On Risk Webinar Replay: AI in Insurance (Infrastructure)

00:02, Adam Klauber

Good morning, everyone. Thank you for joining. If you don't know me, I'm Adam Klauber. I run our insurance group here on the equity side.

We've got a really, really good call lined up, so again, thanks for joining. The topic is AI infrastructure, meaning, you know, we're just going to get into the guts of what insurance companies are doing with AI.

And, you know, very importantly, we're going to try and, you know, really give you a bird's eye view of what's happened today and, you know, what's actually happening on the ground.

We've got two really good guest speakers to help us figure out what's going on. We've got David Mocklow, who's been around the insurance industry in a number of leadership roles for a good number of years. David is running the insurance group at Growth Protocols. He will tell you more about that.

And then we've got Rich Drab, again, a long-time insurance and technology executive who has had a lot of leadership roles within his career. And currently, and for a while, Rich has been a senior consultant at Capgemini, one of the partners in insurance practice.

So, from a format standpoint, I'm going to have each of the speakers talk a bit about themselves. Then we'll do a good discussion on what's going on around AI for around half an hour, try and keep it short and punchy.

And then we will leave some time if you have some questions, please put them in the chat function. So with that, Rich, why don't you start off telling a bit about yourself and about what you guys are doing at Capgemini?

01:41, Richard Drab

Sure. Thank you, Adam.

Hello everyone. I'll keep it brief, but, yeah, just a quick introduction.

Rich Drab. I live in the Chicago area. I'm an executive at Capgemini and our consulting practice, as Adam mentioned. So what does that mean? I work with many of our top insurance clients around business transformation, technology initiatives, and, as you can imagine, those that relate to AI are very prevalent right now. So really on the front lines of helping organizations in the insurance industry navigate that and, you know, expand their capabilities through AI and other technologies.

My background has been primarily consulting in the insurance space for decades, a little bit of a stint in the retail consumer products industry for a while, which was interesting because that actually, at the time, retail was much more advanced as far as digital interaction with the customer and other things that insurance was a little bit behind on. So, it was great to bring that back to insurance.

I stepped out of consulting for a while when a former client in Peter Serafino asked me to come to Marsh, and I was the CEO of their insurtech, kind of technology-enabled MGA that ran all of their small commercial business. Dovetail Insurance was the name of that. And they had asked me to come and, kind of, run that for a while.

I've been back in consulting for a number of years now. Som it's great to meet you all. Thanks for having me, Adam.

03:16, Adam Klauber

Great, great. David, you want to tell us a bit about Growth Protocol?

03:21, David Mocklow

Good morning, everyone. I'm delighted to be here. I am David Mocklow. I'm the head of insurance for Growth Protocol. I was a former colleague of Adam's at previous investment bank.

Some of the folks on the call may actually know me. I'm a little bit of a unicorn in the insurance industry.

I started my career in underwriting with Ace many years ago. In fact, I was the 30th employee of Ace. I moved to the States to help Aon build Aon Capital Markets, which was the very first investment bank involved in insurance-linked securities.

I transitioned to an investment banking career where I actually worked with Adam, spent ten years in a variety of different senior roles around financial services, whether it was M&A, advisory, capital raising for public and private companies.

That was during the financial crisis, so I got to see a lot of very interesting disruptions in the marketplace. I took some interesting companies public, learned a lot about them, both from the management team and from the shareholder community, and how they looked at those companies. After my ten-year career in investment banking, I did what every investment banker does, I decided to form a third-party administrator, a TPA. I did that with a family office in New York. And then, in 2021, I was recruited to help the Mitsui organization restart their presence in the U.S.

At the time, they had a small, sleepy company here doing about 400 million at premium. We supercharged that into a high-growth business.

But, of course, we left the decision about how to build the technology to tomorrow's problem. And that problem got bigger and bigger and bigger.

And, at the end of my career at Mitsui, while we were wading through hundreds of different AI-related vendors and use cases, I ironically tripped over and discovered a company called Growth Protocol.

And I was lamenting to them that everybody's agentic AI solution wasn't solving my bigger problem, which is, "How do I make big decisions to move the business?"

