How Michael Saltzman Built EvolutionIQ to Win in Enterprise AI | Ep 15

Is Enterprise AI adoption failing? You've seen the MIT data, you've seen the Salesforce pivot, but we think there's a big misconception here. The adoption is slow because technology isn't ready. That's just not true. The reality is much simpler. The Enterprise Vertical AI Playbook simply hasn't been written yet. When you're selling the 10,000 personal organizations, you need something a bit different. Today's episode is about what actually works in Enterprise Vertical AI. Deep, intentional hands-on iteration with real enterprises. And we'd argue today's guest is one of the most important authors of this emerging playbook. Michael Salzman is the co-founder and co-CEO of Evolution IQ. Evolution IQ is one of the first vertical AI companies to truly break through an insurance. Evolution IQ today augments claims decision-making and big insurance companies and carriers where even small accuracy gains compound into huge financial impact at scale. A $750 million exit and one of the first breakout outcomes in Vertical AI. So let's get into it with our friend, Michael Salzman, co-founder and co-CEO of Evolution IQ. We're super excited for this episode. Let's dive in. All right, welcome to Verticals, everybody. Thanks for being here. We got an incredible guest today. Got Michael Salzman, who is the co-founder and co-CEO of Evolution IQ. If you don't know Evolution IQ, you should. Because they're probably the first major vertical AI exit that we've seen. And they took a really unique, really interesting approach to scaling, serving mega, mega companies. And we're going to go deep on all things, product market fit, serving the enterprise, product, vertical AI, what it looks like an insurance. So it's going to be a fun one. So thanks for being here, my friend. Awesome. Glad to be here, guys. Cool. Well, let's just kick it off with if you don't mind. Maybe a little bit of your background, like the career arc into eventually founding this company, what made you take the leap? And then we can just dive into the early days. Sure. Yeah, so I appreciate you guys having on this. We fun. So I was an engineer and undergrad, mechanical engineer, knock computer engineer, which probably would have been smarter. I went to Bridgewater Associates after school, which was a big hedge fund. I was there for several years. Had an awesome experience. It's a wacky place if anyone's read about it, but really fantastic learning environment. And besides learning how to think and build models and kind of process data, more specifically, I spend a lot of time in their credit team, which had me focusing on insurance companies and banks. So I got to know insurance companies from a operational standpoint, financial standpoint, a not inside the business, but how they talked about their own businesses, how they reported their performance, et cetera. I left Bridgewater in 2017 to go to Stanford to do an MBA, but kind of took with me a lot of the lessons from Bridgewater, especially around the insurance world. I didn't go to Stanford thinking I have to start a company. A lot of people do. I didn't have to be one of them. I was open to it. I wanted to do something different. I liked my job at Bridgewater, but didn't see it as a multi-decade career path. I wanted to do something a little more active, a little less kind of researchy. So while I was at school, I had a close friend named Tom, who was not a Stanford, a friend from New York. We all met flying airplanes. He's a pilot, a pilot. And we actually met on the tarmac of the East Hampton Airport. And he was at Google doing AI stuff for about 11 years, and he wanted to leave and do a startup. And he was interested in the insurance world as well. He was kind of interested in any industry that had lots of unstructured data both having trouble using it. I was interested in building a startup perhaps, certainly knew a good amount of the insurance value chain. And the basic idea was, you know, if you read a lot of insurance company public statements in like 2014 to 2018, so kind of this time frame, like I did, a trend you would realize is that the CEOs of the companies were talking about technology transformation. They were making big investments in technology, spending more and more on it. The CIOs were talking about the use of data across their businesses. But really, all of that investment was going into distribution and underwriting, meaning pricing the policies and selling the policies. The claims operation was like the ugly stepchild back office. And claims was like, you know, if a claim happens, our call center handles it. The, you know, it's cost doing business, just kind of be as efficient as we can get through it. And just Mike, for listeners, what do you mean by claims? Okay, someone doesn't know anything about insurance. Great, yeah, okay, good question. So you buy insurance policy and it covers workers' conversation. If you get hurt at work, you know, someone should cover your medical bills, your, your loss income. And so if you ultimately get hurt and hopefully only a small percent of people who buy the insurance never do, if you ultimately get hurt, you file a claim. The insurance coming in has to review the claim, get your documentation, talk to your doctors, figure out that you actually got hurt, you know, on the job, do all that kind of stuff, ultimately pay your claim. And what we realized was, you know, there's a big gap between saying we're going to transform our businesses, but not spending a lot of time where 80% of your revenue actually goes. So 80% of insurance carrier revenue is paid out in the form of a claim. And we're trying to optimize the other 20%. And our thought was, well, there's actually tons of unstructured data in claims, a claim is unstructured data. Its correspondence, its documents, it's all this kind of stuff that is hard, especially in 2018, very hard for computers to process. It's challenging for humans to process. You got to read endless documents. And so what we thought is, look, the reason they're not innovating with technology in the 80% is because they don't think they can. For two reasons. One, technology, there isn't enough technology that can process this type of data, unstructured data. And second, you know, they think it's a, it's the claim outcome of set and stone. It's hard to impact, possibly the claim, the goal is just to do it efficiently. We had a different take, which was, if you can put guidance and intelligence into the hand of the person working the claim, the, the, the adjuster, offering to Charlottes, they can actually do a better job, help someone recover faster, avoid longer, you know, medical recovery times, or, you know, whatever the insurance is, ultimately help the claim actually cost less and get someone back on, on their way and their life. And so that was kind of the founding idea. Get into an industry with massive unstructured data, the industry was struggling itself to kind of build systems to use it. Really understand the problem? Like sit with claim adjusters, understand literally what they do, where the challenges are. Don't be arrogant thinking that, you know, they don't know what they're doing. They certainly know what they're doing, but can we help them do it better? What, what are the challenges they face? And then ultimately build technology to actually change the cost curve of the cogs of what are the largest