"A lot of the early-stage company ideas I see in the application layer these days sound like they'll be an awesome product for the user — and a terrible business for the founder. It’s going to cost $1 a month because you’ll have 100 people who take a weekend to build a very similar thing."
This warning from EvolutionIQ co-founder Mike Saltzman captures a central paradox facing founders in 2025: AI has made it easier than ever to build powerful products, but harder than ever to build lasting businesses. When new entrants can emerge seemingly overnight, competitive advantage teeters on a knife's edge. (As the recent DeepSeek news showed, even the biggest players in AI might not be immune from this challenge.)
For EvolutionIQ the answer lay in an unlikely place: the staid world of bodily injury claims. And the team certainly built something durable. Last December, EvolutionIQ's $730M acquisition by CCC Intelligent Solutions marked one of the first major vertical AI exits.
EvolutionIQ started with an obscure but interesting idea to apply AI-native software to create the basic framework of a claim, filling in the gaps by turning messy unstructured data (including documents like detailed medical records) into clear next steps so frontline claims experts could see the full picture and focus on helping more people get back on their feet.
But the journey from idea to big-ticket acquisition was far from smooth sailing. What the headlines don't capture are the real stories behind these outcomes — stories that often go untold until years after success has smoothed over the rough edges.
That’s why we set out to document EvolutionIQ’s unique path to product-market fit in full, so other founders could grab takeaways from that deeper (and admittedly much longer) story right away. Luckily, we’ve seen this team’s journey up close over the years, ever since First Round partner Bill Trenchard led EvolutionIQ’s seed round back in 2019.
The founding team — Google ML engineer Tomas Vykruta, Stanford MBA Mike Saltzman, and enterprise veteran Jonathan Lewin — had chosen to tackle an unsexy, relationship-driven industry where only 150 people in the entire country could greenlight a purchase. "The pushback from investors was brutal," says Saltzman. "We heard a lot of, ‘There are only 20 or 30 carriers in U.S. disability insurance, each of them are going to pay you X, so the whole TAM is only 20-30X.’ A lot of VCs couldn't get over that hump — but Bill could."
Given these dynamics, EvolutionIQ's journey was a tightwire act where every element had to be precisely calibrated — from a complex technical product that could deliver immediate value, to a go-to-market motion built on long sales cycles and white-glove service. There was no place for growth hacks or standard SaaS playbooks — and no room for error.
But these constraints became their advantage. By focusing intensely on this narrow market, they dug deep moats through proprietary data, nuanced workflows, and tight customer relationships that would be nearly impossible to replicate.
After six years, with only $63M raised and a team of 200, EvolutionIQ had created — and won — the category of "claims guidance," with nine products, international expansion, and most importantly, a sustainable model that didn't rely on technical novelty alone.
Their story offers more than just impressive metrics — it reveals what it truly takes to get a company to work in a tricky space. We sat down with the founding team to understand how they threaded this needle, documenting their unconventional choices and hard-won lessons for founders building in the AI era.
EXPLORING IDEAS: EVOLUTIONIQ’S ORIGIN STORY
For co-founder Tom Vykruta, the path to founding EvolutionIQ started at Google, all the way back in 2010. Selected by Jeff Dean to join an elite group of "Machine Learning Ninjas,” he landed on a team of 15 engineers who went through intensive AI training before fanning out across the tech giant’s org chart.
“I was hopping around different parts of Google, bringing machine learning to groups that were using older heuristic methods," says Vykruta. “I felt like I was 12 years old again, learning how to program for the first time. I worked on everything from self-driving cars to improving Maps’ business recommendations to search optimizations. At some point, I started to realize that this technology had insanely broad applications, and it was going to just take over the world."
With that conviction, Vykruta spent a couple of years exploring opportunities for a company of his own. “I went through probably dozens of ideas at different levels,” he says. To narrow in from a sea of possibilities, he established a clear criteria as he tried to find a startup idea:
- Repeatable problem
- Massive dataset
- Using primarily manual or rules-based solutions
- Not much competition yet
- Interesting mission
“I eventually landed on this idea of bringing AI to bodily injury insurance — to essentially build an AI that can understand the human body, learn about injuries and recoveries, and then use that knowledge to drive better outcomes by getting people to recover faster and shorten their disability time,” he says.
This space checked all his boxes: predictable patterns with finite outcomes, digitized but underutilized data, and an industry that was just starting to dip its toes into the waters of machine learning. “Some carriers were running experiments, but the claims were all mostly handled by non-medical experts: insurance adjusters. Their job is to review the file, make sure that it meets the regular requirements and move it forward,” he says.
"I also had a good feeling about the mission," says Vykruta. "Every year millions of people get injured and become disabled in work incidents. This is a huge multi-billion dollar problem for the economy — not to mention the tangible impact that getting healthy and back to work can have on people’s quality of life.”
For co-founder Jonathan Lewin, what stood out most was the underserved market. "Insurance is the second or third largest industry, if you count government. It's huge. But there are lots of underserved areas. People gravitate towards health insurance because it's something that we all interact with, or auto insurance because most of us drive,” he says. “Disability insurance is fairly esoteric — most people probably don't know exactly what it means. But it's meaningful — the U.S. disability industry alone has tens of billions of dollars of revenue, they're far from small entities.”
A massive part of what made EvolutionIQ work is that we jumped on something narrow to start with and understood it really well. The economy is so complex, and every little facet of it is going to change with AI — there's tons of opportunity. But founders tend to gravitate towards the same set of things.
– Jonathan Lewin, co-founder of EvolutionIQ
ASSEMBLING THE FOUNDING TEAM: A DELIBERATE BUILD
Finding co-founders can often be a rushed affair, with founding teams coming together through serendipity or mere proximity. For EvolutionIQ, while the founders already knew each other through their shared passion for aviation (all three are pilots), the process of formally joining forces was more intentional, with each addition carefully chosen to complement existing strengths and fill crucial gaps.
This more methodical approach was driven by the unique complexity of the business. EvolutionIQ was tackling multiple challenges that increased the level of difficulty: working in a heavily regulated environment with sensitive health and financial data, executing a relationship-based B2B sales motion, and taking on the high-risk technical problem of building claims guidance from unstructured medical data. This stands in contrast to the more typical early-stage startup profile of low-regulation spaces with lead-gen driven, high-volume sales and lower technical risk.
