I remember sitting in a Sand Hill Road boardroom, listening to a pitch from a startup that promised to revolutionize radiology. They had a slick deck, a team of brilliant data scientists, and an algorithm that could detect lung nodules with 99.9% accuracy. On paper, it was a slam dunk. They were going to save lives, make doctors obsolete, and generate a 100x return for their investors. I passed on the deal.
That was three years ago. Today, that company is a shadow of its former self, struggling to get its algorithm adopted in a single hospital. Why? Because they made the same mistake I’ve seen dozens of other AI radiology startups make: they believed that a great algorithm was enough.
After two successful exits and over 200 angel investments in companies like Anthropic, OpenAI, and Scale AI, I’ve learned to spot the difference between hype and true potential. The hype around AI in radiology is deafening, but I’m here to tell you it’s a bubble about to burst. Building a strong AI model isn't the full story. From my experience, success in healthcare AI demands navigating tough regulations and real-world challenges that most founders overlook.
The Allure of the Algorithm
It’s easy to see why everyone is so excited about AI in radiology. The potential is enormous. We’re talking about technology that can analyze medical images faster and more accurately than the human eye, catching diseases earlier and reducing the workload on overworked radiologists. The market is projected to be worth billions of dollars, and every venture capitalist in Silicon Valley wants a piece of the action.
I get it. I’ve been there. When I was building my first company, RemoteTeam, I was obsessed with the product. I thought that if we just built the best remote work platform on the market, customers would flock to us. But I quickly learned that a great product is only one part of the equation. You also need to figure out distribution, sales, marketing, and a dozen other things that have nothing to do with code.
It’s the same story with AI in radiology. These startups are so focused on perfecting their algorithms that they completely ignore the messy reality of the healthcare system. They forget that their technology has to be integrated into existing hospital workflows, that it has to be approved by regulators, and that it has to be used by real doctors and patients in the real world.
The Messy Reality of Healthcare
I once invested in a promising AI diagnostics company that had developed a groundbreaking algorithm for detecting early-stage pancreatic cancer. The team was brilliant, and their technology was a decade ahead of its time. But they hit a wall when they tried to get their product into hospitals. The IT departments were a nightmare to deal with, the doctors were resistant to change, and the regulatory hurdles were a maze of red tape.
Here’s a little secret most people outside of healthcare don’t know: hospital data is a mess. It’s fragmented, it’s siloed, and it’s often stored in outdated systems that don’t talk to each other. You can have the best algorithm in the world, but if you can’t get access to clean, high-quality data, it’s worthless. This is a lesson I learned the hard way with one of my portfolio companies. They spent two years and millions of dollars trying to integrate their system with a major hospital’s electronic health record (EHR) system, only to give up in frustration.
And then there are the regulations. The FDA is notoriously slow and cautious when it comes to approving new medical devices, and for good reason. These are products that can have life-or-death consequences. The approval process can take years and cost millions of dollars, and there’s no guarantee of success. I’ve seen more than one promising startup run out of money while waiting for the FDA to make a decision.
Where We Should Look Next: The Promise of Biotech and Drug Discovery
So if AI in radiology is a bubble, where should we be looking for the next big breakthrough in healthcare AI? I believe the answer lies in biotech and drug discovery.
Unlike clinical diagnostics, where the bar for accuracy and safety is incredibly high, the world of drug discovery is all about experimentation and iteration. It’s a numbers game. The more shots on goal you can take, the more likely you are to score. And that’s where AI can have a massive impact.
AI algorithms can analyze vast datasets of genetic and molecular data to identify new drug targets, predict how a drug will behave in the human body, and design clinical trials more efficiently. This is not science fiction. Companies like Recursion Pharmaceuticals and Insitro are already using AI to discover new treatments for rare diseases, and they’re making incredible progress.
I’m an investor in several biotech AI startups, and I’m incredibly bullish on this space. The challenges are still immense, but the potential for impact is even greater. We’re talking about curing diseases that were once considered incurable, and extending human lifespan by decades. That’s a vision worth fighting for.
A Dose of Realism
I’m not saying that AI has no future in radiology. It does. But I believe the path to adoption will be much slower and more difficult than most people think. The winners won’t be the companies with the fanciest algorithms, but the ones who can navigate the complex and messy reality of the healthcare system.
As for me, I’ll continue to invest in founders who have a deep understanding of the problems they’re trying to solve, and who are not afraid to get their hands dirty. The road to revolutionizing healthcare is a long and winding one, but with the right team and the right approach, I believe we can get there. The real impact won't come from a single algorithm, but from a deep and nuanced understanding of the entire healthcare ecosystem. That's where I'm placing my bets.
Frequently Asked Questions
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.
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 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.