economies in the US. So the cogs are that 80% of them. So if you think about like, if, if you just look at the standard kind of P&L of, of, you know, a commercial, or not commercial, just an insurance, you know, conglomerate, how much of their costs are tied to paying out claims? So, right. So like, you know, let's say they take $100 of premiums. So depending on what I've business, between kind of $1680 of that 100, are going to be paid out to the claimants, like actually paid to a person outside. And another $15 or so dollars will be spent on the claims operation staff, all the skilled people who are doing that job. Then they got to go sell the insurance. They got to market it. They got to buy super bowl ads. They got to buy skyscrapers of Manhattan, all that cost of money. And so interestingly, a lot of insurers actually break even on selling insurance. And the idea is to invest the money and make a spread between the time they receive the premium and they actually pay out the claim. And like, that's what actually makes or breaks their business or is a significant contributor to it. So no matter how you slice it, the claim cost is by far in the way the biggest lie item in an insurance company's, you know, Pino. And so, were you guys scratching your head in like this time period when you're doing the research? Not just on the unstructured data side, I totally get that massive opportunity. But you're like, wow, if we actually move the needle here, a percentage point or two percentage points, like we can fundamentally transform how these businesses make money and not just tie it to the investment side, 100%. So if you think about like, you know, the P&L stack we just talked about, let's say an insurance carrier is like a really good insurance carrier and they run what's called a combined ratio. So expenses and claims of 96, meaning $100 in premiums, $96 or we pay out an acclaim order on their business, $4 in profit. At that time, so 2018, 2019, there was tons of attention on the distribution side of insurance. Tech brokers, tech agents, better underwriting, better marketing and targeting. And all this was in service of selling more insurance, which is not a crazy endeavor that makes sense. But if you sell an extra dollar of insurance, you make an extra four cents, right? So that, you just scale the existing problem, bigger and bigger, you make a bigger, bigger insurance company. Nothing wrong with that. But that was basically where the focus of the industry was at that time. And what we said is, okay, but if you save a dollar in claim cost, you save a dollar that you increase your profit by a dollar. That's, you know, it's one for one. And so the leverage on an industry running at a 4% earnings market is just tremendous if you can actually attack the cost side of the business versus the revenue set. - Which is interesting because I feel like most founders in investors have a bias towards initial strategies that are growing revenue. but. in a world where a lot of the customers have incredible scale already, that first year of impact can be wild. You guys were starting more pre-LOMs. How did you think about, I guess, drilling down into the product? How did you think about how you could move the needle? Yeah, so a couple things. First of all, just about the team. One thing that we really started with, and Tom, my co-founder, is a part of this, is we knew that this was actually going to be a technical product, and technology limitations would be part of our constraint set. The only reason I mention that is, their most insurance technology is not at the edge of technology. We're going to build a CRM, we're going to build this new workflow tool, or that software. We don't have feasibility concerns over, can we build a CRM that you do a drop-down, put in someone's name, and tomorrow, it's still there. That's not a feasibility concern. We went out and we said, "Look, we are going to have a feasibility concern. We are not sure this is possible." At every point in time, the one thing we knew about technology, even then, it's not like LLMs are the first technology to dance, we knew that the ability to process on the structured data at the edge of technology would get better. So what was most important to us was to build, especially in early team, but even today, but it starts with the early team of truly incredible technologists, like top, top, top. For example, our first 10 hires, I was the only person talking to customers. We had one product manager and four X-Google employees, two of which were PhDs. Pulling that out of Google was not an easy feat. It was not just a money thing. It was, let's go totally solve this industry thing. There's a lot of that that had to be done. What we knew is that, however feasible, processing deep-limstructured text data, etc. was in 2019, we had to be the company that would always provide the leading edge of that to our customers as the leading edge evolved. So that's the first thing. And then in terms of how we built, our products at the time focused a lot, not just on predictive accuracy, which they did and still do, but we focused a lot on the human component. So one of the really important insights we had was and this is applicable for a lot of industries, not all, but a lot is we decided not to build an automation company. And a lot of other software in the space, whether it was vertical or horizontal, horizontal like, you know, UI path and companies like that, I think you're compiling. And companies trying to more specifically serve the insurance world, they were just trying to automate stuff, which is fine. It's great. There's a lot of stuff to automate. Two issues that we saw. One, that rock had been squeezed a lot. You know, the first version of automation was, don't automate, just make it cheaper. So offshore stuff. Then these coming, went into their offshore team and tried to automate the stuff that the offshore teams are doing. And so like, you were just kind of at the tail end, just like squeezing and squeezing and squeezing. Technology was getting better, but like basically that was a problem that had a lot of attention on it. And the second was the biggest lever in claim outcomes was not automatable. And that was the expertise of the human claim adjuster when they apply that expertise and what do they do on claims. And we didn't think you could automate a building rapport with a claimant in a really hard moment in their life, building trust, helping them navigate the US medical system, and ultimately giving them the motivation to return to the workforce. So what we decided to do is augment the human expertise. And what that means from a technology standpoint is you don't just have to be predictively correct. It's not just that you're saying, yes, this claim can resolve today or if you do this action, the claim is better off. Someone has to then go do that or else there's no value created. And so, early in our product life cycle, it's still very true today, we had to really balance and build expertise and a real motion around not just being predictably accurate, but being compelling to a human user that they would actually take an action differently than the otherwise might have. Or they would look at a claim that wasn't scheduled to be looked at for another 30 days or whatever the thing was. And so getting both of those tight together is the magic of our product. And then being able to show up for a client for an insurance