Data First, AI Second: Stan Smith, Founder & CEO of Gradient AI

Speaker A: This is the Leadership in Insurance podcast, um, brought to you by FinPro Search Partners.

Speaker B: Insurance companies are businesses and they need to look for the long term and be sustainable.

Speaker A: We went from 0 to 1 and

Speaker C: now it's going from 1 to 100.

Speaker A: Insurance as a concept, as a kind of service is brilliant. The execution is what we're looking at now. I think the companies that are going to succeed are the ones that are going to understand and master the art of intent. When we talk about innovation, we lean too heavily to think about technology and we don't think about creating a culture of innovation. I think innovation is essentially continuous improvement of existing processes and platforms and products.

Speaker B: Right.

Speaker A: It's got to be easy, it's got to be seamless.

Speaker C: Welcome to the Leadership and Insurance podcast. I'm your host, Dan Bristledon. Today I'm joined by Stan Smith, CEO, uh, uh, and founder of Gradient AI. Stan, welcome to the podcast. How you doing?

Speaker B: Great. Dan, great to be here. Appreciate you having me on.

Speaker C: Likewise. Thank you. Thank you so much for joining. Um, for those listeners, uh, out there who listen to us regularly, you may well be aware of who Gradient AI are. Uh, they've got an incredible brand and an incredible story, uh, that's sort of unfolding in the market. But for those who don't know, Stan, if you could just introduce yourself and tell us a bit more about your background and how radiant AI came to be sure, Dan.

Speaker B: Uh, you know my background is startups. This is my sixth startup. I didn't start them all but uh, I've enjoyed the idea of new ideas, innovation, fast moving companies delivering great value and having exciting results as a result of that. A couple IPOs along the way. Um, and uh, one of the things uh, about three companies go learned about this machine learning thing before it was cool enough to be called AI, but understanding that the math could discover very accurately patterns and data. So I started a supply chain related company that was using machine learning. Um, it's a very powerful technology and it was powerful back then and it's just gotten more and more powerful as we're all seeing around us these days. But that led me to looking at a market called insurance and trying to figure out how to get in there, which led me to Milliman. You want me to go on with the Milliman story?

Speaker C: Yeah, we'd love to know a bit more about it. Just how you sort of pivoted, how that came about and how you pivoted into Gradient.

Speaker B: The conversation started with Milliman was going to buy software, uh, a venture Backed AI software company because one of the biggest competitors had just bought a software company and they thought they should do the same thing. And I came in to answer some questions to some of the folks that were looking at the deal and they decided that that didn't make sense for them. And they asked me if I joined Millman to start an AI business. I sort of, my first response was no, thank you. I'm one of those startup guys, this is a big company. I don't, I don't do big companies. And they said well actually our business model is more like a franchise, uh, model which is you can start your own franchise of Milliman and it's sort of wholly owned by you and you do your own thing. And all the other franchises, they keep their money, you keep yours, you keep your losses. If you have them, you pay them off, um, or you go out of business and it, and they don't, you know, nobody covers anybody else's business. And I was like m, that's sounds like a startup again. And I said exactly. So I joined Milliman to build an AI business, a software as a service, AI powered, um, insurance related business. And so that's, and the advantage of joining Milliman, um, data is very, very key for AI. I think everybody knows that in the insurance space all those companies are very, very protective of their data. Whether it's underwriting data, claims data, they don't want anybody to know about it, what it is, how they, how they price policies, how they manage claims. And as, as a guy trying to figure out a way to get data to use to build models and insurance, I thought Millman's blue chip brand would allow me to you know, build trust with companies to provide me their data. And it did. So it was that, that part, that part of the thesis was turned out to be very true. You know those companies that trusted Millman for years, they still trust Millman. Um, and, and an actuary can't do your work unless you give them your, your, your information, your data. And so that trust was established at Nilman and I was able to you know, make, make a lot of progress uh, because of that trust. So that really helped me get this business started.

Speaker C: Yeah, and I can imagine as you said, that the minimum brand enabling you to have that trust to go in and probably do a lot of discovery, a lot of fact finding, a lot of um, interaction with customers and carriers to understand what their, what their core pain points are, which is kind of, I'm curious to understand a bit more about what led to the specifics of what you were building at Gradient. I know there's lots of people in the market trying to look at, you know, transforming or disrupting the core system space, but you've gone into sort of like a best in breed, uh, decisioning platform. Can you tell us a bit more about why you made the choice to go into that space specifically?

