I get pitched on “AI for Radiology” at least three times a week. Seriously. Every founder thinks they’re the one who will crack it. They show me a slick deck, a model with 98% accuracy, and a massive TAM slide. I pass every single time.
Most people in Silicon Valley are chasing the same shiny objects. Right now, that shiny object is using AI to read medical scans. It makes sense on the surface. You have a ton of data (X-rays, MRIs), a clear problem (doctors are overworked), and a seemingly straightforward application for computer vision. It’s the perfect recipe for a pitch deck.
But it’s a trap. A bubble. And it’s about to burst.
I’ve seen this movie before. I’ve built and sold two companies, RemoteTeam to Gusto and MovieLaLa to Gfycat. I’ve put my own money into over 200 startups, including some of the biggest names in AI like Anthropic, OpenAI, and Scale AI. I’ve learned a thing or two about what separates a good idea from a real, venture-scale business. And I’m telling you, most of the “AI in Radiology” companies are not it.
The Uncomfortable Truth About Your "Groundbreaking" Model
The first thing every founder tells me is how great their model is. “We have a state-of-the-art deep learning algorithm that can detect tumors with 99.5% accuracy!”
I have to be blunt: nobody cares.
Your model is not a moat. It’s a commodity. In 2025, building a high-performing computer vision model is table stakes. You’re likely using a tweaked version of a well-known architecture like a Vision Transformer or a ConvNet, trained on a public or semi-public dataset. Guess what? So is everyone else. The small accuracy gains you’re so proud of are rounding errors in the real world. Within six months, another team will come along and beat your benchmark.
At MovieLaLa, we had a real moat. We had a unique dataset of movie-watching habits from millions of users. That data was ours, and it allowed us to build recommendation engines that Netflix couldn’t just copy. Most radiology startups don’t have that. They’re building on quicksand.
Welcome to the Hospital Sales Grind
Let’s say you have a genuinely better model. Now comes the hard part: selling it. Founders who have only ever worked in tech have no idea what they’re in for. Selling to hospitals is not like selling SaaS to other startups. It’s a soul-crushing grind.
You’re not dealing with a single decision-maker. You have to convince the radiologist, the department head, the hospital administrator, the IT department, and the legal team. Each has their own incentives, and most of them are not aligned with yours. The radiologist is worried about their job. The administrator is worried about the budget. The IT department is worried about security and integration. The legal team is worried about liability.
I once invested in a promising healthcare AI company. They had a fantastic product that could save hospitals millions. It took them 18 months—eighteen months!—to close their first major hospital contract. That’s an eternity in the startup world. By the time you’ve closed a few deals, you’ve burned through half your seed funding just on sales and marketing.
The FDA Is Not Your Friend
And then there’s the big one: regulation. If your AI is making a clinical diagnosis, you need FDA approval. This is a long, expensive, and unpredictable process. You’re not just submitting code; you’re submitting a mountain of documentation, clinical trial data, and quality control procedures.
I’ve seen companies spend years and millions of dollars navigating the FDA, only to be rejected or asked for more data. It’s a massive barrier to entry, and it slows you down to a crawl. While you’re stuck in regulatory limbo, the market is moving on.
Where the Real Opportunities Are
So if AI in radiology is a bubble, where should ambitious founders be looking? Instead of chasing the same overcrowded, overhyped idea, focus on the less glamorous but far more valuable problems in healthcare.
1. The Boring Backend Stuff
This is the unsexy stuff that actually makes hospitals run. Billing, scheduling, patient records, insurance claims. It’s a mess of legacy systems, manual data entry, and endless paperwork. This is where AI can have a massive, immediate impact.
Think about it. You don’t need FDA approval to build a better scheduling system. The sales cycle is still long, but you’re selling efficiency and cost savings, which every administrator understands. The data is messy, but if you can clean it up and automate these workflows, you can build a huge, defensible business.
2. Reinventing Drug Discovery
This is the moonshot, but it’s a moonshot worth taking. Developing a new drug costs billions of dollars and can take over a decade. AI has the potential to fundamentally change that. My investment in Anthropic is a bet on this future. Their work on constitutional AI and large-scale models can be applied to understanding biology and designing new molecules in ways we couldn’t before.
This is a much harder technical problem than reading X-rays, but the prize is infinitely bigger. You’re not just building a feature; you’re building a new paradigm for an entire industry. The moats here are deep—based on fundamental scientific breakthroughs and massive computational scale.
3. AI for Mental Health
The world is facing a mental health crisis. There are not enough therapists to meet the demand. This is a perfect place for AI to step in. AI can provide personalized, accessible, and affordable mental health support to millions of people.
This could be anything from AI-powered chatbots that provide cognitive behavioral therapy to tools that help therapists be more effective. The need is huge, the market is growing, and you can often build products that don’t require the same level of regulatory oversight as diagnostic tools.
Stop Chasing the Bubble
Look, I get the appeal of AI in radiology. It’s a clear, understandable problem. But the path to building a successful company there is a minefield. The competition is fierce, the moats are shallow, the sales cycles are brutal, and the regulation is a killer.
Instead of being the 100th company trying to read a chest X-ray, go find a real, unsolved problem. Go tackle the messy backend systems. Go for the moonshot in drug discovery. Go help solve the mental health crisis.
Stop chasing the shiny object. The biggest opportunities are always in the places no one else is looking.
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.
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.
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.