carrier and educate them and work with them on how to transform an existing workforce into an AI-augmented workforce. In a way that's not threatening that is not antagonistic. It's no, we're here to help you do your job better and spend your expertise where it can have the biggest impact. We're not here to allow your job away. We're here to put you in a position to have the biggest impact. And so getting those three things right together is I think what made the magic for the product. So given there had been a bunch invested in this RPA layer, did you make a conscious decision to be less app player forward? Like did you have to integrate with a lot of stuff? Or because I like the flip side, that would be, hey, we've got a chain-jogment, human behavior. We need to have that UI. We need to be that everyday login. We struggled with the strategy run integration for a while in their latest. And the easy answer, which everyone was telling us right at the beginning was. So they've single core systems, they're the CRM, they're the sales forces of insurance carriers. And all the claimant, it guy wires the biggest one. And that category, it has others, finios, and claimantage, and all these companies that have been around for a while, you're very kind of key building block and have their companies operate. But they're not exactly changing the world in terms of the edge of AI usage. And so we had a conundrum, which was, everyone was telling us in the industry, oh, we got to integrate with these systems, where I'll see no one's going to use you. And there's some truth to that in terms of it, probably would be easier if we were natively inside these core systems. Because look, that's where the eyeballs were, that's where people were trained to use, when they started their day. The problem is, we were on a development path that was so rapid and so iterative, and so responsive to not just customer feedback, but where we wanted to take the product, that locking ourselves into their box and saying, okay, we're going to give you these fields, can you please display them in your core system? Introduce tremendous bottlenecks in our development cycle. We didn't own our destiny anymore. We had to sit beside someone else on their bus, and we had one seat on the bus, and we didn't drive the bus. And so what we basically decide to do is, for most of the history of the company, it's changed now because we have a lot more market presence and influence with our customers. We didn't integrate with these core systems at all. We built our own standalone application, which in my estimation was gorgeous, and of a quality and usability that far exceeded legacy CRMs. It was intuitive. It didn't take a lot of training, unlike these huge model FX CRMs. And our users liked to beat it, because it did a really specific thing really well. It oriented them to where they should spend their time, and it helped them navigate a super complex claim. And that's all it did. And so we built a really elegant application, and there was concerns about, you know, I'm too screening, or I'm swivel chairing, or whatever. And we entertained those concerns, and we said, "Look, that's the negative. The positive is you are going to get an application and a service that gets better every month. And if we go the kind of fully integrated way, it might get better once a year." And once it had that way, it was not a very hard discussion with clients. Were there any challenges that that presented in terms of getting the right data easily? Like, what would an implementation look like? So what we basically learned is getting data out of these big systems can be a little challenging, you know, depending on exactly their age and all that. But it's not an impossibility. I mean, they have millions of different reporting requirements. They do getting data out of these systems is a thing that happens. And so that wasn't the challenge. The challenge we had to deal with, which is a real investment of technology that we made, is we wanted to build a product. We wanted to have a real product-led architecture. We didn't want to be building a different solution for every customer. We wanted to solve certain types of problems and solve them across the customers. And so what that meant for us is we wanted, especially in our AI layer, we wanted to build models the way that we thought were best. And we wanted to find universal truth around how to interpret medical documentation or how to interpret, you know, this type of data or that or whatever. That said, we also had clients who had meaningfully different data schemas on their side. And the way they had set up their core system was totally different because someone did it in 1997, someone did it in 2006, someone did it in 2012, someone used Accenture, someone used Deloitte, it's totally different. Today I'm, it's like, well, this is just what it looks like for me. I don't know how different it is from your other customers. I don't really care. And so what we had to do is we had to build an interface layer and that was a real technical project. We did this in like 2021 and it still is, it kind of underpins the, the scalability of our company that we can take in data from all these different core systems and we can normalize it into one common data format at within EIQ. Downstream of that data is still segregated because we don't, you know, there's a lot of not commingling upper comments we have, but it looks the same. And so we can basically run a model on that data, run it separately, but equivalently on that data, on that, on that, on that, make improvements that go across our stack. And so, you know, normalizing their data and building technology to do that. was kind of the way that we It almost feels like a digital twin kind of approach Like you take the data set that they have and then you you basically run on it But you're not you're not necessarily integrating and updating fields in the CRM and causing Disrupture interesting so we did we didn't we didn't write anything back to the theorems metal You know, we had a digital twin, but we it was you know in our format and all of our tech ran from that So well, what was the I mean obviously the overall goal is you know, we're gonna produce your expense ratio increase profits All that was there kind of a Near term like time to value metric that you guys worked against So I think one of the things that insurance is known for and insurance technology is long sales cycles And there's real truth to that these are large organizations. They have certain decision-making habits They you know, they're different. They're not all the same But you know, they're in in general is there's different stakeholders. There's approvals. There's budgets and all that and so what we You know really endeavor to do is to demonstrate the impact measurable impact of the solutions within the first year of someone implementing us which did a couple things one it meant that Our expansion motion could rely on the results of the first thing not just selling harder You know, it's it's hard sell someone the next thing when the first thing takes four years to bear fruit The second is because you could see the results the first thing could literally pay for the second thing Many times over and so we had this notion and we still do of like when you work with evolution and key We're a modular product and so you can start you know in different places and some clients start with multiple