Speaker B: Yeah. So you know, to me I go, I'm looking for business problems and the business problems that I saw in insurance were the core system is a, is a challenging area. There's you know, Guidewire and others in PNC space that dominate. But it's big, it's heavy software. I've been in those market, those kinds of markets before. Um, it's, there's good things and bad things about it and just from my perspective I find that to be kind of heavy kind of business. And that wasn't something that attracted me. But I also saw a big need outside of that area in the insurance companies which is they're just simply put their data rich, knowledge poor. They have a lot of siloed information so that they can't see in some cases a full policy set of data in one time. Which sometimes sounds hard to believe but I've seen it firsthand with customers and, or can they manage claims effectively or can they bring the claims data back into the underwriting uh, front end effectively? It's just there was a lot of challenges and as uh, somebody looking for those kind of problems, uh, I saw it everywhere. So to me it just drew me towards that part, that sort of higher level than core system. It's sort of underwriting workflow. Some people call them underwriting workbenches. There's a layer above the core systems where they can and move quickly in a repeated managed fashion. And we found to me it was plugging into that workflow was the best and fastest way into the insurance space to add the most value that I could.

Speaker C: Yeah, and it's an incredibly um, fast growing area of the insurtech space, isn't it? I'd love to know a bit more about how the gradient product slots into that underwriter's day to day workflow and how it kind of makes their life easier essentially.

Speaker B: So it's. What's fascinating about insurance to me is that you know, the business that comes to these carriers comes typically through a third party called a broker or an agent that has a relationship with that insured. So the insurance companies come working with the third party to get to their primary customer. So they, they are given a set of information that is typically given to all their competition at the same time. So they, um, they're trying to then make a kind of two decisions. One, should I look at this and quote it? Because if it's not something I want to write, I don't want to spend the time on it. And then if I quote it, how should I price it based on the risk characteristics? And so, you know, we see opportunities to help them in both those cases where we can actually help them with kind of a quick pass. You know, green, red. This is not a piece of business that's going to fit your risk appetite, or it is. And so move it along. Because we have clients that only, of the quotes they send out, they only bind or win about 2.5%. Some are as high as 15 to 20, but there's a lot in between. And so they have to do a lot of work to quote each one of these submissions. And if we can help them quote fewer but win more, we think that could help. But it's really inserting a risk score at the right time in that underwriting process where they have other information from the business, size of business, type of business, number of people. In healthcare, we get a census file on the individuals for the insurance, so we can go find identified data to help support the risk assessment for that group, but do it in almost real time so that they can move these quotes through quickly, because time is one of the enemies they have. If a group comes through a broker and gets sent to a number of markets, which are basically insurance carriers, sometimes the first one back can win the business. And so it's not just about making the right decision on your side. It's doing it in a timely fashion so that you have an opportunity to win that quote.

Speaker C: Yeah, absolutely. Time is money, as I say. So it's about like keeping the speed, keeping the pace. And, um, where have you seen that roi? Because you've obviously got a huge customer base now, but as the company's grown, where have you seen, I suppose having implemented that and now seeing the kind of fruits of those implementations where we've seen the biggest ROI for your customers.

Speaker B: So we, we measure it in what's called, on the healthcare side, MLR medical, uh, loss ratio. And that's a very direct measurement of if we identify the risks appropriately and accurately and they price it per that accurate assessment, we're seeing our clients have steady improving MLRs so that they have a nice, healthy business making decent margin and profitability for them. And so that's really, it and if it's a small book, the ROI will be in the single digit millions typically. But in large books it can be tens of millions of dollars for ROI per year. So I've been very, I think to me, the one thing I want more than anything else is to see a strong return on the investment from my clients. I want to know that I added more value uh, to them than I extracted from them. So I want to be, if I can, in double digit roi. And we've got a lot of situations where it's, it's 10x plus, um, which gives me it, you know, it makes me feel good about it. I like if it's better, it's better. Um, but uh, I mean some serious large impacts on and on a recurring basis. So that's, yeah, it's really, it's the, the reason I started the business and it's really gets me uh, excited every day to come to work to do more of that.

