

Makes sense. And where are all of these AI start-ups concentrated? Are they mostly focused on collecting incremental data points? Or are they focused model development? Or how are you thinking about the landscape?

Yes. So what we have seen, we've seen a real concentration in workflow functionality and which is interesting because those applications still require a lot of data. And so they're going to often draw data sets that our clients have or that we can provide to automate a component of the functionality in an underwriting process -- it may be data ingestion, but I think in many cases, it's an integration of data and a reasoning function that produces an outcome that then would be reviewed by a human underwriter or a human claims professional to evaluate. So I think a lot of micro functions that might be built into as components of an overall Agentic strategy. That's been the primary focus that we have seen.

Got it. So it's more like the layered analytics layer above Verisk data for insurance carriers data.

Absolutely. I'm focusing on claims functions, underwriting functions, risk assessment functions.

Got it. And how about Lee, how are you thinking about the build versus buy conversation for Verisk?

So when we look at the components, certainly, we've made the decision we're not going to build a large language model. An we don't think that it's necessary for us to provide that. There are plenty of good models. In fact, we have worked with almost all of the models, in some cases, using a combination of them in some of our applications. I think what we are dedicating time to is working with individual clients that have a clear use case that they want to pursue, where they feel we can support or develop that for them. And then if we prove the functionality of that, then it gives us the ability to roll that out to the industry as a whole. So I would probably describe it as our focus is around applied technology as opposed to technology and integrating our data sets with that applied technology approach to meet our clients and then work to scale that across the industry.

Got it. Very clear. So something I've been thinking about is Verisk has actually started talking about deploying AI in different models and analytics from 10 years ago. So over the last 10 years, I guess, what are some of the key initiatives that you are most excited about, looking for in the future? What kind of AI investment strategy, we should describe that Verisk has, and how should we think about the top line and margin opportunities from AI?

Yes. So I think there are a couple of levels to where I get excited about the opportunity. The first is that basic level that I think as clients are using more AI [indiscernible] for data will increase, I think the value of the data improves. And so I think it is the distribution model where we benefit from that. The second element is kind of more specific to the industry is how do we improve the overall productivity in the underwriting function or the claims function. And this is where concept that Anthropic has talked about, which is the diffusion of the technology into the enterprise is a limiting factor at the rate of which is a limiting factor. And I think we'll see a slow absorption of the technology that will produce improving productivity over time, but which will have a significant benefit to the industry in terms of that efficiency of the underwriting, the accuracy of the underwriting, the ability to assess risks more effectively and to manage claims. And that from a scale level, while I think an intermediate-term opportunity, I think, is the largest -- the largest foreseeable commercial opportunity for us to pursue. We've shown in a variety of instances where we can help deliver value to the industry, we have an opportunity to participate in that value. And because of our scale and utility model, our ability to make that investment and then deliver to the industry tends to enable us to produce very high value to cost equations, which encourages it. So that's level 2 that longer-term the intermediate-term opportunity to improve process. And I think the third intermediate longer-term opportunity is the ability for AI to integrate data sets across disciplines and find connections or new products that are going to be additive to what we do already. So this isn't just improving an existing underwriting process or a claims process. But improving the overall -- the overall economic enterprise of insurance. And one use case that I would point to is the ability to connect the life insurance life cycle and manage risk more actively. Certainly, insurance is a marketplace for risk and pricing risk. We have seen led by the reinsurers a more active engagement in managing that risk actively either in selling reinsurance contracts or laying it off in the insurance-linked marketplace. And we think that as the ability to manage real-time exposure data, underwriting information and claims information, -- we're going to have the ability to help the carriers really benefit that Chief Risk Officer or Chief Portfolio Manager within an insurance context to manage that risk more actively. In many ways, we're setting up synergy studio to be that platform to serve that cheap risk officer or risk portfolio officer. And that's kind of an incremental opportunity that we think the industry will benefit from.

We'll come back to Synergy Studio in a second. But 2 quick follow-up questions to that. First one being, have you thought about sizing the incremental revenue or margin opportunity in the medium term? And secondly, you mentioned this distribution of AI technology within each vertical. Could you just elaborate a little bit more about that? Is that just like getting all of your employees familiar with using AI? Or what has got?

Yes. So it is actually, again, 2 that. I think the question that you're asking is, when I refer to the diffusion of the technology, it means the adoption and the integration of that AI technology across the enterprise in terms of the number of employees that are utilizing it. And I think we're at a very early stage in a very limited stage simply having a cloud license can certainly experiment with the model. But until you tie it to a function and a workflow in the data set, the value creation is probably more limited, but that demonstrates it. And I think different industries will have different diffusion rates into that. I think insurance will probably take a little bit longer because of the regulatory process because I think natural caution around new technologies. But that diffusion rate, I think, is going to make this a longer adoption rate to scale of the technology within the industry. We saw it in cloud computing. Initially, insurers were very cautious about the movement of their data from mainframes into the cloud. But once some pioneers had demonstrated, we were one of the first to move a lot of -- all of our data sets into the cloud. And when the efficiencies and the benefits of associating that data became clear, I think the industry followed us and really realize the benefits of it. I think we'll see something similar in AI.