modules some start with one that's all great But the modules you have Once you implement the first one you're basically you know from a financial standpoint you're playing with house money The first thing you bought from us Pays dividends and you take a small portion of those dividends and you could buy the next thing the dividends then you know You know grow exponentially because you're kind of doubling up And so to super highly levered sales cycle to the results the results are super super important and so we we really in our product Romance we only took on problems that we thought we would have a measurable impact end within it within air how did you test that advance? So like even even a year is a pretty long feedback cycle from a product development standpoint So you know where you guys kind of doing you know dummy data tests in order to see. Hey what module do we start with? Well, yeah, so there's a couple different like flavors of results, especially insurance world that is Actuarially driven that you know they study their statistics a lot so when I say results I mean like you see it in the financial outcomes. There's a lot of leading indicators You can see to make sure you're on the right path right so you can look at cohort analysis You can look at month by month changes in activity So we were like maniacal around we wanted to know if our systems were working or not as soon as humanly possible And correct and change or you know say yeah, this is gonna be awesome and let's get our clients ready to Understand you know how we're measuring it and all that so you know It didn't take us a year to know if it was gonna be happier sad It just took a year to like literally show up in the numbers or sooner sometimes it was six months seven months I mean insurance world that's like lightning fast like remember the previous generation of technology project and insurance Work core systems that cost a hundred million dollars take four years to implement and by the time you've implemented them It's time to buy the next one and You know when you talk about a CRM like what is the Salesforce for example? Incredible product company, you know obviously ubiquitous Ask someone who uses Salesforce even someone who loves Salesforce. Where's the value of their Salesforce relationship? Very hard to answer they they're kind of like well. I can't run my business without it valid What how many what is the dollar value of Salesforce to you and so we didn't want to have that problem We wanted to be able to say here are literally the value of this relationship and It's a dex what you spent on it and if you take an aphid you can buy the next one It's it you know to be super super tight to the actual value group. All right We're gonna take a short break to give a big shout out to our sponsor paraphernal They're the ones who make verticals possible if you haven't heard of paraphernal go check them out They build embedded finance tools that really help your merchants grow They're the leader in vertical software and embedded finance and so if you're building in the space Haven't explored paraphernal yet go check them out. They've got embedded credit embedded spending a lot of different tools to really drive revenue growth Drive tam expansion and really improve retention. So big shout out to paraphernal go check them out now back to the episode So just to over like I'm gonna oversimplify it But I want to just try to set the stage of of kind of the product suite that you guys you know had and have today so A good way to think about it is almost like It's basically like a co-pilot for the claims professionals and so you you create this twin and this separate You know web application that the claims folks are really living in and basically as they're Receiving claims your technology is like analyzing them helping them think about you know ways to Respond, you know ways to go this path or that path that that you know ideally benefit Both the claimant but the company and and and not dispersing cash based on like human intuition But like real company policy that's written into the technology and then what so that's that's kind of your let's call it You know product a or like what you get in the door with you prove that that pays eight times over after the first 12 months What are you expanding them into after that? So it's a good it's the equation So if you think about the types of claims we solve They're they last for a while. It's not like you know an auto vendor vendor where it's like give me 800 bucks And like we're good and do it you know instantly these are complex bodily injury claims. So our product suite Basically aligns to the length of the claim but the stage of the claim that it's in the types of activities done there Earlier, you know younger claims more complex, you know longer duration claims Because the type of work going on on those claims changes a lot You know earlier less severe claims will have a certain workforce in the claims organization working it Higher complexity claims maybe they have a lawyer attached to it if it's in the you know auto or casually world In more sophisticated more senior adjuster will work it so we've kind of split our products Mostly to mirror the organizations we serve so you have a workforce doing this type of work on this type of claims Great we'll build the products we that that they need you have a different workforce working on these types of claims We'll build the solution that they need They're basically land and you say hey carrier like take a shot with this one claim in your organization We're gonna prove it and now you have 25 other claim types to expand into you know, and sometimes it's multiple if they say hey, this is great. Let's let's move them to all of them got it interesting Well, what was that so the kind of that period from founding to 21 when you decided to have this kind of middle layer ontology Because I guess what I'm I'm trying to like Extrapolate a little bit to other very heavy enterprise vertical AI And I think there is this pull to say like hey we're in this brave new world where we can abstract all this data But doing everything and putting everything into that perfect ontology upfront is like a slow process. So like how did you I guess how do you navigate this first couple years? Um, so you know The insurance world is is big in a lot of ways. It's small in other ways. You know, there aren't 50,000 insurance carriers in the US And so for those first two years I don't I'm afraid exactly the breakoff point, but like we had one customer right? So you know and their offices were in Philadelphia and ours were in New York and so we were in their office like many times a week And so you know, I think we started relatively small like we weren't trying to shoot the lights out We were trying to solve a specific problem for that specific insurance carrier and You know, I think a lot of the choices we made There's a way to tell the narrative now which is that they all lined up perfectly and like look at this great architecture we built But like at the time it was like well Their data is crazy and messy and comes from seven sources And yet we want to build an eye around it. It's like what do we require and so okay? So we require that like translation layer So like it was built out of necessity in the moments much more than any grand strategy now when we got to you know six seven customers which are we started to invest more in you