Speaker C: Yeah, absolutely. I mean as you said, there's so much opportunity to really add clear ROI to people and you've clearly got a great product does that. There's a lot of skepticism around um, AI and underwriting. I know that we're kind of moving towards that curve of adoption, but could you maybe speak to a little bit more about model accuracy in general, but then more specifically gradient models?

Speaker B: Ah, I think, I think the black box has been uh, the term many have used for a long time. I think actuarial math that has really dominated insurance is more m about knowing how, what math was used than actually uh, trying to get an accurate outcome. So it's been about transparency. So the regulator could see that you use the proper method to project your losses so that I can sign off on your, your loss projections and things like that. There were reasons for this, but when you get into other things like pricing risks or estimating the duration and, or complexity of a claim, some of that math, that's not the best math for those different uh, applications. And so that's where AI starts to come in. Um, I think there's a real big term in insurance called credibility where the actuaries always want to have historical information so they can project it forward again. That's rational, is valid. AI comes into play in a lot of cases in small insurance policies, uh, small business policies. We've got young companies that don't have any history or have limited history. Um, they may be in a class of business that's got some risk, but they've never had a loss. So how do you price A business that has no experience if you will. But also there's millions of them and you'd probably like to write a lot of premium if you know how to do that. And so it's uh, a place that we have really uh, done a lot is small business, uh, insurance. Whether it's small group, uh, like under 200 lives in uh, healthcare or actually small commercial in the PNC space where if you look at 100 companies only one or two out of that 100 are going to have a loss. But how do you price for the majority not having a loss but absorbing those two losses? AI is kind of a really powerful uh, way to do that and to observe characteristics of companies that will have more losses than others and understand that as opposed to have to see direct losses by that insured. So there's just the inference area of AI is really powerful. And when people ask about accuracy you can back test it to know how accurate the models are. You can look at a model and say okay, I'm going to run it a year ago in time and you can back up the data so that's done accurately just as if it was in production and then you can test how it actually performed. There's also on a go forward basis we've had clients that have done a B testing. They've actually put, put a series of claims for example. Uh, they've kept it out of our models and not run through the models and not manage with the data from the models to give them a real clear roi. It's been fascinating and very rewarding. I mean that's one of the clients that had 10x greater um, ROI on what they spent with us because they got to measure it. Their own internal data science team did the measuring, not us. Um, it was very satisfied to see that kind of impact.

Speaker C: Yeah, that's really interesting. What about the um, governance then? Because there obviously uh, have been lots of um, events where there's hallucinations, there's things that can possibly go wrong in the back end. Is there anything that can, it has to be in place specifically to stop models from drifting or making bad decisions. And what does that look like?

Speaker B: So you started off sort of the governance and the bias stuff. So one of the things we, we've done is we, and we can do because we got data from a number of companies, we have a contributory business model suite data both in the property casualty space and the health space from multiple companies. Um, we can see across companies, we can see across companies and across uh, geographies and then measure by type, type of company profiles and other things that to measure for and test for bias. Um, even if it's something that that company can't see themselves because they only have their data to compare themselves against. This is an area where having sort of contributory data set and sort of a benchmarking capability really does come into play. Um, I think what the regulators are always trying to do is prevent, uh, somebody being disadvantaged by some sort of, uh, insight that the models or the insurance companies shouldn't have. And that's fair. So I think what's been happening and will continue to happen, I think you're going to see less concern about how these models work and more concern about what data you feed them and what outcomes are you allowed to use. Because, uh, these things are going to have. I mean, you just think about all the data that they keep training these models on. It's going to have more and more data about all of us. And when you think about the marketing files that most of the marketing companies have on all of us, it's incredibly accurate information. So when that data starts to flow through these models, the models get to know Dan and Stan. You know, pretty good idea where we live, how much we make, how many people in our families, how many cars we have. All those things that are just, you know, you'd be pretty surprised at how accurate it is. And uh, that's. I think, I think the regulators just have to say, okay, that data is out there. Either you can't use it for certain applications or you can't, uh, have these kinds of outcomes, uh, to protect people, protect people's interests. So I think, I think that's been in place for a while and I think it's going to stay in place and the models will just get better and better, but the regulators will sort of, you know, box them in, I think.