And what they built for us at Mitsui really blew me away. And, in fact, it was sort of one of those aha moments that I look back on my career, similar to what we did in the assurance and security, that I was so enamored with what they built that I decided to join and head up their insurance operations.

And I'm delighted to be here and talk about the world of AI in all its forms.

06:09, Adam K

Great, great. Thank you, David.

So, let's dive straight in. Maybe, Rich, I’ll start with you again. I think from the outside, everyone's really interested in what's actually going on. So, maybe, if you could, in a concise fashion, maybe talk about, you know, what are two projects you're actually working on involving AI?

Again, we don't need to know the names of the companies. Share if you want, but that's not the point.

But yeah, I mean, you know, two, you know, granular examples. What do you actually doing right now?

06:40, Richard D

Yeah. Sure.

I think I would select, maybe, you know, briefly, I think I'd classify projects into two categories, right?

One that’s more on the software development side, which is, I think, where a lot of companies focused first was, “how do we improve our software development lifecycle?”  So there's a number of projects, you know, I mean, briefly, you know, to describe some of the specific impact that that's taking place.

Being able to create test cases, test scripts…this is a highly important aspect of the software development lifecycle. And also, you know, quite intensive as far as labor hours and so forth. So, what we're able to do, and what we're doing with a number of clients in specific projects, is to extract any documented, and sometimes not documented, business rules, business requirements.

And through AI, create the test plan, the test script, the test data, everything that, you know, people normally do. And then allow those users to focus more on, you know, just validation. So, very simple, you know, use case. But, you know, something brief I can describe is that AI-automated test script and test case generation is a great one.

There's a lot of examples on the software development lifecycle side where, you know, companies are seeing great value. That's just one simple one, or one concise one.

08:24, Adam K

Sorry to cut in.

So, you know, they're still using their existing core software, but trying to build, you know, functionality or business practices, either testing those or building them with AI. Is that what you're saying?

08:43, Richard D

Yeah. So, still building, you know, kind of the typical, you know, whether it be custom-built or using some of the, you know, Guidewire, Duck Creek, etc. systems. There's always, then, a process, okay, when there's a new release, how do you test, right?

So, yeah, still within that core that you're talking about.

But the efforts that go through around making sure every business requirement is covered, and talking to the right people, and documenting all that, and then translating that into test scripts and test cases that will cover the actual requirements that are built, you know, by the developers.

It's a big process, and you miss things, right?

So, being able to literally take documents, business requirements documents, and develop an AI system that takes those as input and creates these test cases, is a great example of AI in the SDLC, software development lifecycle, space.

But yeah, that's all centered around, kind of, existing software development stuff. But how do we improve it, right?

And there's so many ways to improve it. I'm just picking off one, kind of, easy-to-explain component. I don't want to take too long, but I do want to say that I think what has transpired is, that's where a lot of companies started, was related to IT and technology, because AI is initially perceived as an IT, you know, piece.

And you had CIOs that are very interested in it. And the CIOs world is centered around software development, so that was, maybe, where a lot of experimentation began.

And where I think we've gotten to is more of a business focus. There's a lot of opportunity around business function, business processes, right? So, other projects that, you know, I think are very impactful, and frankly have a clearer ROI, are on the business side.

So some of the things we're doing there are around, you know, customer service, and call center effectiveness, and being able to analyze all policy data, previous interaction with the customer, all these kind of things, and serve it up to a call center rep before they even answer the phone, right? With a summary of, here's what's going on, with suggestions of what to do, with that customer.

And, you know, claims processes, underwriting processes, are huge. Document summarization and recommendations for actions, for claims, and for underwriting, other areas that we've got massive projects that I think generate a lot of ROI and a lot of opportunity for those experts, whether they be underwriters or call center reps, to focus on more value-added functions as opposed to the administrative stuff.

So those are some of the things we're doing, and there's a lot more exciting stuff that we are looking to do, and we should be doing, that I think is going to evolve, and we'll probably talk about that, as the call goes to, you know, what companies should be doing with AI that they're not doing yet.

12:03, Adam K

Okay. Great.

David, Growth Protocol, one or two interesting examples of what you guys are doing right now.

12:09, David M

Yeah. And it's interesting, right?