know pre-emptive scalability for the sake of scalability But in the early days with one or two customers Like that would have been a mistake and the most important thing was to build the ultimate product for them That radically changed their economics their ability to serve customers their ability to control their costs keep insurance affordable That made them and this gets into the commercial side made them just like incredible sponsors of ours And willing to talk to their direct competitors and say how incredible we work because they wanted us to stay in business And so like that side of it was just much more important than the perfect architecture side as long as The architecture was good enough to solve that customer's problem and so you know We never had a CFO thank you know It's probably a good thing because we rewrote our product like six times And so it's not like we said okay, if you know we're going to build the perfect hyper scalable AI platform on day one We really start investing in scale in my year free. I can only imagine the VC conversations right now in the first 36 months because You know you guys buck like every quote-unquote vertical software playbook right which is like wedge product expansion blah blah blah like I'm curious done to just zoom out for a second and kind of think about vertically, I, in the enterprise and how you guys kind of like swat teamed into this, built it hand in hand with them for a few years. Like, at what point? Yeah, go ahead. - Yeah, just like maybe to start, 'cause I think that's great. Let's like transition into this enterprise AI playbook. How did you get customer one? - Yeah, good point. - I mean, it's not rocket, we're cold calling. And I had a network from Bridgewater and I knew some people, my insurance were all to kind of banking insurance companies and we were networking like, you know, like our lives depended on it, basically. We had a lot of people who actually did depend on it, Michael. - He did, I'm like, "I'm not a weird person." - I'm not a weird person. - We had a lot of people who said, "Look, this is really interesting." You guys obviously are like three dudes, a PowerPoint and a dog, like come back when there's something I can use and work on and that was mostly a nice way of saying like, "Okay, you guys are not real." Which we weren't, top still worked at Google. I was at Stanford, like there was no company, we hadn't incorporated the business yet, he hadn't left his job yet. And then we met this company, you know, look, it comes down to a little bit of luck. We met this incredibly innovative company in Philadelphia and they said, "When can you start?" I am like, "What do you mean?" They're like, "Yeah, we want to do this." We're like, "Do what?" They're like, "Well, this vision you're talking about, what's the first step?" We were like, "Okay." And so, you know, honestly, there's a lot of, you know, pounding the pavement and then we got this counterpart with this insurance carrier that, you know, you're talking about us breaking them all, they broke them all, they were innovative, they were happy to be a first customer, they knew the product didn't exist. We didn't hide it from anyone. They said, "We want to build this with you." I think they saw us as a high leverage, high skill tech team that at worst would waste $50,000 of their dollars. And at best, would sit in their office, dedicate themselves for years and build something that they couldn't buy. And the latter is what happened and they took the risk on the former. - And was it just, hey, $50,000 and we'll see where we are in a year or is basically exactly that? - Yep. That's, yeah, I mean, next natural question for me then is, how did you guys know when it was ready for customer number two and then, you know, two to six? - So we didn't even try to sell it until we had results in the first customer. So talking about like bucking the VC model, like we were a deeply unsexy company for the first like three and four years of our growth. Credit to our early investors, first round being the largest, like they were totally cool with that. And Villa first round is just a savant investor and company builder. And he was happy to not look at the revenue, he was happy to look at the product and look at the results, like the claim results of the carrier. He wanted to get into that level of detail to underwrite how he was thinking about our business. And so, you know, the just the priority was just build the kernel product that is just awesome. And if we do that and only if we do that, do we have the right to talk to other carriers better? - That's amazing, 'cause it's so easy for founders to get kind of fall in the trap of like quarterly numbers. I know I did it and sometimes, you know, as a CEO, like you take shortcuts to hit the quarterly number that cost you, you know, over the long haul. And him being product focus is super rare, even in early stage. I mean, our problem in the VC world, you know, God forbid if we start the company today, like, you know, if you're not going zero to 10 million in your first year, like you're not a Sukeh egg, I'm afraid it's like, you know, we're gonna see how that plays out. But, you know, we were a slow burn. And to our credit, we were not very capital intensive, we never were. We were always like, we always basically wanted to run this, like, you know, it had to be a real company, not like a venture, you know, flash napam. And so we invested where we saw progress. And the issue we had was, even we had one product, you know, in one kind of segment of the insurance world, you do the math, okay, you have one product, there's like 40 carriers like this, you're making X from this carrier, they're kind of upper mid size, like, ooh, your market's not that big. So the market size thing was like our, you know, the knock on us through series B, honestly. And then we started to say, well, yeah, if you have five products that are all as good, and they're actually really good, better than we thought. And so the price we can charge is 10X larger than we thought. And we're gonna go into multiple lines where there's hundreds of carriers, like, then the business is like really venture-scale and interesting. But you had to have investors when you go on the journey, when like, it wasn't obvious it was gonna be. - Did you know that upfront? Or did that, where you just like, hey, let's, this is a big problem, we'll figure it out. Like, did you, did you feel like you could, you could ultimately get there? Was it just so, you know, one day at a time? It was more like one day at a time. It was like, yes, we wanted to go build an awesome company, but we didn't set out saying like, we're building a stripe or nothing. We were like, let's go build a business. And there's like a business can have lots of different sizes. We wanted to build as big of a business as we could. We wanted to help as many carriers as we could build great products. But like, on a random Tuesday afternoon, I was not thinking about that. We were thinking about like, how can we get this customer to think this product is great? And then talk to the next customer and the next one. And so it was, it was an iterative process. It wasn't like, we were not always working backwards from how to build a decadent. - You guys like nailed the Peter Teal, monopolize the tiniest possible market ever, move to the next one, kind of these. - Yeah. - Like, so I, I think the market is even