Got it. So Lee, when you look at various portfolio across different segments and with AI, there will be new competitors or new solutions in the marketplace. Which product category do you feel that Verisk has the most -- the best defensible competitive mode over the long term versus having more room for improvements?

Sure. So we understand that's a question that's on investors' minds for a lot of companies. I think from our perspective, as we talked about at Investor Day, we believe that 90% of our revenues are tied to proprietary data sets, contributory data sets, proprietary intellectual property such as our catastrophe models. And we feel very comfortable that those data sets not only are defensible, but they're critical in the realization of value from the application of AI technology, which requires good data sets specific to an industry functionality to do that. We are mindful that there's a lot of concern about the exposure of some software businesses, particularly horizontal software businesses to the ability to code more quickly or to develop internal applications. We do have some independent software businesses, but we believe they're very vertically anchored. They're tied to proprietary data sets in many instances. And so we think are defensible on that basis. We've also deployed lighter architecture to them. We're using low code, no code, which is more conducive to flexible data applications. So we're watching our exposure within those businesses. But to date, we haven't [indiscernible] and we believe that that's -- we're still in a competitive position, but that's one area that we watch.

Got it. I guess in my discussion with investors, for issues loss costs, anti-fraud, property estimating solutions and catastrophe risk models are probably the most defensible segments of Verisk, but there is one segment that I do want to dig a little bit deeper into, which is underwriting data analytics. Maybe just talk us through and remind us why Verisk competitive moat is really strong in that segment.

Yes. So in our underwriting data and analytics solutions business, we have a business called Prometric, which collects commercial building. So in this building, this hotel, we have a very insurance-driven data set that focuses on the engineering, the fire suppression systems, the ground floor use of the business, the general application of the business, everything that a carrier is going to want to know about ensuring this property or general liability associated with businesses in this business. Now that data set is gathered through a field force of individuals that survey the property, gather know what information to gather, and we have expanded the data sets that are relevant and included in that database. This information isn't publicly available. It's acquired by insurers. We're again acting on behalf of insurers. So we have the efficiency to gather this data once on this building. And that way, any insurer that wants to have access to it can do it more efficiently than sending their own team to do diligence the building. So we believe that that's unique proprietary data that we're able to build. We also have in underwriting data and analytics solutions our 360 value, which is a replacement cost analysis. So for an underwriting function, what would it cost to rebuild this building. That is built upon proprietary information and intellectual property on materials costs, labor costs and what's required to rebuild that business. And it's used as an industry standard for underwriting purposes within the industry. And then we also have contributory data on a auto-related information driver history, policy information. That's a business where we have faced competition as we've spoken to in our calls from another player in the industry. it's one of the businesses where we don't have as strong a competitive advantage, but we provide an alternative to the industry as a whole. So I think in the first 2, we believe we still have a very strong proprietary dimension of the data set. And then the third one, we are -- we have been working to build competitive advantage part of what we are doing is trying to leverage some of the data sets that we have in other personal lines businesses to provide more insight around the underwriting -- the auto underwriting function.

Got it. That's super helpful. I guess the other question on investors' mind is really just ferrous pricing power in the medium term as carriers kind of rethink about their allocation for data and analytics budget. A percentage of that probably goes to newer AI-powered solutions. So just help us think through various pricing power over the next few years.

Yes. So we tend to think about our pricing as being value driven. And so we want to make certain that given our relationship with the industry, that we are delivering incremental value to our clients relative to what their pricing increases are -- and I think as great evidence of that, we have, as we talked about in our first quarter call, the fact that we have seen a very consistent trend of renewal of our multiyear contracts at the same or longer terms, than had existed before as well as with attractive pricing improvements that recognize the value we've created from investments that we've made in access and usability of the data sets, the frequency of our updates, which our clients have asked for as well as increased insights that we're able to provide our clients on a product line and a state-by-state basis. So that's our starting point from a value perspective. The second point I'd make is that we don't think that AI and data are mutually exclusive expenses. And in fact, in order to generate value out of we believe that our data sets are going to be required to make the AI investment effective. And so I think rather than one displacing, I am very much of the view that AI is actually going to expand the value that our data can provide and expand the utilization of those data sets, both from a quantity of data as well as a leader that we are providing. So I think that's where we see the opportunity. And then finally, it's that quantum of what do we think can be achievable when you combine AI with high-quality data and integration into workflows. And that's where I think when we look at the breadth of what we've experienced on a micro level, on a function-by-function basis with the anthropic application, we believe that our opportunity is to deliver that value to the industry. and to participate in sharing in a portion of that value that we're creating. And we think that could be a meaningful new growth opportunity for us.

