Building a great AI model is the easy part. I know that sounds crazy, but after writing checks to over 20 AI healthcare startups, I can tell you it’s true. The hard part? The really hard part is everything else. The data, the doctors, the regulations, the reimbursement. That’s where the real battles are won and lost.
I’m Sahin Boydas. I’ve been in Silicon Valley for what feels like a lifetime. I’ve built companies, sold them, and now I invest in the next generation of founders. I’ve been lucky enough to back some incredible companies like Anthropic, OpenAI, and Scale AI. But my real passion these days is at the intersection of AI and healthcare. It’s a messy, complicated space, but it’s also where I see the most potential for real, tangible impact on people’s lives.
I’ve seen a lot of pitches. A lot of brilliant founders with brilliant ideas. I’ve also seen a lot of them fail. And it’s almost never because the tech wasn’t good enough. It’s because they didn’t understand the brutal, regulated world of healthcare. I’ve got the scars to prove it. Here’s what I learned.
1. The Model is the Easy Part
I can’t tell you how many times a founder has come to me with a beautiful, elegant model that achieves 99.9% accuracy on some curated dataset. And I have to be the one to tell them: nobody cares. Seriously. In the real world, that perfect model is going to get punched in the face by messy, incomplete, and biased data. It’s going to have to integrate with ancient hospital IT systems that look like they were designed in the 90s. And it’s going to have to win over skeptical doctors who have seen a dozen “revolutionary” technologies come and go.
One of my first AI healthcare investments was a company doing AI-powered radiology diagnostics. The founders were brilliant PhDs from Stanford. Their model was a work of art. They thought they would waltz into hospitals and doctors would be lining up to use their product. They were wrong. They spent the first year just trying to get access to enough data to even start a pilot study. Then they spent another year trying to get their software to work with the hospital’s PACS system. The model was the easy part. The rest was a street fight.
2. Data is Everything (and it's a mess)
In healthcare, data is the lifeblood of any AI company. It’s also a complete and utter disaster. It’s siloed in a dozen different systems that don’t talk to each other. It’s unstructured, full of errors, and riddled with biases. And it’s protected by a fortress of privacy regulations that make it incredibly difficult to access.
I once backed a company that was trying to predict sepsis in ICU patients. The idea was simple: pull data from the electronic health record (EHR) in real-time, feed it into a model, and alert doctors when a patient was at risk. The problem? The EHR was a black box. Every hospital had a different version, with different data fields and different APIs. It took them two years and a team of engineers just to build the integrations for their first three customers. The AI part was a solved problem. The data plumbing was the real challenge.
3. Doctors are not your enemy
There’s a certain arrogance that I see in a lot of tech founders. They think they can just come into a new industry and disrupt everything. They see doctors as obstacles to be overcome, rather than partners to be won over. This is a fatal mistake in healthcare.
Doctors are the gatekeepers. They are the ones who will ultimately decide whether or not to use your product. And they are rightfully skeptical. They have been burned by technology before. They have seen too many products that promise to save them time but end up adding to their workload. You have to earn their trust. You have to show them that you understand their workflow, that you respect their expertise, and that your product will actually make their lives better.
One of the most successful companies in my portfolio has a team of doctors on staff. They are involved in every stage of the product development process, from ideation to implementation. They are the ones who go into hospitals and talk to other doctors. They speak the same language. They understand the culture. And that has made all the difference.
4. Regulation is a beast, but not a monster
The FDA. The very mention of it is enough to send a shiver down the spine of most founders. And for good reason. The regulatory process in healthcare is long, expensive, and confusing. But it’s not impossible. And it’s not a black box.
I’ve seen too many companies try to avoid the FDA. They try to position their product as a “wellness” tool or a “decision support” system, rather than a true medical device. This is a short-sighted strategy. If you want to have a real impact on patient care, you have to be willing to go through the regulatory process. You have to be willing to do the hard work of clinical validation.
One of my portfolio companies is developing an AI-powered tool for diagnosing skin cancer. They knew from day one that they would need to get FDA clearance. So they hired a team of regulatory experts. They designed their clinical trials with the FDA’s guidance in mind. It took them three years and millions of dollars, but they got it. And now they have a product that is being used by dermatologists all over the country. They have a real business, with a real moat.
5. "Workflow" is the most important word
This is the one that trips up almost every single AI healthcare startup. They build a product in a vacuum, without a deep understanding of how it will actually be used in the real world. They think that if they just build a better mousetrap, the world will beat a path to their door. It won’t.
In a hospital, every second counts. Doctors and nurses are already overwhelmed with work. They don’t have time to learn a new piece of software or change their entire workflow. Your product has to fit seamlessly into their existing way of doing things. It has to be so easy to use that it’s almost invisible. The best AI is silent.
I saw a company that had a brilliant AI for detecting diabetic retinopathy. But to use it, an ophthalmologist had to export an image from one system, upload it to another, wait for the analysis, and then manually enter the results back into the EHR. It was a disaster. They had a great model, but a terrible workflow. They went out of business in a year.
6. Reimbursement is your real product
You can have the best technology in the world, but if you can’t get paid for it, you don’t have a business. It’s as simple as that. And in healthcare, getting paid is a labyrinthine process that makes the FDA look like a walk in the park.
There are a thousand different payers, each with their own rules and their own codes. You have to figure out who is going to pay for your product, how much they are going to pay, and what you need to do to get them to pay. It’s a full-time job. And most founders are completely unprepared for it.
One of the smartest things I’ve seen a company do is hire a reimbursement consultant on day one. Before they even wrote a line of code. They figured out their reimbursement strategy before they built their product. And it paid off. They were able to get a dedicated CPT code for their technology, which made it much easier for hospitals to get reimbursed. That was the key to their success.
7. The "AI" is silent
Here’s the thing about AI in healthcare: nobody cares that it’s AI. They just care that it works. They care that it helps them make better decisions, that it saves them time, and that it improves patient outcomes. The AI should be invisible. It should be in the background, quietly doing its job.
I’m an investor in a company that uses AI to optimize hospital operations. They help hospitals reduce wait times in the emergency room, improve patient flow, and predict staffing needs. They don’t talk about AI in their marketing materials. They talk about the results. They talk about the ROI. And that’s what hospitals care about.
So, there you have it. Seven things I’ve learned from investing in over 20 AI healthcare startups. It’s a tough space, no doubt about it. But it’s also one of the most rewarding. We are on the cusp of a revolution in healthcare, and AI is the driving force. The founders who understand these lessons, who are willing to do the hard work, are the ones who are going to build the next generation of iconic companies. And I, for one, can’t wait to see what they come up with next.
Frequently Asked Questions
Which item on this list has the highest impact?
It depends on your stage and context, but in my experience, the items near the top of the list tend to have the broadest applicability. That said, sometimes the less obvious items create the biggest breakthroughs for specific situations.
How were these items selected?
Each item on this list comes from direct experience, either from building my own companies or from patterns I've observed across the 200+ startups I've invested in. I prioritize practical, actionable items over theoretical concepts.
How do I know which items apply to my situation?
Start by honestly assessing where your biggest bottleneck is right now. The items that address that specific constraint will give you the highest return on your time and energy.
Are these recommendations still relevant in 2026?
Absolutely. While specific tools and tactics change, the underlying principles remain consistent. I update my thinking regularly based on what I'm seeing in the market and across my portfolio companies.