"Mike joined in early 2019, and Jonathan came on full-time in 2020. I don't know if I could have handled the pressure and workload without them as co-founders," Vykruta says. “While I had the technical chops, I hadn’t started a company before and I had no experience working in the regulated insurance space or with B2B relationship sales.”
Here’s how Saltzman and Lewin filled those gaps: “At Bridgewater, Mike worked with insurance companies, so he had some sense of how they worked and built up a network in that world," says Vykruta.
"Jonathan was a good friend, and as a repeat entrepreneur, he'd been an advisor for me for many years. In particular, he was helping us think through how to structure B2B enterprise sales because it was a very special type of sales motion — email templates wouldn’t work. He also has a lot of legal expertise, which was crucial as we were dealing with a very regulated space,” says Vykruta. “I just kept asking him, can you just come join full-time?”
“I told him, ‘No, I've done startups before, they're totally exhausting and I have a two-year-old right now, but I’ll help you,” says Lewin. But what started as advising quickly turned into something more. "Tom and Mike were calling me 10 times a day," he laughs. “I just got sucked into it, and I felt like I could add more value by coming on full-time.”
For the first several years, Vykruta was the technical founder in the CEO seat. “Tom gravitated towards tech, so he was also the CTO, overseeing the engineering, data science and machine learning parts of the company,” says Lewin. “Mike gravitated towards go-to-market, so he was out looking for buyers, and figuring out how to reach real product-market fit with major customers. I gravitated more towards the back-end of the business, stitching everything together and figuring out the strategy — what we went after and how hard we went at it,” he says. “I also brought an understanding of how to grow teams, how to incentivize and motivate, how to interview people, and how to manage customers.”
As their first investor, First Round partner Bill Trenchard had unique insight into where each member of this trio spiked: "Tom was clearly brilliant, and his background in ML at Google was impressive. Mike’s experience at Bridgewater was very relevant, and he’s proven to be aggressive and very strong commercially as an executive. Jonathan was amazing on the business side and very strategically strong."
Unconventional advice: Split the CEO seat — and double your bandwidth
As EvolutionIQ evolved (the only time we’ll use that formulation, promise), the co-founders' roles naturally started to specialize further. “Over time, Jonathan's role morphed into wider responsibilities, and I started doing less of the managing of our people ops," says Vykruta. “And while in the early days, Mike and I were in all the sales calls figuring out founder-led sales together, around 2022 I started focusing more on just managing engineering and technology.”
This led to one of EvolutionIQ’s bigger break-the-startup-rulebook moments: opting to move to a co-CEO model, with Lewin and Saltzman teaming up to share the top seat.
“Everyone told us that it was crazy and made no sense — but it ended up being a great decision,” says Lewin. “The startup CEO role is almost an impossible job because you never have enough hours in the day. Having two people share that role gave us incredible leverage. I didn't have to do everything, and Mike didn't have to do everything.”
He points to their different backgrounds as key to making it work — and why YMMV. “Mike and I basically share the same brain and talk 10-20 times a day,” says Lewin. “But we're pretty different — we each have our own focus areas, and there's an age difference, too, which perhaps creates less of a competitive dynamic. He was a first-time founder who saw things in a fresh way, while I was able to bring a lot of pattern recognition and put frameworks around our challenges from my previous companies.”
Most startup advice says that co-CEOs are a terrible idea. For us, it was a force multiplier. We could get so much more done.
– Jonathan Lewin, co-founder of EvolutionIQ
VALIDATING THE IDEA: THE CUSTOMER DISCOVERY THAT UNLOCKED THE FIRST DESIGN PARTNER (AND DATA SET)
“Tom had been having conversations before I was involved, but we started talking to insurance companies together around January of 2019,” says Saltzman. “When Tom first called me about this insurance idea he had been working on, my initial take was ‘I don't really know if it's a good idea or not, but let's just talk to insurance people and get their thoughts,’” he says.
“I thought it had all the hallmarks of a good idea,” Lewin adds. “And Tom is at his happiest knocking down doors, he runs straight through them like they don’t exist. So he started calling massive Fortune 500 companies, getting Prudential and those types of big entities on the phone and pitching them the idea. It wasn’t even close to being baked yet, but it sounded like there was a real need. For me, that validated that there was something to it.”
Despite early glimmers of interest, the risk-averse insurance world was shaping up to be more of a founder-led sales slog. "A lot of companies said, ‘Look, this is interesting, but we're not going to be someone's first customer,’” Saltzman says. “There were a lot of no's. We probably heard ‘No thank you,’ or ‘Come back to us when you have something more mature,’ 100 times,” Vykruta says.
There was this inherent chicken and egg problem: You needed the data to prove the value but you couldn’t prove the value without the data.
– Tom Vykruta, co-founder of EvolutionIQ
“The insurance industry is very opaque — they're not used to sharing data with vendors. A lot of them have never done it before,” he says. “We interviewed a lot of the insurance executives and operators, and built a compelling fully-clickable demo without a real data back-end to sell the vision. But without that data it was really hard.”
The EvolutionIQ team's patient approach to customer discovery paid off when they found their first design partner, Reliance Matrix — which remains one of their largest customers to this day. “They were interested because although what we had pitched them wasn't what they wanted, they saw similarities," says Vykruta. "They said if you guys can change this a bit, it looks like the perfect solution for us."
The team later learned that Reliance had previously worked with another vendor on a similar problem without success. "They didn't want to give up, because they thought this was a big opportunity. They just needed the right technology partner,” Vykruta says.
The demo alone was a low hit-rate. It took a year to get the first customer who was willing to share their data set. Once we got it, we used that to launch a product, build a case study, get a reference on the customer. That’s the recipe for getting more data sets — you’ve just got to hang on and keep going.
– Tom Vykruta, co-founder of EvolutionIQ
“To their huge credit, they were very excited about the promise of the tech and wanted to go on a journey with us,” Saltzman adds. “This was in the spring of 2019, right before we raised from First Round in the early summer. This first pilot was a small contract — I think they paid us only $30K — but critically, they eventually gave us access to their claims data. And we agreed to work together for several months to see what kind of prototype we could build off of that would improve their business."
Saltzman and Vykruta walk us through how the design partnership unfolded, sharing advice for other founders bogged down at a similar stage:
Show up (literally)
What followed was an intensive period of collaboration and learning. To build up expertise as outsiders, the team went what we here at First Round like to call “unreasonably deep.”
