Most founders building AI for healthcare are going to fail.
That’s not me being a pessimist. It’s a hard-won lesson from over a decade in the Silicon Valley trenches, two successful exits, and writing checks to 200+ startups. I’ve seen more pitch decks than I can count, and the pattern is always the same: a brilliant team, some sexy tech, and a fundamental misunderstanding of what it actually takes to win in this brutal, unforgiving market.
They all think a great AI model is enough. They chase that extra percentage point of accuracy, convinced that if they can just build a better algorithm, customers will line up.
They’re dead wrong.
I’ve watched founders with genius-level IQs and tech that could literally save lives crash and burn because they couldn’t get their product into the hands of doctors. They couldn’t navigate the labyrinth of the FDA. They couldn’t figure out a business model that wasn’t DOA in a hospital CFO’s office. It’s a painful, expensive, and incredibly common story.
One of my angel investments—a company I won’t name, let’s call them ‘CardioVision’—had an AI that could predict heart attacks from a standard EKG with stunning precision. The team was stacked with PhDs from MIT. Their model was, without a doubt, a technical masterpiece. They came to me for a seed round, and I wrote the check. A year later, they were on life support, burning through cash with almost zero revenue. Why? They were trying to sell it as a SaaS subscription to cardiologists. The doctors loved the demo, but they wouldn’t pay for it. It was just another piece of software in a sea of software. They were selling a feature, not a solution.
This isn’t a one-off story. It’s the default outcome. And frankly, I was tired of seeing it. I was tired of seeing incredible potential evaporate because of avoidable mistakes. That’s why my team and I embarked on a massive project: to stop guessing and start knowing what separates the winners from the losers in healthcare AI.
The Signal Through the Noise
We didn’t just want to build another black-box model. We wanted to decode the DNA of success in clinical trials. So we got our hands dirty. We pulled the data from over 10,000 clinical trials, a messy, sprawling dataset of successes, failures, and everything in between. We looked at everything. The specific biomarkers targeted. The design of the trial protocol. The patient demographics. The founding team’s background. The amount of capital raised and from who.
We then built an AI to analyze it all. We didn’t train it to think like a scientist. We trained it to think like me—like an investor who has to decide where to put real money. We forced it to weigh not just the scientific plausibility but the entire ecosystem around a drug.
The result? After months of training and validation, our model can predict the likelihood of a drug passing its clinical trials with 87% accuracy.
Let that sink in for a second. The historical success rate for drugs entering clinical trials is less than 10%. The industry literally burns billions of dollars every year on compounds that are destined to fail. An 87% predictive signal isn’t just an improvement; it changes the entire game. It’s like going from navigating by the stars to using a GPS.
How is this possible? Because we looked beyond the molecule. Our AI found that factors most data scientists ignore are often the most predictive:
- Team Composition: A founding team with a mix of a seasoned pharma exec and a machine learning expert is dramatically more likely to succeed than a team of only academics or only business people.
- Regulatory Pathway: Companies that have a pre-submission meeting with the FDA and a clearly defined regulatory strategy have a 3x higher success rate. It’s not about having the best tech; it’s about knowing how to get it approved.
- Funding Quality: It’s not just how much money you raise, but who you raise it from. Startups backed by venture funds with a dedicated healthcare IT team and a track record in the space were far more successful. Smart money is real.
- Trial Design: This was a big one. Using adaptive trial designs, where the protocol can be modified based on interim results, massively boosts the probability of success. It’s about being nimble, not just rigorous.
This isn’t just about picking winners. It’s about making drug development fundamentally more efficient. It’s about redirecting capital from doomed projects to promising ones. It’s about giving smaller, more innovative companies a fighting chance. Ultimately, it’s about getting life-saving treatments to patients years faster.
From Signal to Business
But remember CardioVision? An 87% signal is worthless if you can’t build a business around it. This is the second, and harder, part of the equation. This is where you have to stop thinking like a scientist and start thinking like a killer entrepreneur.
After their near-death experience, we sat down with the CardioVision founders for a brutal, honest strategy session. I told them to stop selling software. Nobody wants more software. I told them to start selling an outcome. What’s the outcome? Preventing heart attacks.
We completely changed their business model. Instead of a SaaS fee, they started a risk-sharing model with hospitals. They would get a small fee for every patient screened, but they would get a massive bonus for every heart attack they successfully predicted and helped prevent.
Suddenly, the entire conversation with hospital administrators changed. They weren’t buying software anymore. They were buying a guaranteed ROI. They were buying better patient outcomes that they could market to their community. They were mitigating their own financial risk. Within 18 months, CardioVision was in over 70 hospital systems and was acquired by a major medical device company.
They didn’t change their model. They changed their business. That’s the lesson.
The Future is Not Evenly Distributed
We are at the very beginning of a complete rewiring of the healthcare industry. The fusion of AI and biology is the biggest investment opportunity of my lifetime. I’m not just talking about drug discovery, but diagnostics, radiology, personalized medicine, and hospital operations. Everything is on the table.
But the winners won’t be the ones with the fanciest algorithms. They will be the ones who are obsessed with the problem, not the solution. They will be the ones who live and breathe the messy, complex, human reality of healthcare. They will have the humility to listen to doctors and the audacity to challenge the status quo.
I’m putting my money where my mouth is. I’m actively seeking out and funding the founders who get this. The ones who are building not just models, but real, scalable businesses that can survive in the healthcare jungle. The ones who understand that 87% accuracy is just the starting line.
If that’s you, find me. The world needs what you’re building. Just don’t try to sell it to me as a SaaS subscription.
Frequently Asked Questions
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.
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.
What's the most common pushback you get on this?
People often push back by citing exceptions or edge cases. And they're usually right that exceptions exist. But building a strategy around exceptions rather than patterns is a losing game for most founders.
Do all experts agree with this view?
No, and that's fine. The best ideas in business are often contrarian. I share my perspective based on my experience and data, but I encourage you to seek out opposing viewpoints and form your own conclusions.