Most founders in the biotech space think that building a fantastic AI model is enough to guarantee success. They are mistaken. I’m going to share the unvarnished truth about what it takes to truly succeed in the highly regulated and competitive world of healthcare AI. After more than a decade in Silicon Valley, with two successful exits and over 200 angel investments in companies like Anthropic and OpenAI, I’ve seen firsthand what separates the winners from the losers. Here are the five unspoken rules for building a defensible AI moat in biotech.
1. Data Isn't Just King, It's the Entire Kingdom
Everyone says data is king, but in healthcare, it’s the entire kingdom. The quality and uniqueness of your data are the most significant factors in building a defensible moat. A model trained on generic, publicly available data will never outperform one trained on a proprietary, well-curated dataset. I remember one of my portfolio companies, a promising AI radiology startup, that learned this the hard way. They had a brilliant team of data scientists, but they were using the same public datasets as everyone else. Their model was good, but not great. They only started to see real breakthroughs when they partnered with a large hospital network to get exclusive access to a decade's worth of anonymized patient scans. That’s when their accuracy jumped, and they were able to secure a massive Series A round.
But it's not just about getting access to data; it's about the right data. You need data that is clean, well-annotated, and representative of the patient population you're targeting. This is often the hardest part. I've seen companies spend years and millions of dollars just cleaning and preparing their data before they can even start training their models. It's a long and arduous process, but it's absolutely essential. Don't underestimate the importance of having a dedicated data team that is responsible for data acquisition, cleaning, and annotation. This team is just as important, if not more so, than your data science team.
Another thing to consider is the ethical implications of data collection. In healthcare, you're dealing with sensitive patient information, so you need to be extra careful about privacy and security. Make sure you have a robust data governance framework in place and that you're complying with all relevant regulations, such as HIPAA. A data breach can be catastrophic for a healthcare startup, so don't cut corners when it comes to security.
2. Your Model Is a Commodity; Your Workflow Is the Product
Too many founders are obsessed with their model's architecture. The hard truth is that your model, no matter how sophisticated, is a commodity. There will always be a new, better model just around the corner. What isn't a commodity is a seamless workflow that integrates into a clinician's daily routine. At RemoteTeam, which was acquired by Gusto, we didn't just build a tool for remote work; we built a new way of working. We focused on the entire employee lifecycle, from onboarding to offboarding, and made it effortless for companies to manage their remote teams. The same principle applies to biotech AI. Don't just build a model that can detect a disease; build a tool that a doctor can use without adding to their already overwhelming workload.
Think about the day-to-day life of a clinician. They're overworked, stressed, and short on time. The last thing they want is another complicated piece of software to learn. Your product needs to be intuitive, easy to use, and, most importantly, it needs to save them time. I've seen so many promising AI companies fail because they couldn't get clinicians to adopt their products. They had the best models, the most accurate predictions, but their products were just too clunky and difficult to use.
A great example of a company that gets this right is Flatiron Health, which was acquired by Roche for $1.9 billion. Flatiron didn't just build a great oncology EMR; they built a platform that streamlines the entire cancer care workflow, from diagnosis to treatment to follow-up. They understood that the real value was not in the data itself, but in the insights that could be derived from that data and delivered to clinicians at the point of care. That's the kind of thinking that builds a defensible moat.
3. Regulation Is Not a Bug, It's a Feature (If You Play It Right)
The regulatory hurdles in healthcare are daunting, but they can also be a powerful moat. The FDA approval process is long and expensive, but once you're through it, you have a significant advantage over your competitors. It's a barrier to entry that few can overcome. Instead of viewing regulation as a burden, see it as an opportunity to build a defensible position. I’ve seen companies that embrace the regulatory process from day one and use it to their advantage. They design their studies to meet FDA requirements, and they build relationships with regulators. These are the companies that ultimately win.
Navigating the regulatory landscape is not for the faint of heart. It requires a deep understanding of the rules and regulations, as well as a willingness to work closely with regulators. You'll need to hire a team of regulatory experts who can guide you through the process. It's a significant investment, but it's one that will pay off in the long run. I've seen companies try to cut corners on regulation, and it always ends in disaster. They either get shut down by the FDA or they're forced to go back to the drawing board, having wasted years of time and millions of dollars.