Speaker C: Yeah. Nice. Okay. Um, the subject that I'm really interested in that I hear so much about the market is, you know, robots stealing human jobs. You know, so we talk a lot about this. In terms of the human in loop part, where do you still see humans doing things better than AI and um, like, yeah, I guess what's the human edge at the moment that AI can't necessarily replace or automate? Yeah.

Speaker B: So look, you can see I've got gray hair. I've been doing this longer than, than you have. And probably most of the people that will listen m. There's. I've been so many, uh, uh, technology jumps where everybody said, this job's going to go away. This is, you know, it's going away. Software engineers, they were going to be out of jobs 30 years ago or because you know, there's greater, better language, faster compute time, all sorts of different things that was going to, was going to hurt them. There's not enough software engineers. There still aren't enough software engineers. And I think what this, I think what you're going to find is jobs are changing and they always change. When we all got different levels of technology or our cell phones or laptops, it changes how we do our job. But the jobs are just, you know, they're just uh, slightly different. And so we are hiring differently here at Gradient because AI allows us to have people do different things, things that they normally. For example, we would hire some junior software engineers that usually would sort of work in conjunction with one of our senior software engineers. Now they tell me it would take them more time to tell a junior software engineer what they should do than to have the AI, uh, a code assist do it for them. And so, and, and the stuff they're left with uses their judgment, their experience, their knowledge and they're actually enjoying their jobs more. So you know, all this, it's going to take jobs in some cases going to take the drudgery away from jobs. You know, there's not a software engineer I've ever met that wants to do documentation of their code. Guess what? AI does it for them. I mean there's, there's things that make their job better. I think the more we can do that, the more we can deliver to an adjuster. Not just this claim is going to be high risk, but here's, here's by the way, here's a care plan you should consider. Here are the milestones you should consider for this claim, for this person on this injury. So it makes them, they have judgment. They'll say, I've seen this before. I don't think that's exactly right because of some, something that I see or I know or I suspect or you know. I agree and I'm going to do that using their judgment. I think the judgment becomes more and more important as AI takes away more and more of the mundane. And I think that

Speaker C: it's an enabler like you're actually allowing people to do more high value tasks, look into things in a bit more depth. They wouldn't necessarily used to and actually that can have a positive impact on the customer care part of the journey as well, which is obviously, you know, integral to, to um, you know, retention. So I think that's really good. Um, have you seen anything in terms of how. I think it's easier to do when you've got, uh, companies that are starting looking. We obviously work for a lot of MGAs that are building AI native businesses and beginning the journey with technology at the forefront. I think the harder part is actually when you're looking at legacy, bringing people on a journey that have got very archaic processes or people that have been in their jobs for 20 years that either maybe don't want to change or are afraid of change, what do you kind of see and hear from the customers that you talk to and what you see as that kind of balance that's being struck?

Speaker B: I think I started this business over 10 years ago. I think the openness and the, and the acceptance and, and actually excitement about taking on AI now, I've never seen it before in the space before it was. What is it? I would have meetings with people, you know, I thought were interested in my products and no, they're just interested in learning what AI was, you know, 10 years ago. So that, that was interesting, you know, to educate people. But you had to find the few that were willing to say, hey, I want to get, I want to take a jump ahead. I want to get a competitive advantage. Um, I think everybody's realizing that it's great for a lot of different things, some of it just getting rid of busy work and nobody wants busy work. So I think they're, I think the adoption rate is pretty fast there. I think you're going to see a lot of internally built tools for workflows and things like that. Um, but I think they're also going to try and build some pretty complex things. So. But they're always going to have their data, you know, sort of hermetically sealed behind the firewall. So that same, uh, security also prevents them from seeing a lot of things outside their four walls.

Speaker C: So.

Speaker B: And that's where someone like Gradient can come in to help there. But I, I still think it's the best time ever for being in the insurance space with AI because I think they're, they're now realizing this is, it's here to stay. It's going to make a big difference. And if they don't get on board quickly, uh, they can fall behind.