Because Rich talked about what companies could use AI for in a, if you will, organizing their IT delivery perspective. But we think very differently from virtually everybody else in the agentic AI space, right?

We think about enterprise decision-making. And the way we approach our interaction with our client is to try to lift the conversation up from one that gets bogged down in the, sort of, if you will, forest-for-the-trees discussion, where you've got lots of individual agentic ideas that people want to prosecute.

And even though, you know, AI can help the PRD perspective on getting that project started or run, it runs into all sorts of inevitable challenges, whether it's resources, roles, etc.

So, what we approach our clients with, and there's some on this call, there's some private equity folks on this call that we're working with, we look at this as an enterprise endeavor.

And what we want to do is we want to empower our enterprises to make, you know, big decisions, where they're not constrained by legacy core systems, or how they build their testing process, or political constraints.

So, a good example, and one that's very public, and if anyone knows the company called Pro Assurance, was just acquired by the Doctor's Company.

Pro Assurance was very public about their work with Growth Protocol. And, in that case, even though we were working with them across their entire enterprise, their bigger concern was, "How do I use AI and, in particular, neuro-symbolic AI?"

And we haven't talked about what that is compared to everybody else's AI. But Growth Protocol's neuro-symbolic AI was used by Pro Assurance in their eventual discussion with The Doctor's Company about the acquisition.

And, in particular, they wanted to be able to, A, have confidence in the reserves that they had set. In other words, the strength of the balance sheet of the company. And be able to convince the doctor's company that the merger would be of a company that had a very, very solid reserving base.

And anybody on this call could Google those outcomes.

We validated, with a great deal of confidence, using our platform, that The Doctor's Company was buying a company that had an extraordinarily over-reserve position, in fact, measured in tens of millions of dollars.

And so, when I think about AI, and practical applications of AI, we're looking for situations where our neuro-symbolic AI can actually help companies solve big issues rather than deal with the agentic elements of process automation or speed, if you will, as core thing.

So, Pro Assurance is a great example, but we do that across many, many industries, Adam.

We are not a claim-centric or an insurance-centric company, even though I head up our insurance vertical.

We do that for manufacturing companies. We do that for consumer product companies.

We help them make cross-domain—we call it across domains—so, make decisions where you're looking at information out of your underwriting organization, out of your claims organization, out of your reserving organization, out of all of these different organizations, and looking at it holistically, together with some context and some reasoning, to make big decisions that move the business forward.

A very different approach, and requires a different discussion, go-to-market, conversation, delivery, all of those things are very different about the way we go about it.

15:58, Adam K

Thank you, David.

Yeah, I think one of the big questions for people in and around this business is that there are a couple of dominant core providers, Guidewire, DuckCreek are the big ones.

And, you know, there are some other ones. So it's not just those two, but those get a lot of the headlines.

Rich, I know you've dealt with those and some of the other big core systems for years and years.

So, you know, is AI, and I'd say maybe near-term and long term, do you think AI is positive or negative for these big core system competitors?

And, again, if you could maybe give some granular examples, what's helping them or hurt them with AI? And then maybe some thoughts on the long term for the big core guys.

16:42, Richard D

Yeah. Interesting. I think there's two answers. So positive or negative, right? is your question?

16:55, Adam K

Yeah.

16:56, Richard D

I think initially it's integrated fairly well and fairly positive. But there's a different answer to that, which is it's negative. And I'll explain why.

So, I think the initial positive piece is AI, to be most effective, needs great data, right? We know that, right? The better the data is, the more effective insights you can gather through AI.

And these, you know, our modern core systems that you're talking about, one of their greatest functions is to provide very structured, incredibly, you know, consistent, great data.

So, there's a good marriage there, right?

And so the value that some of these core systems can add, I think it's augmented because they generate such great data.

AI depends on structured data. So this is a positive because they become more valuable. If you have one of those, and it's working well, then you can get more out of your AI.

So that dynamic is very positive.

18:09, Adam K

But in a certain sense, you know, right now, they're sort of dependent on the core systems because that's where the structured data is.

18:16, Richard D

Yeah, exactly.

And so those that, let's just call it, best-in-class implementations with core systems that are providing structured, very positive, you know, data, have a great foundation to generate incredible insights through AI.

