General Catalyst's Alex Tran sits down with Max Kauderer and Ryan Lee to explore why America's health insurance system is broken, how a fintech mindset is reshaping claims processing, and what Yuzu is doing to rebuild health plan administration from scratch.
This interview has been edited and condensed. Watch the full video below.
.png)
Alex Tran: You're both repeat founders. Before we get to Yuzu, how did you meet?
Max Kauderer: We met about five years ago through our third co-founder, Russell.
Ryan Lee: Russell and I ran a dating site in college called Data Match, a matchmaking service that had been around Harvard for years. We put it on the internet and spread it across the US, and it's still active at a couple dozen schools. I was the engineer and he was the operations guy, even then. We kept in touch, and one day he said he knew the perfect person to complement us and introduced me to Max. We got lunch, and within 30 minutes we'd decided to do this together.
Alex Tran: And you both came from fintech.
Max Kauderer: I started my first company at 18 in legal tech. It didn't work out, but I got the startup bug. I was excited about fintech, joined Capital One's strategy group, a hundred-billion-dollar bank that's still founder-led, then went to Bain before jumping into Yuzu.
Ryan Lee: I spent a couple of years at Lithic, a payments API. The company had transitioned from a consumer product, Privacy.com, into API-driven backend infrastructure. A lot of how we thought about Yuzu from day one came from watching that.
Alex Tran: Tell us about the problem Yuzu is solving.
Max Kauderer: It's clear to anyone in the US that something isn't working. Premiums in '26 saw the steepest cost increase in 15 years, and '27 looks worse. NPS among the major five insurers is at an all-time low. There's no transparency into how dollars are spent, what things cost, or when claims are paid. It's a black box. Yuzu's mission is to bring trust and transparency back into healthcare.
We got here by accident. Four years ago, we set out to build a new health insurer for startups, asking why there was no universal health plan for them. As we dug in, we found a health plan is really a triangle: plan design (the network, the deductible, how care is navigated), risk (who pays a catastrophic claim, usually outsourced to reinsurers), and admin (processing claims, eligibility, moving money, reporting).
On plan design and risk, there were turnkey options. But admin, third-party administration (TPA), was the problem. The only TPAs were legacy companies with 15 solutions stitched together, outsourcing claims and reporting to five different firms behind the scenes. None could support the nuances we wanted. So out of necessity we decided to rebuild the TPA from first principles. That's what we are today: an operating system for health plans.
Alex Tran: Who are your customers, and what can they do now that wasn't feasible before?
Ryan Lee: We're enabling people who thought running a health plan was too hard to actually consider it. One of our customers was formerly a brokerage. They saw a gap in the plans available and wanted to become the health plan, but the blockers were enormous. Our technology lets them do it.
A lot of what the software industry takes for granted, like smooth onboarding or a website to sign into, didn't exist here. Onboarding a group used to mean PDFs back and forth, double-checking by email, calling the HR rep. Just putting it online was revolutionary. TPAs would refuse to go down-market because of all the upfront work, and anything under 50 people was considered a small group. Because we built the software, our customers can send a link and sell to much smaller businesses.
Max Kauderer: Our customers are often on the bleeding edge of how to put members first and contain costs. Before Yuzu they were stuck if they wanted anything other than a standard plan. We give them a nimble, flexible platform where they can configure their own plans, fully white-labeled.
Alex Tran: So you're powering this groundswell of innovation in healthcare, building the foundational technology that lets these new plans exist. What about the other end of the spectrum? Today, there are also very large insurers serving hundreds of millions of Americans. Are those in scope too?
Max Kauderer: The biggest insurers are still pushing the edge of innovation. United's fastest-growing product, Surest, is a dynamic copay plan, and most of that claims processing relies on third-party administrators. Whether it's a massive carrier testing a new incentive structure or a startup, they all need flexible core infrastructure. That's what we offer.
Alex Tran: You've been opinionated about owning every piece of software in-house. Why?
Ryan Lee: Traditionally, a company like us would outsource the big components—claims processing software, especially. By bringing it all in-house, we get to be the responsible party. A lot of healthcare frustration is just not knowing what's going on. You call about a claim and support can't answer, because in the traditional playbook it's someone else's problem. When it's in-house, we can say exactly why a claim was adjudicated the way it was. That keeps the whole company sharp, and it lets us move fast and automate instead of building integrations with everyone else.
It surprises engineers when they join. They ask, "So you integrate with an insurance company?" No. "You work with the claims adjustment firm?" No. We print our member ID cards in-house, on a machine in our office. That's how deep this goes.
Alex Tran: Where does AI fit into the system of record you've built?