Speaker C: Absolutely. Well, that's, that's another point because, um, I think we're just seeing a bit of a domino effect across the market, aren't we? With people that are kind of early adopters, you've got to read all the reports that are Coming out of the, you know, the various different um, uh, publications around where we've gone from pilot to adoption over the last three or four years, there's many more people that have had those strong use cases now. So it's table stakes in a sense that you know, are we moving towards a market where having AI technology and AI automation is par for the course and if you don't you're essentially going to fall quite far behind quite quickly? Yep, yep.

Speaker B: I think one of the decisions these companies have to make is even if they can make it, should they make it? Because I think it's going to stop being can I or can't I? Because the answer, I think the more these things continue to get more and more powerful, the answer is oh, is it going to be. I probably can. So the question is, should you. Is it really as a insurance company, should you actually build which parts of your IT stack should you build and maintain and own versus uh, not. And I see it all across the board. There's a lot of companies we work with that homegrown core systems, they built it themselves. They're always going to want to build to themselves. That's the way they think. It's a differentiation for them. And you go down the street to a company they compete with head to head and they buy third party software off the shelf and they think that's a advantage there and they're both convinced they're right. Um, ah, there's going to be a wide array. Um, but I think you're going to just, I think it still comes down to what should you build and, and what should you license. And by the way here at gradient we don't build everything. We, because it's, you know, it's a build or buy on everything. Um, you know, stick with your core competencies, do what you should be doing and need to do. Um, and what you can license in many cases still is a better choice in my view.

Speaker C: Yeah. What, what are you seeing? Because obviously we've, we've got a lot of noise uh, around, you know, people like anthropic making big moves in the financial services space. A lot more companies having that technology readily available at their fingertips. What are you seeing in terms of actually buy versus build and what's working and what's not? I guess

Speaker B: so. I think the thing I see the most of one category out there is document summarization. It's gone from a pain for people to do and everybody wondering how to kind of know what, you know, if I'm going To if I'm looking at a claim, whether it's because I'm taking care of the person or there's a lawsuit or whatever, there can be thousands and thousands of pages of medical documentation documents and typically that's handled like PDFs and you just get, you know, thousand pages. Well, in medical documentation a large portion of that document is a repeat. So if I go see my doctor 10 times, the first visit is there when I'm there for the second visit, those first two visits are there on the record when I go for my third visit and so forth and so on. And so by the 10th visit I've got multiple documents that have the same visits in copies. But those will all be a document has to be, you know, uh, you know, pulled and I, I get charged per page. There's a big business in insurance around document retrieval, which is kind of interesting. It's like I didn't realize that that was, but it is, it's a huge business. But then Now I got 4,000 pages and I'm trying to figure out what I'm going to do with your, your claim, Dan. Now what do I do? Well, I used to have to read it or I'd send it to a low cost region for them to convert it to an outline. And now document summarization instantly, um, I get an outline, I can, I get all sorts of great information and I know what's going on. So that, that is the, probably the number one use case across the board that we've seen is it started with those medical documents. Now it's going into data ingestion for underwriting and just anything that there's multiple documents where you need a good, consistent, accurate summary and you need key insights into all those pieces of information. LLMs are great at that. And then we come in downstream from that to say, okay, here's outside information, here's a risk score or set of risk parameters or insights that you need to further that, that process and to make that final decision. But up front, those LMS are getting better and smarter daily. So it's, it's, it's a, it's incredible. And again, we're taking advantage of it here, uh, and we're including some of those capabilities in our products as well. Because why not?

Speaker C: Yeah, see, and how do you sort of um, build a moat then to kind of keep, keep ahead of the curve and the rising tide of AI.