But I think there's a tug-of-war that's happening, right?

Which is, many carriers are looking to use, you know, platforms such as Guidewire, but would prefer to use them as their ledger system, as their, you know, system of record.

But, "let me, as a carrier, own the data."

So, there's a tug-of-war over data. There's a tug-of-war over, "let me be in control of my insights, of the more impactful AI-related components."

And, "let me just use Guidewire over here as the ledger system."

19:24, Adam K

Sorry to cut you off.

I think that's a key point. I’m not the most deep tech guy.

But because, right now, a lot of those relationships are structured, essentially, the software vendors, they control the data. Is that right? Control, own, the data?

19:40, Richard D

Well, there's options. I have clients where that's the case. And I have other clients that have frankly refused, "No, Guidewire cannot hold my data. I'm going to hold it separately. So, I'm going to use this piece of Guidewire."

So, this is where the tug-of-war is, right?

Guidewire wants to, and is doing a great job of, "let's infuse AI, let's infuse insights."

But a lot of the pushback is from a client perspective of theirs, from a carrier.

"I don't need to use all the components of Guidewire. I don't want to use all ten slices. I just want to use these two slices, and let me then build my own proprietary components for some of the more higher-value items around Guidewire. But Guidewire wants to charge me a licensing fee for everything. I don't want to use everything. I want to use part of it."

So, this is where the tug-of-war is happening, right? So, the tug-of-war is around huge, you know, large licensing fees for the whole set of capabilities, where I'd like to pay less licensing fees and just use a piece of it.

And so this is where I think we're seeing the short-term, high-value, mid-term, a little bit of this tug-of-war around, “How many pieces of the Guidewire or Duck Creek puzzle am I going to use versus create myself.”

And then you have, I'd say, a longer-term outlook, which is, you have carriers out there that have been searching for, "Hey, what's the silver bullet that allows me to just create my own?"

And the answer to that has always been, "Well, yeah, do you have 100 million and five years to develop it from scratch, because it's really complex, right?"

So, I think that AI might eventually give a better answer to that question, which I think, kind of, puts some of those core system vendors in a risky situation longer term.

So, three answers, maybe. Short term, mid term, longer term.

So hopefully that's clear. I think it's great in the immediate term. But I think there's challenges that are going to evolve over time.

21:55, David M

Yeah, I would echo.

21:56, Adam K

David, I want to hear your position.

You know, just as background, David, actually, in his current last two roles was actually selecting software vendors.

22:06, Richard D

And I'll say this too, I mean, we're one of one or two of the biggest Guidewire partners. They would probably not be too happy to hear from my perspective. But I'm just sharing my client's perspective, right?

So, yeah, I mean, I'm coming from this, just my personal insights.

You know, we are, frankly, a lot of our business comes from Guidewire implementations.

22:38, David M

Yeah, but Rich, you know, even if you didn't hold that view, the evidence is becoming obvious, right?

Guidewire has made an investment in Sixfold, right, an agentic AI underwriting workbench. Majesco has purchased Send, right, an agentic AI underwriting workbench. Duck Creek has issued a press release saying that they want to build a neuro-symbolic AI reasoning tool to sit on top of their products, right?

The evidence is out there that they all know that some form of AI is going to be important for their survival.

And if you think about the legacy systems, what are they? They're really two elements. There's a series of workflows, some of which are easy to implement, some of which take a long time and a lot of money, and are difficult.

And they are a repository for records, right? They drive financials, they drive statutory filings, they drive regulatory issues.

If the workflows increasingly get done in AI solutions, what are the core systems left to do?

Be systems of record, right? Depositories for systems.

I agree with everything you've said, but I wanted to add that nuance because I think they know their world is changing rapidly and they have to adapt to a new environment, and some of them already are.

Now, let's assume that they all go that distance and adapt to an AI world. Part of what I think, and this is, again, another commercial for us, what makes it interesting for us is, "Will agentic AI be the toolkit that allows insurance companies to build all the workflows they need to sit on top of those legacy systems?"