Got it. The other hot topic in the insurance industry is really the transitioning from hard to soft markets. From -- I guess, from your seat, are you seeing anything that's fundamentally different about this cycle versus historical cycles? And how is that impacting Verisk?

Yes. I don't think that we're seeing anything. It's -- I think we're still early in this cycle, but I don't think that we're seeing anything in terms of severity or breadth -- this is, I think, a natural cycle where we had a strong hard market that was fueled in some part by inflationary dimensions, some catastrophic losses. We've had a very mild weather impact, no hurricane that made landfall last year. And so that tends to have a softening effect on the overall market. But all of this is very consistent with what we have seen in the past. We've talked about historically that when we've looked at the variation between hard markets and soft markets, we've seen our organic revenue growth vary between 7.2% in hard markets and -- or 7.3% in hard markets, 6.8% in soft markets, it still rounds to 7% and we think doesn't capture the fact that what drives our growth is the increasing adoption of our data and technology in the insurance space.

Got it. So coming back to Verisk Synergy, Studio. I think you're launching this new product next month. Maybe just talk us through the pricing strategy, target audience, how is this new product different from legacy models and things like that.

Yes. So the intention with Synergy Studio is to move what had formerly been a client hosted or a Verisk hosted modeling environment into a cloud-based SaaS solution that enabled us to deliver our models much more quickly and update them. So again, we're addressing the currency of the tools that we're providing to the insurance industry to apply all of our cat models to a consistent economic model. And this is where I want to explain, catastrophe modeling is not weather prediction. It is taking a view of weather variability and applying them to physical structures and insured assets and estimating what the potential loss is going to be. And so our ability now on a SaaS-based platform to integrate or to tie together a massive amount of exposure accumulation and subject them to a wide range of natural catastrophe risks so that our carriers can have a more accurate outcome, a faster outcome, the ability to apply more robust stress testing against that and get more current and updated models is the value proposition. We've also opened up our model -- our ecosystem to invite other models to participate so that our carriers can utilize a range of models in areas where we may not be offering a model. So it is, I think, a more technologically robust product set. It expands the functionality of what we are doing, and it opens up that ecosystem, we believe delivering more substantial value.

And I guess the other exciting transformation initiative at Verisk is really the core lines reimagined program. So maybe tell us more about where we are in the core lines remain program? And are you seeing this program does actually delivering tangible benefit in terms of pricing or customer retention and so on and so forth.

Sure. So I think we are -- we have accomplished everything that we set out to. I certainly think there will be an ongoing effort to take client input and integrate that in a reoffering to enhance that -- but in terms of digitizing much of the data that we provide, the ability to interact with our forms gather data through APIs, that infrastructure has been established. We'll look to continue to enhance and refine that. But the feedback from clients is even stronger than we'd hoped for. That's translated into consistently stronger renewal experiences and more engagement from our clients and our ability to help them utilize and get more value out of the data. And I would say what I'm excited about is that now having that infrastructure a lot of which is API-driven, its readiness for AI and the ability to apply generative AI large language models to those data sets through MCPs for which the 2 initial anthropic connectors are just to start in terms of what we can do -- and we are just pricing that on a free for a trial basis. But I think our view is that there's going to be real value that we'll be able to price to over time. Now that has created a great foundation for us then to demonstrate value from AI investments with our client -- with our client base.

Got it. I guess, switching gears a little bit. Verisk is obviously a very important data and analytics provider in the P&C insurance industry. But are you looking outside of P&C insurance industry for some of the future growth initiatives that you're thinking about, whether this is international markets, the life insurance vertical or anything else that you're excited about?

Yes. So I'd say all of the above, Kelsey. The Certainly, there's a lot to do in P&C. And I'll give as one example, which we've been thrilled at the results from is A lot of our loss cost data came from what's known as the admitted market, which is the regulated market where you have to go to state regulators. But there is what's also known as the excess and surplus market. And we've had clients come and say we'd like to have more data sets from you, and we're -- we would like to contribute data sets so that we can evaluate our excess and surplus writing, which has been growing at a faster rate than the admitted market. And so we've been thrilled to be able to get more data contribution on that front. So that's an example of just within the P&C marketplace where we're doing more. We also have been doing a lot in the specialty insurance market, the Lloyd's market in London, and have developed a transactional platform called white space that has received high adoption for facilitating greater efficiency [indiscernible] and underwriting by underwriters within the London market. That's an example of both P&C expansion, but also in international market development. And as we talked about at Investor Day, we have demand to roll that out in the Middle East and in Asia, and specifically the Singapore market, which we think will be a natural expansion. And we also see the opportunity to bring a lot of the technology that we've deployed within the U.S. and find solutions within a European -- within a European content. And then finally, we've had great success in the life business with our fast element. We've added a regulatory component to that with the acquisition of Assurance Bay. And we've also had clients interested in how we might deploy AI against the data sets that we're collecting in the life side as well. So I think we feel very comfortable with multiple vectors from off of our very strong base in U.S. P&C.