Speaker B: So that is the key and I think our go back to the beginning of the business when I wanted to start this business, uh, My thought was that, you know, AI needs data and if I, if I just have AI, eventually everybody has AI, just the way technology works. And here we are. I wish I invented some of it, but I didn't. Um, but I always thought that the, the data would be the key. So when I started the business and part of the reason I joined Milliman was to build a sort of independent trusted company that could in fact be trusted with data from multiple companies. And uh, it was also part of the reason when I started the business to go after small companies because they didn't have enough data on their own to compete against the big, the big companies, big competitors didn't have the resources to build a lot of the stuff themselves, didn't have the people or the, or the funds, but they could if given the opportunity to take advantage of greater insights at a reasonable cost. It just was the right starting point for the business. And so I got a number of companies to start trusting me with their data and sort of built that flywheel effect and they got more and more and then more and more bigger companies to build sort of a critical mass so that we could bring a lot of value to each new customer and all our existing customers without sharing anything sensitive, without providing anything that's not allowed. Um, there's just so it's been really about model training pattern, uh, observations about you know, different things that go on in different places for different reasons and just having really well trained, uh, insightful models and then being naturally pulled up, up market. So you do that from smaller companies and the mid sized market folks realize that you could help them too and you start getting the mid sized folks and then you start getting to the, the large um, carriers and we've got a growing number of large carriers now. It's just, it's sort of a natural progression because it turns out that a lot of those MGAs that you probably bring on your podcast are using somebody else's paper. So when we got into some of the underwriting in group health, we were helping some of these MGA's right business and have great MLRs. And the carrier paper that was, they were using took notice and said what are you doing? Oh, it's gradient. Let me look at it. And guess what our first carrier deals were basically through those kind of warm introductions we were already indirectly working with those carriers and they could see how we performed and it was a great Torah, uh, a couple first deals because it was totally natural and my board was surprised that we sold some carriers because we, I sort of Said to him, they take a long time to make decisions. I'm not trying to do that now. And suddenly we land some carriers. I'm saying, well, we already had our, uh, you know, chance, uh, to, to uh, show them what we could do indirectly. And so it just made it straightforward. And they're like, well, let's go do more of that. And so that's what we're doing, more of that, more carriers.

Speaker C: So it's a really smart strategy, isn't it? You've kind of got your, your use case on a smaller scale, the familiarity. You can see the work that you're doing. And then that's what I was curious about is how do you make that transition upstream? And does that layer in complexity then? So, versus a, uh, sort of an M and a, an MGA project versus a, uh, large carrier. Ah. How does that sort of transpire?

Speaker B: Yeah, there's. There. Well, the MGA's tend to be very focused. As you've seen, they've got a market area, whether it's geographic or lines. You get to the carriers. One of the things that happened to us is we walked in the door in a particular line of business and the carriers thought, well, that's where gradient fits and we'll just leave them there. And then we've discovered that, oh, by the way, in many of these carriers we can help them across multiple lines. So the difference when you go upstream is it's sort of an opportunity to land and expand. So we start to, we, we typically go to the line where they have the greatest challenges from a profitability perspective from the underwriting side, or if it's a claim, uh, opportunity to help them better manage the, the losses that have occurred. Um, and then typically once we help them there, we, you know, we ask them, so what, what else is going on here? It tends to be some other things happening in the marketplace that people will bring us over to those operating units and we start to have more conversations and start to expand horizontally as well as vertically. Because typically the first thing we sell to a client is new business underwriting. Uh, because that's typically all of them are trying to grow their top line profitably. Uh, but they also, the bulk of the revenue is on their existing book of business and they go through a renewal cycle. So we actually have a growing number of really sophisticated AI solutions on the renewal side to help them understand retention and other risks that, uh, will help them do a better job of pricing their existing book, uh, and maintaining and maximizing the profitability across that book.

Speaker C: And how does it work from a product perspective? Are you constantly sort of plugged into that conversation, iterating, coming up with stuff in the background to make sure that your product Runway is matching the needs of the clients. How does that look?

Speaker B: So, absolutely. So what happens is, if we go into an area that's not doing as well as they'd like and we help that it usually is like the domino effect. They say, okay, guess what? We have this other business line here. It's different. Here's how it's different. We need your help. Can you help us? And so we're actually expanding from the group underwriting that we've been doing for a number of years into a number of individual lines because our clients have said we have some challenges here. So Medicare Advantage, the US Government has shaken up that whole market segment, which, you know, so some carriers are exiting after having been very big players in that market. So there's a lot of, uh, uncertainty and a lot of, uh, volatility. And that tends to be where we shine because we can actually go to companies and say, hey, we can help you get through all this and still build a profitable business. And so we're sort of being pulled to the next hot area. Um, and they will tell us, your products don't work for this line for these reasons. And we're like, great. Uh, but if there's a reason for it to make it work, we will, and we do. And we tend to start adding those lines. And it just, we're looking, you know, two, three years down the road at more, more business lines that we want to get into because we sort of see how this will, you know, where this will take us.