We do not believe that's the case. We think they're a part of that puzzle, but they don't help insurers. Those tools don't help insurers with things like complex decisions, edge cases, explainability, keeping costs under control. Those are elements that we are increasingly talking to clients about, so the answer to the discussion about whether those core legacy tools survive is partly influenced by how much the alternatives cost.

And, you know, "what do you do with them in terms of compliance, right?"

There's another layer of operating complexity that will change the way we talk about Guidewire and the like, and whether they're dinosaurs or not.

25:08, Adam K

And then, Rich, there's a lot of questions coming in around this topic. So, you know, I think we'll continue to dive in here.

You know, of the big core again, it could be Guidewire, Majesco, you know, or Duck Creek, which components are being prioritized right now?

And then two, you know, when they are doing more custom DIY, are they using Claude?

What are the two, you know, two or three that are being used more?

And I know, David, you guys are punching in there also. But, Rich?

25:43, Richard D

So, yeah, I mean, I think there's a number of clients that have either implemented or are committed to implement, you know, the full Guidewire suite, for example, ClaimCenter, PolicyCenter.

As far as the component, you know, I mean, I have a client who specifically has created a PowerPoint deck that outlines the multiple streams of the system components and is targeting, more as David said, more as a system of record, right?

More of the ledger piece. "Let me build my own. Let me own the data. I'm going to build my own AI components around it, my own insights. I'm going to integrate it with our underwriting system as opposed to using all the provided components."

What are they using?

I mean, I don't think there's one answer. I mean, I have a couple clients that are highly using Claude. I have some that have developed their own. I have one that has developed their own and is using Claude.

I think Claude is probably, you know, the one I've seen the most, as far as, you know, insurance enterprises, signing up to licensing and utilization.

But I don't think there's one answer.

27:08, David M

No, I think that's right. I don't think there's one answer.

Some companies are going all in. Brown and Brown announced a partnership with McKinsey and Anthropic, right? They're going all in on Claude as their agentic AI partner.

Others are sort of agnostic. They use different tools with those systems of record and what they're trying to build, or they're looking for vendors to do, like the Vala, right?

They're looking for vendors to help them put rules around how those tools work, right? What those agents can do…Claude, Gemini.

So, there's increasingly a nuance to the question of, what should we use? But I think the answer is becoming, the question of what they should use is not the totality of the question, right?

Because what we've seen in the last year is, when people think about going down the journey of using AI, I think there's been some sharks, right?

Maybe I'm being hyperbolic, but two of the sharks that are starting to show up in every meeting we go into now are…we meet the CTO or the CIO, and they said, "Yeah, I just went to the board and explained to the board that my token budget for the year, I blew in the first two months of the year, and I expect that number to double next year because of the increasing complexity that those AI models are being run at, right?"

The frequent calls, the compliance elements, the rules, right? Those elements are driving people to sit there and go, "Okay, I could use Claude. It's a good choice. But wait a minute, what's my effect on my cost and maintenance?"

The other thing that I think is going to shock people, and will force them to be thoughtful about what they use, which of these models they use, is the increasing requirements under the regulations to talk about explainability, right?

How does an AI tool, in a workflow that might have been in Guidewire before, like ClaimCenter, how do they make a decision or a recommendation?

And it won't shock this audience to, you know, when you said, Adam, which models are people using first, or which tools are using first?

I like to think about which business units they're going into first, right? And what we've seen is a plethora of people going into underwriting with these questions, "Should I use Claude or should I use other tools?"

We haven't seen as much work yet on the claims side, but what we're seeing is the biggest evidence of where the wrong choices are being made, right?

There's thousands of court cases now where judges are literally now moving from a slap on the wrist to, you know, sanctions against defense counsel and plaintiff counsel showing up in court with a work product produced by, say, Claude, which has a significant hallucination. And the plaintiff argues, "Wait a minute, is that the precedent you used to deny the claim that you were going to pay to my 80-year-old client who relies on the income that you've denied her?"

There are lots of downstream, unintended consequences. The regulatory one is one I think people are just getting their arms around, and that will drive them to be more careful about the kind of tools they select to work on top of those core systems.

30:35, Adam K

To the extent that you guys, whether it's hearing, seeing, or, you know, industry scuttlebutt, you know, which of the, I'd say two or three, public companies, stocks, you know, are being out front and maybe, you know, maybe have a lead in using AI.