Ryan Lee: I think of an LLM like a brain floating in a jar, dying to do something but unable to act on its own. We've spent the past couple of years building the body: claims adjudication, invoicing, the workflows our team and customers operate. Once that's fleshed out with the right checks and balances, you can attach the brain and let it automate safely, suggesting edits rather than going wild.
Max Kauderer: Early on we tried to partner with other TPAs and offer them agentic solutions, and we realized it's impossible. Their data lives in 14 different systems and no one knows what's accurate. You can't unleash an LLM without a source of truth. In a world where software is cheaper and faster to build, the way to differentiate is by being the system of record. It's harder and slower to build, but once you have everything in one place, slotting in an LLM becomes almost trivial.
Alex Tran: Some might call you outsiders. What first-principles thinking did you bring from fintech?
Ryan Lee: At Lithic, I was shocked by how much happens when you swipe a credit card, a million things in under a second, all of which we take for granted. That trained a muscle: dissect what something actually is.
At Yuzu, the standard is that a claim must be paid in 30 days. My first question was, why 30 days? Why not 30 seconds, like a card payment? No one had sat down and studied a claim the way fintech studied payments for a decade.
There's real nuance. For example, a hospital sends a claim, then an edit to it: "we forgot a service, you owe a different amount." Most systems treat those as two separate claims and void the original. Our engine ties them together as one claim, which has major ramifications for how we adjudicate. The last couple of years have been about defining those fundamentals, like what is a medical claim and what is a pre-certification.
Max Kauderer: We bonded over a book called The Anatomy of the Swipe, about everything that happens behind a card transaction. Card payments are almost a solved problem, and a million fintech companies do it well. But claims processing, which is far larger in dollar volume, has nothing close to that sophistication. We had the naivety to think we'd take 30 days to three seconds out of the gate. It's much harder than that, but that naivety is the first step to actually doing it.
Alex Tran: The TPA business is famously low-margin and services-heavy. What changes?
Ryan Lee: In the early days I was taking support calls and Max was initiating invoices. We did it manually to understand what actually needed to happen. The existing software was written for a world that no longer exists. Coming in with a clean slate, we saw most of the service-heavy work could be automated, and a lot of it could simply be deleted.
Max Kauderer: Our principle is delete before automate. The best thing isn't automating a process, it's removing one that shouldn't exist. Many of the plans we power use cash pay: a member gets a credit card, pays the provider upfront, and pulls funding directly from their plan. The provider gets paid in real time and takes a discount, and the member shops for care. On the back end there's no traditional claim, so the filing and prior-auth processes disappear entirely.
On margins, these TPAs run 10 different systems and each one costs money, before you count integration and complexity. We rebuilt them as one integrated system, so our cost basis, and our margin profile, is fundamentally different.
Alex Tran: Tell us about the team and culture.
Max Kauderer: Team is everything. Because we don't come from a TPA background, we over-index on aptitude, curiosity, and slope: how quickly someone can dive in and contribute. Every person who joins has to get in the weeds of health insurance, paying invoices, printing ID cards, handling member and provider calls. In a world where AI makes product development cheaper, it's critical that smart, eager people have high context, because then they know what can be automated and what can be deleted.
Ryan Lee: Late last year our ID card machines literally melted from overuse, right before the holidays. It was a code red. Engineers dropped everything to stuff cards into envelopes and get them to the post office. Was that the highest-leverage use of an engineer's time? I'd argue yes. A couple of them immediately realized it needed to be automated, and that became a tentpole project. Feeling the pain firsthand is the best way to build empathy, and empathy is exactly what we hire for.
Max Kauderer: At our offsites we have engineers and ops folks swap roles for a few hours. Engineers answer support calls and print ID cards, and ops folks go into the codebase and use Claude to write code. A number of our ops people are in the codebase now, even adding small features. That's been surprisingly successful.
Alex Tran: If you do this right, what does the healthcare system look like in 10 or 20 years?
Max Kauderer: I don't want to be prescriptive, and that's the beauty of what we're building. Health insurance will evolve, and there will be an explosion of new plans: upstarts, incumbents launching new lines, maybe hospital systems launching their own. The challenge is there's no infrastructure to let them. My hope is we're the platform that powers all of it, a catalyst for change, and that it ends up more affordable and more transparent for everyone.
Ryan Lee: Stripe's early ethos was to not be too prescriptive. When you create a new technology, it's almost arrogant to think you can push it exactly where you want. It takes itself where it's best suited.
Alex Tran: Healthcare is a $5 trillion problem in America, larger than the GDP of all but a few countries. Even a small dent would be good for all of us.
%201.svg)