“We spent a ton of time just reading, reading, reading claims and medical data so we could intuitively understand what was encoded," says Vykruta. “An important part of being a data scientist is becoming an expert in that real world domain so you can double the machine and teach it how to think and predict more effectively."
What made their industry uniquely suited for this approach was the rich documentation already built into the claims process. "During various stages, the insurance industry requires examiners to add notes to say, ‘On this date I made this phone call, I learned this information, I made this decision,’" says Vykruta. "We can learn a ton by reading through it and training our models on this data. In other industries, where they have purely quantitative data sets, they have to go and ask the customer 'Why did you do this?' But we had that documentation layer.”
Even so, the team didn’t skip over the heavy lifting of customer discovery. They in fact went out of their way to embed themselves with their design partner, making regular trips from New York to Philadelphia to work directly with the frontline staff. “This was pre-COVID, so we were there all the time, sitting with their examiners and iterating quickly,” says Saltzman.
This meant hundreds of hours of IRL customer interviews. "Going on-site adds such an important dimension. We still go on-site even today to many customer calls, whether it’s sales calls or post-implementations. The personal relationship piece of it is enormous, you can’t undervalue it,” says Saltzman. “I'm a huge believer that people don’t just buy software, they want to work with people they like. Regardless of what the technology might do, if a group of people like the partners they're working with, they will get more out of it.”
This wasn’t just founder-led sales work from Saltzman and Vykruta either. “We had hired a few engineers — who now lead our engineering team actually — but it was a small team working closely with the end-users to build the prototype product," says Vykruta. “We'd have several one-hour meetings per week for months to tackle the first problem alone and those meetings would include several of our senior data scientists.”
If an AI team is living in a silo just looking at data on their own, they're going to have a tough time cracking these older established industries. Everyone from the founder to the newest engineer has to be deep in there with the customer — otherwise you’re not going to make it.
– Tom Vykruta, co-founder of EvolutionIQ
Win hearts & minds at every altitude of the org chart
“One of the things that I think our business has done really well — and has been important to do well — is we have been good at building and maintaining relationships across the verticality of an organization," says Saltzman.
“From the frontline desk person who could be a recent or junior employee, to the manager, to the director level, to the VP, to the Chief Claims Officer and the President of the business — we built relationships with everyone in the reporting line, and we adjusted our approach depending on what they needed," he says.
This meant paying as much attention to frontline workers as to executives. "I still remember our partners six years later — Gia and T.J. were the frontline examiners we were working with, and they were our confidants. They were into it, we were into it, there was just good energy," says Saltzman. "And importantly, we made it clear that whatever we were building was in no way a threat to their job. It was meant to enhance what they were doing, to support the hardest parts of it.”
The team created feedback loops that spanned the organization. "With the frontline examiners, we were white boarding out what we could do at the desk level,” says Saltzman. “And then once a week, we were meeting with their manager, and once a month meeting with the head of claims and saying, 'Look, this is what we're hearing, this what we think we're building, does this problem work for you as well? Do you want a solution here?' We got all the stakeholders to dive into the process.”
This multi-level engagement approach would become a cornerstone of EvolutionIQ's success as they expanded to new customers. From Trenchard’s perspective as an investor, what stood out was the founders' communication chops here. “They had this ability to really explain at a deeper level what was going on inside of disability insurance, and why AI was going to be a big difference maker for them,” he says. “You only develop that skill from immersing yourself in the problem and going ‘unreasonably deep’ to understand it from every angle.”
Find your feedback rhythm — and the single metric you want to move the needle on:
The EvolutionIQ team eventually established a rigorous weekly rhythm of iteration and feedback. "On Wednesday, we’d receive a new data dump. We would evaluate it, retrain the models, re-score it all, make it available to the frontline examiners by Friday. Then by next Tuesday, they would have viewed our recommendations and given us feedback," says Vykruta.
“Essentially they’d review different medical diagnoses, and would say, ‘Yes, this one is correct, this person could go back to work,’ or ‘No, this one cannot.’ Every single week, the model would get better,” he says. “Say they accepted 40% and rejected 60%. Our ML team would review all the rejections and then look for patterns to improve the model. Our goal was to improve the acceptance rate by 10% every single week. So after five weeks, we would go from 40% to 90% and the model would be at a high enough acceptance rate that the examiners could actually start to use it for day-to-day work.”
Go out of your way to explain the why
There were, of course, plenty of nuances to navigate. “Sometimes we were giving the wrong recommendations. Sometimes we knew we were right and the frontline examiner was wrong — but we had to figure out how to explain the recommendation in a way that would get them to agree,” says Vykruta.
The early EvolutionIQ team focused on making their AI system's recommendations interpretable and actionable. "Rather than just serving up raw scores or simple severity ratings — which was very popular at the time — we would give them an English sentence that would explain why we made the recommendation," says Vykruta.
“It might say something like, ‘For people of a similar age and gender and medical background, it typically takes six weeks to recover from this type of injury. And this particular claim has been on leave for 18 months. Therefore it makes sense to reach out and see if they have made progress.’ Then the frontline examiner would then realize that there was an additional complication they’d missed, like a chronic disease that complicated the medical picture. So they went out and got more information, updated the data and were able to manage that claim better — and then our predictions about outcomes got better too,” he says.
BUILDING THE PRODUCT: AI THAT HUMANS ACTUALLY USE
Most startups face intense pressure to expand their product footprint quickly, especially when tackling a more constrained market. But the EvolutionIQ team tried to not let that sap their focus.
“It's tempting to do many things because there are big opportunities, and you want to open up new markets. But we've remained very diligent,” says Vykruta.
“There were a lot of pressures, internal and external, to try to make the TAM as big as possible and go after everything,” Lewin adds. “The mantra I always share with other founders is to crawl, walk, run. Completely nail one thing perfectly before you expand the aperture. Because stuff breaks in unexpected ways. If we were trying to take more on, we wouldn't have had that core team focus on what turned out to be existential level stuff.”
Saltzman agrees. “Our approach was to build incredible products to start, to achieve great success in a couple of years because our first customers are really happy and the first things we build for them are working great. Then we'd go solve the next problem the next day,” he says. “And that mentality has served us quite well — to date, that first product has never churned a customer, even six years later."
We had a rule to try not to worry about the business issues we didn’t have yet. So while we felt pressure to go multi-product and get the land-and-expand motion up and running, we were laser focused on making our first product an incredible one.
