''' I remember sitting in a pitch meeting a few years back. A couple of brilliant PhDs, armed with a model that could predict the likelihood of a rare disease with uncanny accuracy. They thought they had it all figured out. They had the tech, the brains, and the passion. They were wrong. Six months later, their company was dead.
Why? Because they made the classic mistake that I see over and over again in the world of healthcare AI. They thought the model was the business. It's not. Not even close.
For the past decade, I've been in the trenches of Silicon Valley. I've built and sold two companies, and I've written checks to over 200 startups, including some of the biggest names in AI like Anthropic, OpenAI, Scale AI, and Hugging Face. I've seen firsthand the brutal reality of what it takes to win in this space. And I'm here to tell you the uncomfortable truth: the future of personalized medicine isn't about genomics, it's about AI-driven predictive health. But building a successful company in this space has almost nothing to do with the elegance of your algorithm.
The Data Moat is a Mirage
Every founder I meet tells me they're going to build a "data moat." They believe that once they have enough proprietary data, they'll be untouchable. It's a nice idea. It's also a fantasy.
Healthcare data is a nightmare. It's fragmented, siloed, and riddled with errors. Every hospital system has its own EMR, its own data formats, and its own gatekeepers. I once invested in a company that was trying to build a predictive model for sepsis. They spent the first year and a half just trying to get clean data from a single hospital. It was a constant battle of politics, paperwork, and technical debt. By the time they had a workable dataset, a dozen other companies had popped up with the same idea.
The real moat isn't the data itself, it's the infrastructure you build to clean, process, and learn from it. It's the relationships you build with the hospitals and the trust you earn from the doctors. That's the hard part. That's the part that no one wants to talk about.
The FDA is Your Co-Founder
Another classic mistake is treating the FDA as an afterthought. Founders will spend millions developing a product, only to realize at the last minute that they have no clear regulatory path. I've seen it happen time and time again. A brilliant diagnostic tool that gets stuck in regulatory limbo for years, bleeding money until it finally dies.
You have to think about your regulatory strategy from day one. You have to build your product with the FDA's requirements in mind. This means hiring regulatory experts early, running the right clinical trials, and being prepared for a long and arduous process. It's not sexy, but it's the difference between success and failure.
One of my most successful investments in the space was a company that was developing an AI-powered tool for detecting skin cancer. From the very beginning, they worked hand-in-hand with the FDA. They designed their trials to meet the agency's specific requirements, and they were transparent about their data and their methods. It took them longer to get to market than some of their competitors, but when they did, they had a product that was not only effective but also trusted by doctors and regulators. They were acquired for a handsome sum, not because their model was so much better than everyone else's, but because they had a real, defensible business.
Doctors Are Not Your Users (At First)
This is a tough one for a lot of founders to swallow. They build a product that they think doctors will love, only to find that they can't get anyone to use it. The reason is simple: doctors are not the ones who buy software. Hospital administrators are. And hospital administrators care about one thing: ROI.
Your pitch to a hospital can't be about how cool your tech is. It has to be about how you're going to save them money, improve patient outcomes, or reduce readmission rates. You have to speak their language. You have to show them the numbers. I had to learn this the hard way with my own startup, RemoteTeam. We were building a great product for remote teams, but we were selling it to the wrong people. It wasn
wasn't until we started selling to the HR departments and the CFOs that we started to get traction. The same is true in healthcare.
Your first users are the hospital administrators. Your champions are the innovative doctors who are willing to stick their necks out for you. But your buyers are the people who hold the purse strings. You have to convince them that your product is not just a nice-to-have, but a must-have.
Your Model is a Commodity, Your Workflow is the Product
Let's be honest. For most predictive health applications, the model itself is becoming a commodity. With the rise of open-source libraries and pre-trained models, a team of decent engineers can replicate your core algorithm in a matter of months. I've seen it happen. A hotshot startup raises a huge round based on a supposedly revolutionary model, only to be blindsided by a competitor who built something similar for a fraction of the cost.
Where you create real, defensible value is in the workflow. How does your tool fit into the day-to-day life of a clinician? How does it make their job easier, faster, or more effective? A model that spits out a probability score is useless. A tool that integrates seamlessly into the EMR, flags high-risk patients, and suggests a clear course of action—that's a game-changer. But you can't call it that.
Think about it from the perspective of an overworked doctor. They're juggling dozens of patients, drowning in paperwork, and facing constant pressure to see more people in less time. The last thing they want is another screen to look at or another number to interpret. They want a tool that simplifies their life, not complicates it.
One of the best investments I ever made was in a company that built an AI tool for radiologists. Their model wasn't necessarily the most accurate on the market. But what they nailed was the user experience. Their tool highlighted potential areas of concern on an X-ray, automatically pulled in relevant patient history, and even pre-drafted a report. They didn't just sell a model; they sold a complete, end-to-end workflow. Radiologists loved it because it saved them hours of tedious work. The hospital loved it because it increased throughput and reduced errors. That's how you win.
The Real Business Model is Not SaaS
Everyone in Silicon Valley is obsessed with the SaaS model. Recurring revenue, high margins, it's the holy grail. But in healthcare, a pure SaaS play is incredibly difficult to pull off. Hospital budgets are notoriously tight and sales cycles are brutally long. You can spend years trying to convince a hospital to sign a multi-year contract, only to have the deal fall through at the last minute because of a budget cut or a change in leadership.
I've found that the most successful healthcare AI companies use a hybrid model. They might have a subscription component, but they also have a shared-risk or outcomes-based component. For example, a company that predicts hospital readmissions might get a bonus for every patient that doesn't come back within 30 days. This aligns your incentives with the hospital's. You're not just selling them software; you're selling them a result.
This is a much harder model to build. It requires you to have a deep understanding of the hospital's finances and to be able to track and measure your impact. But it's also a much more powerful model. When you can go to a hospital and say, "You only pay us if we save you money," that's a very compelling proposition.
Stop Chasing the Unicorn, Start Building a Camel
I'm going to say something that might be controversial in the Valley. Stop trying to build a unicorn. The blitzscaling, grow-at-all-costs mentality that works for consumer apps is a recipe for disaster in healthcare. You can't just throw money at the problem and hope for the best.
You need to build a camel. A company that is resilient, adaptable, and can survive in the harshest of environments. A company that is built for the long haul. This means being frugal, focusing on sustainable growth, and building a real business with real revenue.
I'm not saying you shouldn't be ambitious. But you have to be smart. The future of medicine is being written in code, but the story isn't about the code itself. It's about the messy, complicated, and deeply human process of bringing that code to life. It's about navigating the labyrinth of data, regulation, and hospital politics. It's about building something that doctors will actually use and that hospitals will actually buy. If you can do that, you won't just build a successful company. You'll change the world. And that's a hell of a lot more interesting than just building a model.
Frequently Asked Questions
How can I apply this thinking to my own situation?
Start by identifying the core principle behind the opinion, not the specific example. Then ask yourself: does this principle apply to my context? If yes, test it in a small, low-risk way before going all in.
How has this view evolved over time?
My thinking on most topics has changed significantly over the years. Early in my career, I held many conventional views that experience proved wrong. I try to update my beliefs when the evidence changes.
What experience informs this perspective?
This perspective comes from over a decade of building companies in Silicon Valley, two successful exits (RemoteTeam to Gusto, MovieLaLa to Gfycat), and investing in 200+ startups including Anthropic, OpenAI, and Scale AI. I write about what I've lived.