30:59 , David M

Well, I will leave that one to Rich.

I've got some views, but Rich probably has a broader landscape across the industry.

31:08, Richard D

Yeah. So, I guess I have a perspective on this, Adam.

That is, the winners now are the first movers that, you know, where people are at with AI. I mean, certainly there's a lot of companies doing stuff, right? I mean, so it's hard to name whose stock is benefiting now, right?

But I mean Progressive, Travelers, the Hartford, Zurich, you know, these are all companies that are doing things.

But I've left off, you know, 40 companies by naming those, right?

31:43, Adam K

Just stay with those for, you know, don't need granular, but you know, the ones that you're hearing, seeing, is it that they're plowing a lot of money into it? Being really active? Is that, sort of, the bar right now?

32:00, Richard D

Yeah. The bar is, you know, if we rewind twelve months, or what have you, you know, it was experimentation, and, like I said earlier, a lot of it on the technology side.

Where we've come now is a lot more focus on the business side, and a lot more moving from experimentation to actually operationalizing AI, which drives ROI and eventually impacts stock price.

But I think the real impact, we haven't seen it yet, and I don't know who's going to win that one, right?

I think the real impact, from a stock perspective, from a value perspective, is so much of what's going on right now is around existing.

So, yeah, we've moved to business processes, right? Outside of just IT. And we have business leaders interested in driving AI initiatives.

So much of it is around looking at existing processes and improving them. Massive improvements, incredible ROI, but business processes, automation of existing processes—all this stuff sounds great.

Where I think the true bigger value is going to exist, and is really going to be seen in stock prices, is for those companies that will be some of the first to figure out how to reimagine things seriously, right?

So it's not, "Okay, we're taking our existing world that we've lived in for the last, you know, decades and decades, and applying AI to automate it and make it better."

There's nothing wrong with that, that's great. But flip that thinking and say, "Okay, now that we have this tool of AI, what if we started from scratch and reimagined the underwriting process, the claims process, whatever process it is?"

With that as a foundational tool, we might create a completely different process that is so incredibly better from a quality perspective, or efficiency perspective, to allow that company to grow or drive, you know, better expense ratios and so forth. So, I don't think we're there yet.

34:19, Adam K

Can I jump in here? Two or three, and try and be concise, because I want to get to David in this, big topic.

So, one, is there a difference in, you know, near the next couple of years, for, whether these are public, or, you know, we're talking about tier one to two.

You know, if the companies are on more modern systems, you know, small amount of systems, you know, is that a big edge? If you're on Duck Creek or on Guidewire across your suite versus being on ten different systems? So that's one, base ahead.

Two, are you seeing, you know, I would think it's just the nature that, you know, you definitely have some differences as far as one, two, or three companies, if you want to name them, fine, if not, who are, like, going all in, whereas a handful are like, "Well, you know, I just got to cover myself, make sure that we're not falling behind."

And then three, the last point, is to what you're saying, you know, are you saying one or two companies begin to reimagine, really try and change the process, or is that still whiteboard at this point?

So three really quick and then David, same for you.

35:32, Richard D

Okay, so first question: do you need to be on one core system? No. I mean, I think in some regard that helps, because the better your data, the better you can, I think, implement AI.

So I think that data foundation is probably the most important factor. And if you have, you know, a hundred legacy systems with a hundred different data sources, that is a little bit more challenging.

But I think I put that foundation more around, how good is your data, rather than, are you on one core system? But it certainly helps, because it's standardized, and maybe it helps your data, but I don't think that's critical.

I think the critical factor is more centered around organizational aspects of really assessing how to use AI most effectively.

You know, the "who's reimagining" question, which I think was the third question.

Yeah, there's a few companies that are probably pushing the envelope a little bit more, but frankly, I don't think we're there yet, because I think companies need to go through the progression of really getting more comfortable with AI and how to operationalize it with maybe some of their existing processes before they completely rip the Band-Aid off and just start from scratch.

And that's a risk for people, too, right? Executives are like, you know, going to be, you know, considering, how big of a risk is it to really go all in and reimagine everything? I just don't think we've got one that's done it fully yet. But pieces are being reimagined. Sorry, I forgot question two.