– Mike Saltzman, co-founder of EvolutionIQ
The technical complexity that it took to craft that initial product shouldn't go unnoticed. “The biggest challenge we faced — which we've now solved — is that every carrier has a different way to represent the data,” Vykruta says. “They have different schemas, different versions of databases, so we had to work with many custom data sets before we could finally create a unified schema. Then we were able to transform and normalize every carrier's data to look the same. That took us a couple of years.”
Below, Vykruta shares more about the product choices and strategic focus that enabled EvolutionIQ to nail the most important dimension of nascent product-market fit: satisfaction.
Loop in the humans
"The choice between automation and guidance is key for any AI product. Our decision early on was that we don't want to get into automation. It’s the low-hanging fruit. It's easier, it's higher frequency — but it's more competitive," says Vykruta. "The internal teams might not be able to do it as well as an AI company, but they can do it reasonably well. The much harder piece is the tasks that cannot be automated entirely."
Instead, the early EvolutionIQ product focused on helping claims adjusters make better decisions about where to spend their limited time. "A human expert might be dealing with managing 150 claims, or even thousands of claims at any given time. They need to decide where to spend their time. That's hard for humans to do well," says Vykruta. “Once they're in the claim, that's when you want the human expert to make the decisions — to pick up the phone, get more information from the doctor, call the employer, and so on.”
Design for adoption, not just accuracy
A key insight came from observing how other attempts at AI adoption in insurance had failed. "A lot of the clients we spoke with back then had an internal team building a model, or they were working with a ML vendor. And maybe these models worked, but then they gave the frontline examiner a spreadsheet with 20,000 rows that had some score in it," says Vykruta. "A team would try to use it, but it didn't really fit into their workday. They have to take an hour to go into the spreadsheet, look at stuff, and go back to their other UX and try to make sense of it."
The EvolutionIQ team took a different approach. "To get adoption of complex AI, you need to have a beautiful interactive visual tool that people can work within. You can't give them just a spreadsheet of scores and say, 'Go run with this,'" says Vykruta. “We made a big investment from day one in not only developing the models, but also developing that UX they could log into and interact with."
Adoption of AI is actually much harder for technical teams to solve than building the technology itself. You should be spending something like 70% of your time figuring out how to get people to use it. We're all great engineers, we know how to build technology. But can you get somebody to use it? That part’s not as easy.
– Tom Vykruta, co-founder of EvolutionIQ
"Our initial team was purely data scientists. I was looking for AI engineers who had specialty in either unstructured data NLP or tabular data, and specifically sequential data," says Vykruta. "But I also hired React engineers to help build the front-end very early on." The early EvolutionIQ team even went so far as to train special models just for the purpose of explaining what the main model was doing, to get users excited about using it.
The result was a system that felt natural to users' existing workflows. "We wanted it to be where they start their day — they log into our system, and it leads them down the right path to which claimant they should be speaking with. Then they’d jump back into their system to add notes about what they had been doing," says Vykruta. "They actually helped us build it, telling us exactly what elements they needed to see. We were able to win over the adjusters because they loved seeing our UX laid out how they’d envisioned.”
TURNING THE FLYWHEEL: BUILDING A BESPOKE GTM MOTION
After finding the inroad into the industry with their first design partner and building out the initial product, the EvolutionIQ team had to figure out how to scale without compromising the high-touch, relationship-driven approach that made their first deployment successful.
For context, the majority of EvolutionIQ’s contracts are multi-year, six- to seven-figure enterprise sales deals that require nurturing with internal champions in long sales cycles (12+ months).
First Round partner Bill Trenchard recalls being struck early on by the EvolutionIQ team's thoughtful approach here. "For the first two years, it was a very long product build, and the game was making sure it worked and demonstrating efficacy to get credibility in the market," he says. "They truly nailed every piece of their GTM, from navigating the long sales cycles and establishing credibility that led to strong references, to the pricing and positioning with this premium, Rolls Royce-like product. And that matters tremendously in the insurance world."
Ultimately, the product has to work. And if the product really works, you have to make sure that you are telling the story of how it’s working in a way that’s compelling for your specific market. For our market, that was getting people to talk to their peers, building a machine to facilitate that, being taken seriously ourselves, and creating an aura that we were a company that would make you look good if you took a chance on working with us.
– Mike Saltzman, co-founder of EvolutionIQ
Here, Saltzman and Vykruta walk us through the lessons they picked up along the way, specifically breaking down:
- How they approached the market and positioned themselves
- How they got in the door
- How they moved deals forward
- How they delivered and expanded
How they approached the market and positioned themselves:
Aim carefully when elephant hunting
This founding team knew that traditional enterprise SaaS playbooks — armies of BDRs, mass email outreach campaigns, growth hacks — wouldn't work in their smaller pond. To mix our metaphors, they went elephant hunting instead, focusing intensely on the more methodical approach of building deep relationships and delivering measurable value to nab the important players.
The team was strategic when it came to selecting their targets at the watering hole. "It's a careful balance between getting more at-bats, but also not completely whiffing and souring an organization on you forever," says Saltzman. "That’s why we didn't want to try to sell to the biggest companies in the early stages," says Vykruta. "We actually tried our hand at it once, and it was hard. There was too much bureaucracy, and we were a smaller company back then, so they were more risk averse."
Building our account target list was a deterministic effort. Networking to the right contacts, starting the first conversation, navigating the long sales cycle — it all requires a huge amount of time for a 10-person company. So we’d try to find Goldilocks-sized companies until we felt ready for the bigger players.
– Tom Vykruta, co-founder of EvolutionIQ
Price it high — if you can make the case
Another key component was how EvolutionIQ was able to command a higher ACV than most. “One of the things I think we got right is our pricing. We would have to work just as hard and our sales cycles would be just as long, even if our prices were 10% of what they are," says Saltzman. "A lot of founders get this wrong. They think 'If I raise my prices, it's elastic,' and that's not true — with the big caveat that you have to do a good job at articulating the case for it, of course.”
The team moved aggressively here. “We spent a lot of time making ROI models, helping our customers understand what they were going to get, how they were going to get it, and how we would measure it,” he says. “And then we’d check back at three, six, nine, and 12 months into the implementation to say, ‘Okay, we said we'd be halfway there by now. We're ahead of schedule,’ or ‘We're behind but here’s the plan.’”