that tiny in comparison to a lot of like, vertical software winners that like, prove the constrained VC-Tam thinking wrong. - Well, the market is small by number of companies. - Yeah. - Yes. - The carriers, right? The several hundred is not a big number. But, you know, these carriers are financial institutions. And if you move the needle for them, there's a lot of leverage there. And so that was really interesting. And, you know, we weren't just like vertical. We were like vertical of vertical. So like, vertical is like your insurance specific. We were like, you know, we're not in terms of, we're claim specific. You know, we're not claim specific. We're disability claims insurance specific. So we were like, really trying to get into the nuances, which adding is why the products were good, 'cause we were doing something very specific. And even with the net, it was like one segment to start as we talked about. So like, we were really trying to find like, the most specific place. But then, in a very leverage rich environment, like the impact of that focus is large on, you know, the insurance company, you know, I mean, you could look at public insurance companies talking about how this little company from New York like changed their net income for the quarter because it was such a focus on a high leverage base. Are there any comparables that you can think of of companies that have taken similar approaches to kind of crack the end? Like, the only one that is top of mind for me is like a volunteer. It seems like there may be like an open Gov's, Act Bookman where that's kind of the approach that they took to really crack these mega institutions. I don't know. You know, I think we have, we have similar, there's a lot of companies. I think there are differences. Our biggest difference is, I think, just kind of categorically speaking. It's like the focus area we had, had just obscene leverage on it because it was changing the cost structure of a product that had 4% margins. And like, if you can do that, you start moving the overall business really quickly. And that's the most unique thing that we, from a strategic standpoint, that's the most unique thing that made this work. So, well, was there ever along the way any sort of friction from, you know, kind of the incumbent systems or record, if you will, like, or I guess down the line, would you guys have run into a Godwire or like, would they've gotten sharp elbowed if you got bitter? You know, we had to build relationship with those companies and some relationships took off and we have a great relationship with Godwire now. We know they're executive team super well. We partner in a lot of ways, similar with some of the companies in other lines. You know, just depending on what they were doing at the time, I think, in the time frame that we were kind of really growing and expanding rapidly, we were almost growing too fast for them to process and what was happening. And they, all those companies were undergoing big cloud transitions at the time and they had a lot of wood to chop in their own backyard. I mean, they had the businesses and our business was, you know, growing crazy fast, but off for kind of a small base. And so it wasn't obvious to them that like, we were the thing they needed to do next. Like, you know, they are horizontal insurance speaking, but they're doing just, we're doing underwriting there. So they're like, we have a hundred different priorities coming up next. And so we, we didn't, with some exceptions, but we basically didn't feel real competition with them. We felt a little bit like competition for, you know, the air in the room when we're talking about exciting insurance technology stuff, but we didn't feel like we were, you know, building-- - Wasn't direct, right? Got it. - So if you were a founder starting over again, with a similar kind of eventual high-eCV enterprise sort of approach, would you do it exactly the same or, you know, in your mind, what are kind of the non-negotiable is that playbook? - And how do you actually measure product market fit with that? - That's a good question. I hope I wouldn't be driving the same because I think I, we made a lot of mistakes and, you know, found our way. - It wouldn't work out pretty well though. - Yeah, you might. (laughing) - I think, yeah, I think we pretty, like, intellectually lazy to say like, because it worked out there were no mistakes. There's like billions of mistakes. I made mistakes today. So I would do lots of things differently, but I think the things that are non-negotiable that I wouldn't do differently are I would spend the time with the early customers. You know, if the customer concentration is one where there's 30 in the world for the place you're starting, maybe one is right. If you're serving laundromats, maybe 20 is right to start with, you know, whatever. But like, basically, I would continue to go crazy deep with my first customers and basically tell the BC world pound sand about how far my revenue got in the first year. I just wouldn't care about it. I think I think like first year revenue growth is a really shitty predictor of like long term business scale. The other thing that we did is we stayed with founder led sales for a long time and we were in no rush to hire a sales team. Even today our sales seems really small, really sophisticated, senior, fantastic, like client advocates, sales, but it's not big, it's like six people. And I think that also, again, that's a little industry specific because like insurance, you can sell big relationships. You don't need a million sale people. But in general founder led sales as far as you can possibly stretch it, I think is really important. There's no better learning than selling. Both for the product and also just how to sell. So that would be non-negotiable. And I think the team, like I think, you know, we talked about this when we were last together with you guys, you know, finding the right mode is super hard. One thing that's a given is that the technology available is going to continue to change. Like as fast as things are changing now, this is probably as slow as they will change. It's not going to slow down. And so having a team that, you know, knows how to hot swap technology and knows how to like keep a product on the edge is really important. Can we talk a little bit about customer success too? Because this had been really freaking hard from a implementation. We talked a little bit about time to value, but like was it basically founder led customer success too or was it like kind of forward deployed engineering customer success before that was a term? You know, and the early days everything is founder led because they're only founders. But like in general, so okay, so the problem to solve is our product, not automation. And so the only way it drives value is if a large workforce that doesn't know who you are or why their boss bought this thing uses a technology that coaches them on how to spend their time. That's a pretty hard set up. Okay, that's not ideal, right? Like a lot of people and these the folks in this industry have just years of expertise, training, good outcomes, their companies are profitable. Like where's the problem here? And so building a deployment motion and the client service motion that partners with organizations like that builds real trust up and down the seniority level. So like important for us is we get buy-in from the chief claims officer all the