37:08, Adam K

That pretty much answers those.

David, are you seeing, because AI, are insurance companies moving to the big systems, moving to the cloud quicker?

And the flip side, is it slowing down sales cycles for the big systems?

37:33, David M

You know, it's interesting, right? Because the dialogue so far around this has been around what insurers used to do yesterday, right?

We talk about the big core systems. And I'm going to give you some counterfactuals from things that Rich said, right? Because I find this interesting as a student of the industry and a former strategist, right?

I see people like Jamie Dimon get up and say, "We've spent $20 billion on AI over the last three years." They've got a thousand forward-deployed engineers, and we haven't figured out how to extract true value yet, right?

We see AIG with these monster contracts with Palantir, not yet extracting any value. And we, sort of, come at this question through the lens of, "Well, will big core legacy help us do anything, or will it be a victim of this?"

And we talk about process. Let's see who can break away from the pack by having a better process.

But if everyone follows the same methodology—"let's bring in agentic AI to improve, you know, workflow submission," or, "extracting information from a claim"—everybody's chasing exactly the same improvement.

Nobody has built a moat, right? To get back to your question, Adam, nobody's figured out how to build value.

So, how do you build a moat if everyone's chasing the same thing and doing it the same way and following the same path? And building moats is what drives value. And, this is an investment banking call, and a lot of your clients care about who's building the moat, right?

And so I would add to that, very clearly, we haven't seen anybody yet in insurance build the moat.

And I think what the moat comes from—and Rich sort of hit on it by talking about data as the foundation—where the moats are going to come from is not because you can process a claim a little bit faster than you did yesterday. And so you solve a $100,000 problem with a $10 token solution.

It's taking your data and really interrogating it so that you can make big decisions, move into businesses that you're maybe underweight because you've figured out you're really good at underwriting them, and you just haven't allocated enough capacity. Being a smarter reinsurance buyer, so you don't cede as much premium. Figuring out claims information that allows you to manage claims and litigation in a far more effective manner to bring down claims costs.

I'm a big believer that if you look at the economics of insurance, right, we know the big three: premium gives us a hundred cents on the dollar for our combined ratio.

Out of that, we pay losses and we have expenses. What's the biggest piece of the stuff that walks out the door? The loss ratio and LAE, the loss ratio, right? Indemnity payments to our customers.

40:36, Adam K

Sorry. Can I jump in? How much of the moat you're talking about is IP protection? So, that's one question.

40:46, David M

None. None.

40:47, Adam K

The second question is, you know, again, no one's stepping up. Are you guys hearing of any one, any companies?

40:54, David M

I gave you one example, right? ProAssurance. It's been very public about the power of using our type of technology, right? A reasoning AI to make a profoundly different outcome on their balance sheet, right?

And let's be honest, right? This is all we should be talking about, right?

Is, how do these companies invest in technology and software to improve their businesses?

41:21, Adam K

Okay, Rich, just want to go back. Sorry, because I've missed this question two or three times.

So, you know, one, is there an increase in cloud migrations because of AI? Two, is there a slowdown in the sales cycle because of AI, would you say?

41:41, Richard D

Yeah, I think that was question two that I forgot.

No, I haven't seen a slowdown in the sales cycle. Now, the sales cycle is nothing of light speed in the first place, so tracking that pace might be difficult, but no, I specifically have not seen any slowdown in the sales cycle.

As far as cloud migration, I think there's a number of aspects of AI that enable better opportunities to support cloud migration activities.

So, if there's any impact, yes, it's positive. A lot of cloud migration pace depends on other factors organizationally, you know, existing licensing on mainframes that are causing timing and urgency.

So, frankly, a lot of the deadlines, if you will, for cloud migration are gated by some of these other factors, like, "When do we need to get off of our mainframe?"

And that answer still stands regardless of the tools that you have. But yes, the ability to migrate to the cloud more effectively, with better quality and faster, is absolutely supported by AI tools that exist.

And a lot of companies like mine, I mean, this is not a commercial, but there's consultancies. Many consultancies have developed cloud migration tools that are AI-based, that allow that process to be much quicker. So, we certainly are able to do our cloud migration work in a much more effective and efficient, quicker way, with incredible tools that we've developed.