Here’s how they got customers comfortable with the premium pricing models upfront:
- Making it measurable: "On the continuum of venture-backable businesses, we were lucky to be in the insurance business, working on things that are more or less measurable,” says Saltzman. “There are nuances, but broadly speaking, you can measure the impact of what we do. But if you're building a next generation Salesforce killer, it's harder to measure the value of a CRM and figure out how to price that,” he concedes.
- Picking a space where you can move the needle: “Insurance is a big business, so if you have an impact, you have a big impact,” he says. “If you're building a system that makes something that was going to cost $100,000 cost $90,000 instead, and there's 100 of those things per year, you've created $1M of value.”
- No long-contract lock-ins: “We didn't make prospects sign five-year contracts. If it didn't work, they could fire us after the first year — that never happened, but I think that approach was hugely helpful,” says Saltzman.
Break the rules when your instincts tell you to
There was another unconventional pricing move, too. “When we were negotiating with our first customer, we had a choice between a fixed fee, which would probably have been low- to mid-hundreds of thousands, or to have less money fixed but success fees on top of it," says Lewin.
“We were told ARR is very important, and the push across the board was, 'Don't do success fees.' But we realized that the only way to get this deal to be really significant was to have success fees and share the risk with our customer. We decided to go for it, and it was very much the right decision because the software really worked. We took in millions of dollars from that customer, and then it eventually became a fixed fee of millions of dollars.”
Pattern recognition is helpful, but sometimes as a founder you just have a very strong sense of something. The conventional wisdom is there to help you — not to handcuff you.
– Jonathan Lewin, co-founder of EvolutionIQ
How they got in the door:
Score an invite to the party — and don’t show up empty handed
The EvolutionIQ sales motion was high-touch from the very start. "We try to get a warm introduction as much as we can via an advisor or someone we know," says Saltzman. The team looked for inroads, searching LinkedIn for retired insurance executives and sales leaders, but mostly relying on word of mouth. "Once people know what you're doing, you can ask them, 'Do you know somebody who would be interested in talking to us?'"
These connections proved invaluable because they truly understood the space. "They get the pain point because they've been in this industry their whole life," says Vykruta.
The team's approach once they were in the door wasn't drawn from any playbook. "It wasn't like there was a particular sales book that blew our minds. It was a lot of common sense — treat people with enormous respect, expect them to take you seriously and put in the work to earn it and show up with analysis," says Saltzman.
That analysis piece was crucial. “If they're a public company, dive deep into their financials. Figure out where they're doing well and where they need help. Give them a thesis — not quite in a ‘Challenger Sale’ adversarial way, but to really get inside their business to every degree that you can and then layer on top of that deep personal relationships.”
Find the right champion
These efforts were, of course, hypertargeted to get to the right person. But finding the right champion inside these large insurance carriers wasn't just about a specific title — it was about identifying a particular mindset.
"We had to find the person who had both the organizational clout to get something a little risky done, and also was personally open to shouldering that risk," says Saltzman. "They needed to be all in on the bet that they’d look really prescient when this thing started to work."
There were some other non-negotiables in their criteria. “They needed to hold responsibility for the overall profitability of the business line," Saltzman says. “If it was someone who was only focused on one particular aspect of claims operations, like efficiency, they probably weren't the right sponsor for us,” he says. “We needed someone who understood, 'If I save $10 on claim outcomes because I'm getting people healthy faster, but I have to increase my spend on manpower or my adjusters by $1, I'm still saving $9.' But some people would get overly fixated on, 'Oh, this cost went up by $1' — that type of person couldn't be our sponsor."
This often meant swimming against the typical sales process current, for two reasons:
- Most insurtech products are bought by CIOs, but EvolutionIQ never worked with a CIO as their core sponsor. "They're great partners, and they're involved of course, but they’re rarely our core champion," says Saltzman.
- Most software’s business case is built around an efficiency argument, but EvolutionIQ led with the product’s health outcomes. “We were focused on claim outcomes and returning people to health faster — so that meant we were selling software to a buyer who typically didn't buy software,” says Saltzman. “That cuts both ways. It was good because it allowed us to educate each other and build the rule book from scratch. But it was also hard because we had to convince them, ‘No, you are the person we have to work with. There's no one else here who we can do this with.’”
Hire sellers who sound like natives, not tourists
Building these relationships in this specialized world requires a different breed of salesperson. "Hiring for our sales staff has been hard. Finding them has been hard, making sure that they're great has been hard — you just don't know until six or nine months later," says Saltzman.
The team quickly learned they had to show up to their clients with a level of sophistication that matched the industry. “That meant no SDRs, no BDRs, no folks straight out of college mashing the phones,” he says.
Even though we were an early-stage startup with six people, we had to act and be perceived as though we were an established company — not through obfuscation, but by professionalism, clarity, level of service, commitment to customers, and deeply understanding not just the business problems, but how their organizations worked.
– Mike Saltzman, co-founder of EvolutionIQ
Another critical ingredient was finding sellers with deep insurance experience — and the team learned this lesson the hard way with their first sales hire. "He was from the fintech world, so he had sold to banks and financial institutions. We were thinking, 'Great, you're sophisticated, you know banking tech.' But he bounced off this industry like a steel ball bearing. It was almost funny — both of us were like, 'Nope, this is not working.' It was just so clear," Saltzman laughs.
“What we're trying to do is talk people's language and show them that, yes, there's this technology that’s crazy and modern, but we're just solving human problems,” he says.
Bringing a sales mentality too, of course, ended up being an important trait. “You couldn’t just be an insurance operator,” Saltzman says. “Our most successful sellers ended up coming from two specific backgrounds:
- Those who had sold insurance coverage to major enterprises like Boeing and Walmart. “They're sophisticated, they knew how to navigate complex organizations, and they also knew insurance.”
- Those who came from big consulting firms selling into insurance companies. "Someone who can sell a multi-million dollar engagement for Cognizant or Deloitte, get the buy-in and handle the seniority of the buyer — it's actually very similar to what we do."
What we do from a technology and a gross margin standpoint looks like tech, but from the customer standpoint, it looks like change management. The tech is just a tool.
– Mike Saltzman, co-founder of EvolutionIQ
This approach allowed the team to maintain remarkable sales efficiency even as they scaled. "Even coming into 2024, we only had three sales people, plus me. Now we have seven, and we've been able to sell into a majority of the largest disability carriers in the U.S. with just this small team," says Saltzman.
“Our contract values are significant. And so if the product works and people buy it, it doesn't take 100 people to pull it off. Having 100 SDRs annoying our prospective customers is not productive, so we don't do that.”