One company that has done an exceptional job of navigating the regulatory landscape is Tempus, which has raised over $1 billion and is valued at over $8 billion. Tempus has built a massive library of clinical and molecular data, and they've used that data to develop a number of AI-powered diagnostic tests. They've worked closely with the FDA from the very beginning, and they've been able to get their tests approved in record time. As a result, they have a significant head start on their competitors, and they're well on their way to becoming the dominant player in the precision medicine market.
4. “I’ll Just Sell to Pharma” Is Not a Strategy
Every biotech founder's dream is to sell their technology to a big pharmaceutical company. While that can be a lucrative exit, it's not a go-to-market strategy. The sales cycles in pharma are notoriously long, and the decision-making process is complex. You can't just build a cool piece of technology and expect pharma companies to line up at your door. You need a nuanced strategy that involves building relationships, running clinical trials, and demonstrating real-world value. One of my investments, a drug discovery platform, spent two years and millions of dollars trying to sell to pharma with no success. They only started to gain traction when they shifted their focus to smaller biotech companies and academic research institutions. This allowed them to build a user base, gather feedback, and refine their product. Eventually, the big pharma companies came calling, but it was on their terms.
The problem with selling to pharma is that you're dealing with a massive, bureaucratic organization. There are multiple stakeholders, each with their own agenda. The person you're selling to may not be the person who ultimately makes the decision. And even if you do get a deal done, it can take months, if not years, to get it approved. In the meantime, you're burning through cash and your competitors are catching up. It's a risky proposition, and it's one that I generally advise my portfolio companies to avoid, at least in the early days.
Instead of trying to sell to pharma, focus on building a product that people want to use. Sell to smaller companies, academic institutions, and even individual researchers. Get your product into the hands of as many people as possible. Build a community around your product. If you can do that, the pharma companies will come to you. They'll see the value in what you've built, and they'll be willing to pay a premium for it. That's a much better position to be in than begging for a meeting with a mid-level executive at a big pharma company.
5. The Human-in-the-Loop Is Your Secret Weapon
There's a lot of talk about AI replacing doctors, but that's not going to happen anytime soon. The most successful AI companies in healthcare are the ones that build systems that augment, not replace, clinicians. A human-in-the-loop approach, where the AI provides insights and recommendations that a doctor can then use to make a final decision, is the most effective way to build trust and drive adoption. It also makes your system more robust and less prone to errors. I’m an investor in a mental health startup that uses AI to analyze therapy sessions and provide feedback to therapists. The AI doesn't try to replace the therapist; it acts as a co-pilot, helping them to be more effective. This approach has been incredibly successful, and the company is now one of the leaders in its field.
The beauty of the human-in-the-loop approach is that it combines the best of both worlds: the analytical power of AI and the clinical expertise of a human. The AI can process vast amounts of data and identify patterns that a human might miss. The human can then use that information to make a more informed decision. It's a symbiotic relationship that benefits everyone. The patient gets a more accurate diagnosis, the doctor is more efficient, and the healthcare system saves money.
Another benefit of the human-in-the-loop approach is that it helps to build trust with clinicians. Doctors are understandably skeptical of AI. They've been trained to rely on their own judgment and experience. They're not going to hand over the reins to a machine overnight. But if you can show them that your AI is a tool that can help them to be better at their jobs, they'll be much more likely to adopt it. And once you have their trust, you have a powerful ally who can help you to evangelize your product to their colleagues.
In conclusion, building a defensible AI moat in biotech is not about having the best model. It's about having the best data, the best workflow, a smart regulatory strategy, a realistic go-to-market plan, and a deep understanding of how to work with, not against, clinicians. If you can master these five unspoken rules, you'll be well on your way to building a company that can withstand the test of time. It's a long and difficult journey, but it's one that is well worth taking. The future of healthcare depends on it.
Frequently Asked Questions
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.
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.
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.