way to like the frontline claims operator and everyone in between. And so yeah, that was hard in the early days. It was, you know, I was and still am very involved in it. But you know, I think what endeared us to our customers is our product team was involved in it. Our engineering team was involved in it. Like our CTO, Karan, who's incredible. You know, he knows claim handling as well as a 30 year claim veteran in like four different lines of this. Like he could literally be a claims manager. He's in the content. He's also in the tech, but he's like, and he, you know, he had no insurance back on before. He's a pure technologist. You know, he led machine learning teams at Bloomberg, but he liked to spend the time. And so I think we earned the respect of our counterparts because we showed up both with humility, also with a great, and you know, maybe also very importantly, we're going to be able to work with appreciation and a willingness to learn their business. And that I think started with the founding team, not just me, but Karan and others. Benjamin or she product officer started off as a product manager and then grew to the whole client service team, which was like, that's the ego. Like we're here to help. We're here to learn. We're here to figure out how exactly our product is. We're not a consultant company. We're not going to build it new for you. Like how our product fits into your process and what's a little bit of like a process interlinkage we could make. You know, that level of care was super important. Was that a weekly cycle at first? Like when it was just you guys with that first customer, how did you set up that interaction? More than weekly. I mean, we were in Philadelphia probably three days a week in their claim, like this pre-COVID. It's like in their claims call center floor, you know, in a building on Market Street in Philadelphia, Gia and TJ were two claims managers that we worked with and their staff. And so like we were like, we had desks sitting behind them. We didn't know what we were doing, right? So that's probably not the most efficient way to do it. Which is why we had to eat there so much. But after that we started to really understand like what is the leading indicator of success? And a lot of it was adoption, meaning are people using the system you could observe that in an app? Are they agreeing or disagreeing with the recommendations you could observe that because we actually asked them for feedback on every individual recommendation in real time. And then you can see in the data like are they taking actions at the system, you know, recommend. So we can really quantify the level of that relationship. That came later, you know, what came first is like asking them do you like this or not? And better questions. But yeah, it was a frequent motion and now we've built a real process around it. So there's some efficiency to it. But we're still a super high touch technology. Like we're not like we hooked up the pipes. It works. We'll see you next year. It's like we still are working with our clients we can make up. Zooming out a little bit just into kind of the market broadly. Like where do you kind of see opportunities and insurance today? Like where's their surface area? How do you view, you know, AI and plugging in like these broad tools to try to solve some because you guys built a lot of this technology that seems pretty similar before it was, you know, called an LLM. So LLMs are incredible and they super charge a lot of what we were doing and they enabled us to go faster farther into kind of the product core that we were working in. So the things LLMs are great at is they're just incredible at data extraction. Like before LLMs we had to build all this super complex, you know, basic classical ML models to understand what a document's at. You know what people call, you know, NLP. And it was fine. But it was getting better. But it was just, it was not like a human read the document. It was like a machine read it and it kind of, you know, it was a little bit here, a little bit there. LLMs are great at that. So what we use LLMs for, you know, across our stack, but a big part of it is pre-processing. There's a huge amount of extraction technology that we built to understand all different types of insurance data. The other thing that we use LLMs a lot for is guidance. So like we can write guidance to frontline claim adjusters, you know, much richer way that we used to. And so we use it across the stack in different ways. You know, where I see a lot of surface area in the future besides just lines of business and stuff, but from a product standpoint, an x standpoint, you know, there's going to be a real opportunity for an agentic layer to push further to actually doing the thing that the system says should be done. And so still the role of the human claim operator remains robust, which is trust building, partnering with the claimants, understanding motivation, dealing with different stakeholders if they generally talk to the employer, all these, you know, super complex things. But like can an agent ask a medical provider for an updated like, you know, medical treatment plan? Or does a human have to write seven on answer emails on a row? So like there's a lot of room for the an agentic layer to take the feedback from our system and then try to action as much as possible and then kind of tee up the claim operator to like really have a strategic next step in the claim. And so we're super excited about that. I think agentic has a huge opportunity, both inside the claim organization, dealing different stakeholders and also kind of going outside the building and talking to different stakeholders in the claim. I think versions of that exist across a lot of insurance. A lot of insurance, whether you're buying it or somebody a claim, is about assembling records to prove a thing. And everyone's record looked different and how you get them and where they come from is different. And right now there's a lot of people sending a lot of emails. And so just problems like that that can then feed into a system of intelligence that can operate the high value thing, which is what did you next about it? I think that whole space is, you know, very, very wide open. And only now becoming technologically accessible. I think your stories get inspired a lot of founders, specifically Vsass founders, to go attack like real enterprise problems because I feel like most people have just shied away from them forever. Like there hasn't been a really strong clear playbook. And obviously it's, you know, it's, I can't even call it a playbook. But you know, I love the hands-on, the SWAT team, like the sit-with-on and really, really solve the problem. It's awesome. I, you know, I appreciate it. It's surprising to me that more people don't go into the enterprise like something that's scarce so we were at your guys conference, you know, two months ago, whatever it was. And there's a gentleman there who serves long-dramaths, which is awesome and this is fantastic. I was scared of like, how do I get feedback from 40,000 SMBs? Like how do I have, like how do you, I'm, she's doing an incredible job and a lot of companies do that. You know, the benefit of the enterprise is as the founder, you can have a relationship with your customers. Know them. Know their motivations. They know yours. You can like really strategize together. I think it's hard to strategize with 40,000, you