How they moved deals forward and got unstuck:
Battle the build vs. buy dilemma
Customers torn between building an internal version of the software and buying it was a roadblock that frequently popped up. “Build versus buy is our number one competition, that’s been the case since 2019. Now with advances in AI, the dynamic is similar, but it's like both sides got bigger guns,” says Saltzman.
“In some ways the build versus buy question was counterintuitively harder early on, back when we were 10 people, and their internal eng team was 10 people. The answer there was that we just hired better talent than they could — but you can't say that, so we lost some of those opportunities back then,” he says.
“We still win the majority of the time on that question now. Your internal teams are going to build a prototype version, maybe similar to what we have. They're going to ship it. But then they’re going to go start a different project,” says Saltzman. “We've been on that particular value proposition for the last six years — and we have a team that’s going to be on that value proposition for the next six years. That's all they do. That’s the case we make whenever build versus buy comes up.”
Play the (long sales cycle) game
“There were definitely signals to monitor, even with 12+ month sales cycles — for example, if a company was setting the next meeting 10 weeks out instead of one or two weeks out,” says Vykruta. "Even if things were stalling, we would not drop accounts, because in a small industry, you have only so many shots on goal," he says.
Even if an account didn't convert in the first year, getting to know us was still valuable. We knew we would be twice as large and have twice more case studies the following year when we checked back in.
– Tom Vykruta, co-founder of EvolutionIQ
That’s why the EvolutionIQ founders treated every interaction as an investment. "Our belief, even back then, was that there is no future without claims guidance. We didn’t think half the industry would still be doing it the old way,” says Vykruta. “Our view was that every company will end up using the software, and we had a really high chance of being the category winner, because we created the category, and it's hard to catch up."
How they delivered and expanded:
Ship change, not just software
But while EvolutionIQ’s sales team has always stayed lean, the founders made the unconventional choice to invest early and heavy in service, never charging for it and instead wrapping everything into the product price. This included building a substantial customer success team dedicated to the post-sales change management process that many companies pass off to the customers.
"We build a guidance product, which is to say, our system doesn't actually do anything – it guides frontline examiners to take action," says Saltzman. "What we learned early on is the AI part of it, which is predicting what's possible, could be accurate — and also commercially useless if the people using it didn't take the actions that it would suggest.”
“We knew early on that we had to staff up a whole group here. We told prospects, ‘We're going to take your data, give you the software, give you the machine learning models and all that. But we're also going to become a part of your claims team. We're going to help you deploy this. We're going to coach your people. We're going to help your management team keep them accountable.’"
A lot of SaaS companies deliver the software, and then between software and value is the customer's job. But most software, especially in enterprise, falls down in that gap. We got as close as you possibly could to just delivering value itself — there was very little gap between our deliverable and the thing working.
– Mike Saltzman, co-founder of EvolutionIQ
This approach flew in the face of conventional wisdom about maintaining high margins. "If the gap is narrow, you're more likely to jump over it. If it's wide, your customers will feel like they're taking more of a risk. And so we made it as narrow as we possibly could, and that was expensive in the early days," he says.
The team didn't let potential investor concerns about this being a services-heavy model deter them either. "We didn't let VC questions about if this was more of a services business or consulting model get in the way of building. We just didn't care," says Saltzman. "We were quite convinced that if we lost our customers, that would be worse than overinvesting in customer service. We did the math that if we delivered enough value for our customers, and we could capture a reasonable percentage of that value, we could build a business off of that."
Turn your first fans into your sales force
The real engine of EvolutionIQ’s sales motion was making early customers wildly successful. "Our motion has always really hinged on one thing: demonstrating value and making the customer really happy as soon as we could. And this is still our motion when we enter new markets today,” says Saltzman.
It was a race to get a customer so happy that they’d be a reference customer and we could do a case study about their actual quantitative outcomes. Before we got to that point, we basically didn't even try selling. We had conversations, of course, but we didn't put people in a position where they had to say yes or no when they didn't have enough information yet.
– Mike Saltzman, co-founder of EvolutionIQ
This dedication to developing the customer reference motion paid off. "Our best salesperson was our first customer. He's an incredible guy, he took so many reference calls," says Saltzman. "He knew that he needed us to stay in business, given our system was working for him and he was paying a lot less than the X number of employees' salaries it would take him to build something similar. He was like, 'If you guys do well, I'll do well. I know startups go bankrupt all the time, and I don't want you to go bankrupt, and I also don't want to pay 10X what I'm paying, so let's get you some customers.'"
This reference-based approach might seem counterintuitive in such a small, insular industry. But word of mouth and introductions from other customers became a huge lever. "These companies focus on different market segments. They're not exactly directly competitive in all places," says Saltzman. "We were lucky to have champions who could take the long-term perspective of, 'Do you want to support the startup delivering you a 30x ROI, or do you want to have them go bankrupt?'"
Make sure your first $1 funds your next $10
The EvolutionIQ team’s focus on proving value quickly was the foundation of their expansion strategy, too. "Our commercial model centers on land and expand," says Saltzman. "We have different modules of our platform, and most customers start with one or two, and then when it works, the proceeds from using the first or second to fund the next several. It's a compounding partnership process."
He points to a self-funding venture capital portfolio as a parallel. "We wanted to assemble something where you can get the proceeds from earlier funds reinvested to the next fund, it's all snowballing, and our customers — or limited partners in the metaphor — are putting money in their pocket at the same time," says Saltzman. "We did the math for them, very clearly showing that if you spend $1 on one module, it'll make you $10. If you pocket $8 and spend the next $1 on the next module, this time next year you'll have $16."
PQ: Our land-and-expand model started very organically, but we honed it to make it really commercially successful — and that meant making sure that people got to see the value from their first product modules as soon as possible, so that value could be recycled and fund the next piece of value we could deliver. - Mike
"Interestingly, we don't have a real beachhead product, which is different from a lot of platform companies," says Saltzman. "Our products are quite diversified by first purchase, and when they are at scale and working, they have relatively commensurate value impact and pricing.”
This meant their expansion conversations were deeply strategic. "It’is more about what would make most sense for that particular carrier right now, based on where they are and their challenges. It's not like 'We got you to do this, now we want you to do that.' It's more of a, 'Okay, tell us about your five-year strategic plan. Tell us what you're telling your board. What are you working on?'" says Saltzman. "And then we expand to the product that's most aligned with that."