Speaking of the [indiscernible] Bay acquisition and the FaaS acquisition, maybe talk us through a little bit more about what your product road map road map may look like in the life insurance vertical in the next 5 to 10 years.

Sure. So starting with Fast. Fast as a policy administration platform that has replaced a lot of legacy systems and has enabled our clients to dramatically accelerate their product development functionality additional players and life and annuity -- a private equity life and annuity acquirers that want to have a system that is more responsive, much less expensive to maintain. So that's the starting point. And then from that platform, Assurance Bay is a natural regulatory function to make certain that, that brokers are meeting regulatory requirements and how they are describing the policies and the language and getting through all the necessary disclosures. And what we have heard from our clients that have been really happy with that -- those implementations is how do we integrate more life cycle analytics to help them evaluate the maintenance and the renewal cycles for their life products. So I think the life insurance industry has been more actively engaged in thinking about data and analytics whereas before, they were probably more focused on basic mortality tables and kind of annuity calculations, and we're seeing a much more sophisticated demand for data and analytics to help them better manage those long-term relationships that they have.

Got it. Kind of switching gears to a more near-term question. So for FY '26, obviously, in the half of the year, you've had a number of short-term or near-term headwinds. You're guiding for a second half reacceleration of growth. Maybe talk us through what are some of the key drivers of that second half acceleration? And how are you thinking about the full year guidance of 7% type organic revenue growth and the puts and takes to that guidance?

Yes. So I think it's important to understand that the -- while we talk about a -- you describe it as a reacceleration. Our underlying business as evidenced by kind of the strong subscription growth has maintained a very steady and strong growth rate. We had the impact of some weaker weather -- low weather activity, we had a government contract, which was impacted by some policy changes within the organization. And we had some other headwinds that in the second half of 2025, we had to work through. As we've described, those are migrating out of the system. We believe we had the trough in the first quarter. We're expecting continued momentum into the second quarter -- and then the third and the fourth quarter, based upon the guidance that we have described, should show a recovery to our 2026 overall growth rate. But to us, it's less a reacceleration. Well, that's what's happening on a net basis to me, it's more of a natural revelation of the underlying growth the business as a whole as a couple of those exogenous pressures that we had in 2025 work their way through.

Got it. We have 2 minutes left. Maybe 1 last question from me today. How are you thinking about your capital allocation priorities, especially in the current environment where your own AI initiatives probably brings so many new investment opportunities and you're probably thinking through organic investment, inorganic investment. So how do you think about your strategic M&As, your buybacks, your debt prepayments, your dividends and so on and so forth. And specifically for strategic M&As, what kind of new data assets or models or analytics solutions are you looking to acquire?

Sure. So our first priority by far is where we can invest organically in integrating AI for our customers into our data sets, creating value in that broad underwriting or claims function. And I think we're anticipating more intensive investment given the rate of change. You've seen our level of engagement, as I've described, both with Anthropic, our dialogue with other frontier model companies. our exploration of the opportunities in neuro symbolic, project on the agentic AI element. So all of those are top priorities to determine where our greatest monetization opportunity is for AI in association with our data. I think secondly, continued investment in maintaining the strength of our data sets. So investing in data, improving its currency, its relevance, quality, expanding new data sets like our excess and surplus data elements would be a second internal priority A third priority is how do we strengthen the ecosystem and the network dimension of what we are doing to create more efficiency and data sharing across claims ecosystems, fraud ecosystems, underwriting and catastrophe and risk modeling. So there's a lot to do within each of those areas to strengthen our existing business and develop new growth opportunities in AI. Fortunately, we generate a lot of capital. And so we have plenty to fund that, but I think we might see a marginal shift in that internal investment consistent with our overall guidance. So nothing that changes on that front, but I think we want to take advantage of this opportunity that we've seen. And then Secondly, we will look for value-creating opportunities where we believe that there's a business that is delivering value for the industry, but we can rapidly scale it or enhance its effectiveness with our data set. That will be a second priority if the return characteristics are attractive to us. And then finally, naturally, given the pressure that the overall sector has experienced -- we have been more aggressive as investors have seen in terms of taking advantage of that from a share repurchase standpoint. And where we don't see opportunities to invest in good returns, we have a natural inclination to return that through both increasing the dividend and repurchasing shares.

Got it. This is all super insightful and helpful. Thank you so much for sharing with us today. Really appreciate this, Lee.

Thank you, Kelsey. It's great to be here. Thanks for the great questions.

Thanks, everyone, for joining us.