know, stakeholders. But in the enterprise, you can, you can build relationships, especially in the early days where you're kind of fledgling and getting going where, you know, it's if you do a good job, you have to earn this, not given. But if you do a good job, you can have people who run large organizations who are rooting for your success and are willing to like, give you that marginal piece of feedback, tell you what they need, spend the marginal hour or have their team spend time with you. And so the, you know, the opportunity is for relationships to matter in the enterprise I think is overlooked. Like, enterprises are not monolithic skyscrapers. They're people inside and their names and individual people have big roles. And you know, as much as it's intimidating for you to look from outside to the skyscrapers, they're like, oh my god, how can I possibly work with them? You know what's happening inside the skyscrapers? They're like, oh my god, how do we make our business work five years from now? And that's an equally tense pressure that they're feeling too. And so they're looking for you on the street down there, right? And so like, you're trying to find each other. And when you do, it's actually pretty great because you can really build deep partnerships. I think the enterprise, if you're a person who likes people and you like technology, this may be counterintuitive. I think the relationships you build when your business by its nature has a relatively low customer count. You know, we're never going to have 40,000 customers. I think, you know, a forward is a lot of different types of product building and kind of feedback collection and it's a really rewarding experience. No, I love it. Well, let's do a quick lightning round to wrap if we can because we're coming up on time here. So, I got a quick one and then Nick, you can round us out if that works. So, highest high, lowest low from the journey. Highest high, you know, maybe counterintuitive really, has absolutely nothing to do with selling the company. It has to do with like, there was a train ride back from Philadelphia. We had built the first product for this company there and we were there with them and it was my CTO, my co-founder and our header product around the Amtrak. We went down, we did a bunch of stuff around the product that we had built and we talked to them for like 30 minutes about the next segment of their claim process. And we built the product on the train ride back. Built the spec, we built how it was going to work, what was it going to do and like we went back the next day, we were like, look, we have the next thing now. And that next product became a massive success for us and it's still in use for them, you know, many versions later. So like, you know, just the being inspired by your customer and then being able to add value to their business and then do that, you know, trusted way that's kind of fast and like, you feel that don't mean there's nothing more fun than that and doing with a team that you just love to work with. So that is incredible. And we've had a lot of moments like that but that was like the first one where I was like, oh man, this actually might work. There's a lot of imposter syndrome in this business and so the first few moments we realized, oh wait, this actually could work is kind of a magical moment. If you don't go into this thinking it almost certainly will not work. I think you're an idiot. And then the lows are the opposite of that where like nothing is working and there's plenty of that too. And you're like, like, do we have deep holes in our architecture? Like will this product actually work? Was it a fluke? Like, you know, anything that resembles that is a very scary moment and there's plenty of it. And then we have a Fortune 500 insurance CEO calling you and saying what the F is going on. Gangly, we want to lock the customer so we haven't, we've entertained it to the edge but I mean, it's yeah, there are highs and lows, 100%. All right. Well, last one here, this one's a bit of a agree, disagree, and why. And you know, I think you previewed us on this one already. But if you want to build a big valuable and importantly durable technology business, first year revenue is a horrible barometer. Yeah, totally agree with that. Has almost nothing to do with it. I think I first read it was $30,000. I mean, I just, I want to re-highlight that one because I feel like now, I mean, look, this is somewhat VC land, how it goes, like let's simplify it into one benchmark. Everyone has to hit, which is ridiculous. But you know, I think you're hearing a lot more than we have over the past couple years of like, yeah. And I think there's a lot of people dismissing companies that, you know, are building true long-term value for the wrong reasons. Not to say, hey, revenue growth is always great, but- Just to be clear, if you can go from zero standing start to 10 million revenue in a year, that's obviously great. Yeah, you've done something right. There's nothing bad about that. I would just say there's the absence of that is not evidence of much. And you know, I think VCs are in a funnel business. I have respect for the profession. I have a lot of respect for individuals in the profession. They're looking for signal and noise. And so they rely on heuristics. And so that's a fair one. I would just say, you know, the precision of it is very low. I mean, that's okay because they just need something to win other things down. If you had the time to wonder, why does this company have 30,000 revenue? Us, great customer relationship. That one customer is like, if this works, it changes the business. And another customer is 30,000 revenue because everyone said no and the project doesn't work and the last customer is about to turn pretty different. But you've got to spend time to figure out which one it is and that's the problem. And if you had made decisions on like a monthly basis of, hey, if we don't have hyper revenue growth, we're not going to do this. You might not have built the product to the- I think I would sell founders is like, when you get to the growth stage, there's plenty of revenue pressure. You'll get- you don't need to accelerate it, right? Like the benefit of the early stage is no one gives a shit how you got to series A, right? Once you're there and the company is doing something cool and it's maybe sustainable and maybe it's scalable and like all that stuff is kind of starting to come into view, then we got to go. But how long it takes you to get there? That just takes years off of your life, not the VC's life. So like, you know, when you're ready, you go. When you're not ready, you build this and like over thinking that I think is a mistake. Awesome. 100%. Well, so appreciate the time. Thanks for sharing your wisdom with us and the audience and man, I learned a lot. So it means it does- This is one of the best, Mike. So really, really appreciate the time. Founders will love it. So, glad to be here. See you guys soon. Alright, brother. Appreciate it. Thanks. Alright, that's it for today's episode of Verticals. Man, evolution IQ, Mike Salzman, the team over there just absolutely incredible has made me like really rethink just the vertical software playbook in general and you know, think about serving enterprises and just doing it in a different way and proofs in the pudding. So fantastic conversation with him, learn to ton, hope you all did and we'll see you next week.