FOCUSED EXECUTION: MAINTAINING DISCIPLINE ACROSS PLANNING, FUNDRAISING AND RECRUITING
"The company consistently executed and beat forecasts, even in challenging markets, from COVID to the 2022 reset," says Trenchard, EvolutionIQ’s seed-stage investor. "They raised little external capital and were just generally very disciplined.”
That rigor was also evident in how they approached building their team from the earliest days. “Even at the seed round, I was struck by how they already had a great AI product team, largely sourced from Google,” says Trenchard. “I wrote down in my notes at the time, 'I think they will be able to stay ahead of any competition on features / ROI,' and they’ve continued to do that in an impressive way."
The EvolutionIQ team's execution playbook centered around a few key pillars: an obsessive early focus on metrics and planning (even when plans changed), a disciplined approach to fundraising that embraced constraints (despite rather frothy environments), and an unrelenting drive (paired with unusual patience) when it came to recruiting the team that could pull it all off.
The co-founding trio helps us dive deeper into each one below.
Embrace early planning, even when plans change
"Our North Star success metric is how many additional people we are able to return back to work, and how much can we reduce the total number of absence days," says Vykruta. "We're very metrics-driven — we believe if it can't be measured it didn't happen, which is definitely something I picked up from my Google days.”
The team adopted Google's OKR system from the very beginning, too. "We started that very early — when we had three employees, knowing that we didn't really need it at that stage, but that it would also be much easier to implement as we scale up to have that muscle early on," says Vykruta.
While the actual OKRs often shifted in their early days, the process proved invaluable. “OKRs helped to motivate people and set a really high bar for delivering exceptional results to our clients. If the client didn't see results, it didn't matter,” he says. “But as far as the actual tactical parts of the OKRs themselves, we were so early and changing our path so quickly that whatever we decided for the quarter would be different even just a week later," he says.
"But the process still was very useful, and almost existential in a way. I don't think that we could have made nearly as much progress without it. We needed that kind of vision reset every quarter to keep pushing ourselves to set really high bars."
In the very early stages when things are chaotic, it's very important to have a plan. It’s okay to not follow it, but it's not okay to not have it. The plan is extremely valuable, even if it's only used as a starting point that you diverge from.
– Tom Vykruta, co-founder of EvolutionIQ
This became crystal clear when they briefly abandoned OKRs as an experiment. "We did try one or two quarters without OKRs, and it was very unpopular. We got a lot of complaints — as much as people didn't love the OKRs and complained that it was getting in their way, when we took it away, people were saying, 'We don't understand what the company is doing. We don't know what other teams are doing.' So we brought it back quickly."
Let constraints be your guide
While many AI startups chase the biggest rounds possible, EvolutionIQ took a more measured approach to fundraising. "We raised in 2019, before the big inflation in valuations. And I think it actually helped us, because we didn't have an infinite budget. We were very tight with hires and very careful with everything. We didn't have a choice," says Vykruta. "We were very swift with decisions like letting people go. We set hard limits on the time and resources dedicated to new bets, time-boxing initiatives and tying ourselves to agreed-upon metrics in order to move forward,” he says.
"You can end up getting addicted to fundraising. Things get bigger than you can control, and you end up not being master of your destiny," says Lewin. "So we were always running the business in a way where we could be profitable within 18 months. Our approach was always very managed, which meant the metrics on the business stayed healthy. It ended up making it easier to raise money because our metrics were good and our burn wasn't crazy."
This disciplined approach extended to how they sized their rounds. “In each round, we had a strong sense of what was sensible and limited the amount we took on,” says Lewin. "Our Series B was only $7M, which is small, but it was all we needed at the time.”
As a result, the team also implemented strict guardrails around new initiatives. “Once things start to get crazy, when you're hiring more people who haven't gelled, and you're going after too many new ideas that may not work out, the whole thing becomes a house of cards," says Lewin. "But if you're solving one thing at a time — you have a new product idea, you put it into market, you see product-market fit, then you scale it and see that working — the risk you're taking on with the next round of funding is contained. You're not taking on so much that if these bets don't work out, it all comes crashing down. Our business was never at risk.”
We didn't take the position of ‘Let’s raise as much as possible.’ We raised as much as we thought we needed to get to a potential place of profitability. Once we get to that place, we would decide, ‘Do we want to slow things down and be profitable, or do we want to turn things up and raise again on the new story with the new opportunity?’
– Tom Vykruta, co-founder of EvolutionIQ
Saltzman adds: "I saw so many founders contorting themselves into a position where the only successful outcome was an extremely, extremely unlikely event. We never wanted to put ourselves in that position," he says. “We didn't work with investors who would be horrible to work with just because they offered us the highest valuation. And we didn't tell people that we're going to be ‘the AWS for insurance’ or whatever," he says.
“Our belief is that if you build a company that has more ARR than net money spent, that's good. So we've always kept those two numbers in mind as we've built,” says Saltzman. “If you’re on the opposite end of the spectrum of that, maybe it works itself out 15 years from now because you build some crazy thing, but in the medium term, that's going to be very painful.”
Start building systems ASAP — but recognize where to keep the startup spirit
“One thing I think founders miss is changing their managers' mentality from startup to scale-up — and doing that earlier than you think is necessary," says Lewin. “It’s not sustainable for the founders to stay hands-on in every aspect of the business. I can't hold everything in my head that we're doing — we have 200 people, nine different products with a lot of complexity, and too many customer relationships that are all very important.”
Sometimes there's resistance to this shift. “People react like, 'This feels like big company stuff, why are we doing this?’” says Lewin. "It doesn't mean your entire business needs to be scale-up immediately. We still have some teams that are startup-y and some that are scale-up. The goal is to get them all there eventually, but you have to do it thoughtfully.”
At EvolutionIQ, it was trickier to get the sales team to the scale-up phase. “We tended to hire the exact number of salespeople that we thought we needed, and then we would find out that half of them were not the right fit — but then we would have half the number of salespeople we actually needed. It’s also really difficult to get out of founder-led sales,” he says. “But one of our biggest sales victories last year was that there were a couple of accounts where Mike had zero involvement, which was phenomenal.”
Here’s the warning sign of where you need to spend more time on systems: “When you're just fixing stuff and ICing a lot, it's helpful for someone else to try to think through your day with you to figure out where your time is going, and identify what needs to change,” says Lewin. “Wherever I saw people being heroes, or wherever I saw senior leaders firefighting, I would always try to focus on why it was happening: Who else do we need to hire? What are we scoping incorrectly? What are the systems that are broken here?"