Speaker C: Yeah. Are you able to speak to those lines that you're sort of seeing opportunities in?

Speaker B: Uh, well, so, yeah. So I mean, we've been group, as I've said before, and the part of the reason was when we were smaller group, uh, is. Is you don't deal with individuals, so you don't have some of the individual rules and regs that go on with, you know, just all sorts of things when you write individual policies for personal, you know, auto or home or whatever. So we stayed a little bit away that. Just from the regulatory weight. But, uh, in the healthcare side, we, we see a big opportunity to go on the individual lines. So Medicare, Medicare Advantage, Med Supp, and potentially other lines around ACA and possibly even going into disability and, or life insurance eventually, because it's sort of. A lot of our carriers have those lines as well, and that's that's sort of where we see them taking us is down, down those lines in that order. Would life sort of be the last one? Excuse me.

Speaker C: Yeah, it's a huge opportunity in that sense. What about internationally? How does that. If you look at the kind of the product Runway and the international expansion, is that something on the horizon?

Speaker B: You know, it's, it's funny, I've talked to a lot of folks that have entered the US market from either uh, Asia or Europe. And um, you know, they came to the US because it was such a big market relative to where they were. Um, you know, I mean, Europe's overall is a big market, but each country's, you know, on the smaller than the US side from the business opportunities. Um, and so it's really been hard for me to say, okay, we're going to go, because you don't go after Europe. You go after, you know, France or Germany or the U.K. you know, and then you have to build. I've done it before. I think what I've said to my board is that if we think we're going to run out of headroom growth space, uh, in the U.S. i think we should talk about it. But one other way to think about it is possibly an acquisition or merger with somebody who is European based or, or Asian based or some combination global. And there's a lot of those folks that, that are bigger outside the US Are in the US and so if we're trying to build a business for like an IPO or something like that, that kind of opportunity would be something we would probably want to consider. Um, still you don't have to be global. But I think if you want to build a category leading clearly, uh, dominant type of company, sometimes global is required because some of our clients are global.

Speaker C: Yeah, as you said, there's such a big market opportunity in the US and actually expanding line by line. But you know, there's also a huge um, M and a opportunity with some, you know, pretty established players that are coming out of the UK market, for example. So there's, there's always opportunity there to get, to get boots on the ground. Um, what, what do you see then at curiosity? Um, I mean, level with me. How far off are we from sort of fully automated underwriting and where do you see the next sort of three to five years in, in the insurance landscape going?

Speaker B: Well, I think there's been fully automated underwriting for a long time, but some of it's very basic like rules engineering. So a couple simple if then statements where, if it's X or you know, uh, above X and below Y, it's a yes if it's, you know, whatever. I mean some of that has been going on for a long time. You look at some of it's financially motivated just purely if you're writing commercial policies for, you know, hundreds of dollars or thousand, a few thousand dollars. Insurance companies cannot afford to have a person go in and look at each of those and review it for hours and so forth and so on. It's just, there's not enough margin in it. So you just either don't write that business or you have to figure out a way to write it very efficiently. And so the question really with AI is can you do it more efficiently and more accurately? Because I think if you can. We've actually done a proof of concept with a carrier where their internal team had built a certain straight through processing volume. But when they went above that percent, a certain percentage, their loss ratio started to get worse. And so the proof of concept was could we get them to a higher percentage of straight through processing without having the loss ratio get worse? And it turned out we not only could get way past the goal of the percentage they wanted to go straight through, but we actually improved their profit, uh, their loss ratio as, as by the selection results we showed them with our solutions. So it kind of hit, they were ecstatic that they could again put more business through but actually drive better results at the same time. So drive top line growth as well as improve profitability. That's rare to put those two together. I think the insurance world has seen, you know, when companies grow too fast, they tend to not be as selective about the risks they're taking on. In many cases, not in all cases, but in many cases. So the regulators start to get nervous when someone's growing too fast because they just uh, they conclude based on past behavior, they could be taking on a lot of risky business that they've underpriced and that those losses will show up on their balance sheet in the future. And that can put that company at risk and the policyholders at risk. And so it just, there's this behavior characteristic that everybody says, you know, decent growth is okay and safe, but too much growth just looks risky. So we're not trying to say what somebody's growth should be, but we think if we can help them grow profitably, it unlocks a lot of opportunities for our clients.