43:46, Adam K

Okay. And come to the end, probably 1 or 2 last questions. This is a different direction, but if either of you want to take a stab at it.

You know, the huge value with a lot of the AI in the data centers, you know, is that insured value being recognized?

In other words, is their insurance market developing around, you know, these big data centers, the big infrastructure projects?

Do you guys have any visibility into that? If no, that's okay.

44:18, David M

I don't but I have some personal views.

But I think it will be a challenge for insurers to take on insurance for those things in anything other than traditional property-related matters, right?

There are going to be cases where things like privacy, zero data retention issues emerge. And will data center providers that work with people doing the models, will they be dragged into it as interested parties?

I think there's a huge amount of coverage-related questions that haven't even begun to be broached yet that we need to look at.

44:56, Adam K

Okay, another quick one, and then I'll leave you guys two minutes for, you know, anything you want to say in closing.

But this is an interesting question. Does AI make, you know, MGAs more or less relevant?  On one hand, you could say, "Well, they can get up and running quicker," but on the other hand, you could say, "Well, maybe the carriers can develop that product expertise in-house easier."

 So, any two cents on that one?

45:23, David M

I think it's an interesting question, right?

And I don't know that you could look at the answer in isolation around technology, right?

Because you need distribution to match up to the MGA model, regardless of where it sits, either independent or as part of a carrier operation.

And then you need risk-bearing.

Now, what I think AI will do—whether people can afford to implement it is an open question.

But I think what AI has the power to do is make that whole workflow…distribution, packaging, right, propensity to put it to the right MGA, make it easier for the MGA to evaluate, to price it, take that risk, and show it to the carrier in a quick manner with more details, more insight.

All of that improves the process. And, in theory, using things like neuro-symbolic AI make better decisions.

So, the question is, who has the money, and the engineering, and the time to develop that? Is it the carrier, with, in theory, lots of resources, but lots of competing interests? Or is it an MGA that doesn't have a big financial piggy bank, if you will, but a dedicated focus?

And so, if I'm a big MGA plant, you know, I'm backed by a big private equity firm, and I've got a focused AI capability, that might be the person that can exploit AI in an MGA environment.

But I don't think a smaller MGA with, you know, $15 million in premium, $20 million in premium, is going to be able to deploy the AI at the same speed, with the same consistency that the larger players will be able to deploy.

47:11, Adam K

We’ve covered a lot here, but final closing.

Rich, any final closing thoughts you want to throw in?

47:19, Richard D

Sure. Yeah.

I do have one area that I think is a good closing topic.

So, you know, I think as we talk about AI, we are very focused on technology, very focused on the business process, you know, ROI that can be achieved, and so forth. I think that one thing that is being underestimated is, it's not all about the technology or the business process, but about the organizational change.

And, you know, it's not that it's easy, but there's one plane of activity that's centered around operating model and technology.

But infusing AI into an organization is one thing, but changing how 20,000 people work every day, and interact with each other, and maybe have different responsibilities, or the value that they deliver is shifting from one role to another role.

There's a lot of organizational change. And, you know, I think successful implementations of AI need to also look at, you know, investment around training, and governance, and business process redesign.

So, I think, in closing, I would just emphasize that, you know, you need to take your lens somewhat away from a technology idea to just a complete organizational rethink, to be truly successful in implementing AI.

48:51, David M

Yeah. I would add to that, as you're doing that.

My closing thought would be to make sure that you really assess, like, what's the total cost of ownership of going through that journey, whether you impact the humans in your company or you don't, right? When you go on these journeys, the total cost of ownership is, I think, a calculation most companies have yet to do.

And it's becoming a surprise to them as they continue to experiment and think about process. And that goes hand-in-hand with the regulatory side of things.

I think a lot of insurance companies are not yet really being forward-thinking about the impact that regulation will have on that future journey that Rich just described.

49:42, Adam K

Rich and David, thank you very much.

 Audience, you know, we're really glad you could join us, and we'll definitely be doing more of these venues. So, thanks, guys.

49:51, David M

Have a good day.