If you need people to be heroes for the business to work, that means you actually don't have a great business. You're lucky to have them, but you can’t scale a company on that dynamic.
– Jonathan Lewin, co-founder of EvolutionIQ
Make recruiting your #1 execution priority
For Vykruta, talent was the foundation that underpinned EvolutionIQ's disciplined execution. "Our early team was perhaps one of the biggest reasons that we succeeded as a company — outside of everything else that we did correctly. We wouldn’t have pulled it off without Karan, Georg, Benjamin, and Stanley," he says.
This conviction shaped how Vykruta approached his own role."A founder should make 'recruiting top 1% talent' the number one task on top of their weekly priorities," he says. "I’m firmly in the camp that a company with top talent can get a lot of things wrong and still win in the market, but that a company doing everything right while hiring the wrong talent is unlikely to win any market."
On top of the founder-led effort, the EvolutionIQ founders packed in additional investment here. "I hired a full-time in-house recruiter in the first year of founding the company, and immediately added full-time sourcers," says Vykruta. "Everyone thought this was mad — most startups rely on outsourced recruiting until they are north of 100 headcount."
By the time they were 30 people, they had around four in-house recruiters. "Some investors would say, 'Why don't you have more sales people? And why are you running a recruiting agency?'" says Saltzman. "But that was the denominator. We didn't see it as four recruiters out of a 30-person team. We saw it as four out of the 100-person team we wanted to be a year from now."
The team also made the decision to only hire senior talent, even though it meant higher costs. “I think there's a misconception that if you're a startup, you need to hire junior people to stretch your runway,” says Vykruta. “But I think the reality is the opposite. Startups are very difficult because you're solving foundational problems. There's no infrastructure there at all. Everything is built from the ground up,” he says.
Saltzman adds: "We learned that while you'll pay 30% more for someone with five years experience, they're worth 10 times more."
Never compromise — especially on executive talent
The team's conviction about getting the right people meant being willing to move slowly, even when it hurt. "I spent six or seven months not having a team, just recruiting. I could have hired people in the first week if I wanted to," says Vykruta. "I had clients and investors asking me questions like, 'What's going on? Why don't you have any hires?' We had all this money in the bank, our client was looking for results, and I was losing sleep,” he says.
I was convinced that we need to get the right people — not the first people we could find and convince.
– Tom Vykruta, co-founder of EvolutionIQ
This overinvestment in — and patience for finding — talent wasn’t just confined to the scrappy early days, either. “I interviewed something like 40 or 50 people for our VP of Product position, and it took a year to get the right person,” says Lewin.
“It was very painful having the role unfilled for that long, but it makes all the difference in the world. Every time I talk with him, I'm so happy that we waited a year. Virtually all of the startups that I advised that have run for a while have compromised hiring at some point, and it always becomes a massive problem. It takes much longer than you’d think to reset from it.”
Fight for the perfect hire for your leadership team. It’s indescribably better to wait to get exactly the right person.
– Jonathan Lewin, co-founder of EvolutionIQ
Another area to never compromise on is a candidate’s ability to create strong working relationships. “When we hire senior people, we have all the folks who they'll work with interview with them, and I want a very strong yes — from everyone. If someone's not a yes, I'll examine why. Fundamentally, there's usually something there, even if I can't quite understand it — it typically means downstream there's probably going to be a problem, says Lewin.
“If you ignore these kinds of flags, your exec team can easily evolve into a ‘Johnny ate my sandwich’ dynamic, with people fighting all the time and you having to separate them and pour all this energy into just trying to get these relationships to work,” he says.
PARTING ADVICE ON MANAGING YOUR FOUNDER MINDSET AND KEEPING YOUR EYE ON THE (BUSINESS) BALL
There’s a simple truth that's often glossed over in startup circles: The path is always harder than it looks from the outside.
"There's a whole category of problems that are not discussed in management and startup books," says Vykruta. "Founders should realize that 50% of what they need to know will be in the books, and 50% will not be — because it's just not appropriate for book material. Elon put this best: He says doing a startup is like eating glass and staring into the abyss. Accepting that this is the norm — not the exception — was really hard for me,” he says.
“Initially, I always felt like every startup around us was doing so well, hiring quickly and scaling up revenue, and we were just struggling and struggling. Eventually, I came to realize that this is the norm. Every startup is struggling, and it feels like it’s on the verge of collapse and the founder is not getting any sleep. The sooner you can accept that reality, the sooner you can be a little bit easier on yourself,” says Vykruta.
“What I always tell founders who are just starting out is that you have to work really, really, really hard, especially in the early stages. It’s an absolute must. It's not an option,” he says. “And it does not set you apart from other founders. It does not guarantee success. It only gives you the possibility of success.”
Saltzman’s parting piece of advice is for AI founders to direct that energy into a particular area. “If you're trying to build an application layer business like we have using AI, you have to figure out what's going to be your moat,” Saltzman says.
Planning for success is really important. This is obvious, but it can’t be overstated enough because it’s skipped over with surprising frequency.
– Mike Saltzman, co-founder of EvolutionIQ
“Assume it's going to work, because if it doesn’t work then nothing else matters, game over. But assume that it works, and it’s three years from now, when you have $10 million in ARR and 30 customers — what’s going to be your moat so another company can't do a weekend project and build the very same chatbot or agent?”
That answer might change over time. That's fine, Saltzman says, but parts of the answer should include:
- How is my data set going to be materially different than what others can get?
- How is my workflow going to be materially different and more nuanced?
- How is my relationship with my customer from a commercial or distribution perspective going to be materially different?
“If you can't answer those questions, you're probably going to be in a bad spot. I think part of why EvolutionIQ worked is that we did have those answers,” Saltzman says. “Some of them have changed, some of them haven’t. But we have a valuable business today because: 1) the data is something no one else can get, 2) the workflow is intensively nuanced, and 3) there's a ton of work in the application layer — not just the AI layer — and 4) our word-of-mouth distribution has such compounding, almost network-effects because of the insular relationships in insurance. No one would fund a business to compete with us at this point, because we already are in a lot of these companies,” he says.
“When it was just software, it was all about the product. Now, it’s so much easier to build a good product — which is awesome for a lot of reasons — but that also means that building a good business has become that much harder. Figuring that out is where I would focus a lot of my time and attention if I were a founder starting an AI-native company today.”