Speaker C: Yeah, absolutely. Well, I mean, profitability is what we're all here for, right? So you mentioned, um, uh, AI sort of actually breaches that chasm between, you know, profitability, uh, how does that, how does that kind of work from your perspective about sort of bringing that, those sort of two things together?

Speaker B: I think it's, it's interesting. AI, you know, we sort of talk about it, you think about with underwriting, tell me how I should price this risk. But there's a lot of layers you can see about a piece of business that isn't what normally an underwriter would look at. For for example, lifetime value. Can AI understand based on your, your company, who you are as a carrier and the market you serve and the brokers you have and so forth, is this a client that's going to be come in for one year? Is it, is it one of your types of policies that is going to remain with you for years? And maybe you should think about it differently if it's going to be if the lifetime value is a certain level. Uh, again, that's not a direct loss ratio prediction. That's an insight into this type of company, type of ownership, whatever combination of things the models have learned from your book of business that can tell you that you might be more aggressive with your pricing, um, or not that aggressive because this is going to be a high turnover type policy. They're likely to come in for one year and leave after that. So if you don't make money in one year, uh, in the, in the year you write them, you should write them. You know, that's, that's kind of the interesting insights. It's not about what's the price you can write this policy profitably. It's looking at what's the value of this policy to me this year and in the future. So there's a lot of these nuances that AI brings, not just around that risk that will make folks uh, able to again automate some of that stuff. Okay, so I want to maximize profitability, I want to maximize long term relationships. You know, I, I can, AI can be told maximize A over B or, or evenly or you know, a lot of those kinds of things. They can actually companies can look at how they're performing financially and tighten their underwriting rules, uh, very, you know, dynamically say, okay, my range has just got tighter because I, I'm seeing lots and losses in some earlier business I wrote that was, that's been unexpected, which can happen. So I just, I need to pull back, uh, on I need to maximize profitability, you know, for the rest of the year as opposed to maximize growth. And having that ability to dial that in has not been easy. In the insurance space because they've been a little bit, you know, sort of big ships and not that uh, agile. But I think AI allows them to be a lot more agile about things than they've been in the past.

Speaker C: Yeah, I always like to ask this at the end of uh, an episode, Stan. So looking um, back, uh, gradient, how you built it, if you were to go right the way back to the beginning, knowing what you know now, would there have been anything you would have done differently? You know, a market area you'd have entered into or something you would have done sooner? Uh, um, if you had the opportunity to kind of go back?

Speaker B: I think, you know, it's an interesting question. I haven't really, I've never asked myself that and I think my, my gut is I would have done everything I've done faster than I've done it and sooner. I, I think and I, I don't think there's many people that have started companies that wouldn't think the same way. You know, sometimes it just, sometimes it feels like it takes forever to want to get to. Question is if I had it all to do over again, I guess I'd want to get to where we are to now, you know, years ago. So I could be much further along today than I am just because the opportunities are, it's still massive. It's still about out executing the competition. There's lots of new companies coming up all the time. Competition makes you better or it takes you out one of the two. Um, so I think you just, you just, I think speed is still the thing I think about mostly just you know, going faster.

Speaker C: Yeah, absolutely. As also the fail fast mentality as well. Getting you to the um, the no's rather than the yeses and then learning from.

Speaker B: Yeah, yeah, brilliant.

Speaker C: Well, fascinating conversation Stan. I really appreciate coming on. Sounds like you've gone on an incredible journey. Um, and you know you've got an incredible background to really drive this business forward. Your passion comes across through when you talk about this industry and, and uh, the challenges that you're seeing and the changes that you're making. Uh, it's evident that Gradient is doing a fantastic job and having some real success and genuinely meaningful impact in insurance space. Um, and I really look forward to seeing what you continue to do and uh, the continued success of Gradient. So thanks very much for joining us.

Speaker B: Thank you for having me Dan. Really enjoyed it.

Speaker C: Thanks Dan. Cheers. Sam. Mhm.