I’m going to tell you something that might be hard to hear. That shiny new AI platform you just spent a fortune on? It’s probably going to fail. Not because the tech is bad, but because you’re thinking about it all wrong.
I’ve seen it happen more times than I can count. A hospital system drops a million dollars on a "groundbreaking" AI diagnostics tool. A year later, it’s gathering digital dust. The doctors aren’t using it, the data is a mess, and the promised ROI is nowhere in sight. It’s a $1.2 billion mistake being repeated in hospitals all over the country. That ’s a rounding error in the grand scheme of healthcare spending, but it’s a symptom of a much bigger problem.
I’ve been in the Silicon Valley trenches for over a decade, with two exits under my belt and over 200 angel investments in companies like Anthropic, OpenAI, and Scale AI. I’ve seen firsthand how technology can change the world. But I’ve also seen how easily it can fail when it collides with the messy reality of human behavior and entrenched systems. And there’s no system more entrenched or more complex than healthcare.
The Seduction of the Algorithm
It’s easy to understand the appeal. You see a demo of an AI that can spot tumors in an MRI with 99% accuracy, and you think, “This is it. This is the future.” You’re sold a vision of efficiency, of cost savings, of better patient outcomes. And it’s not a lie, not exactly. The potential is real.
But here’s what the sales deck doesn’t tell you. It doesn’t tell you that the AI was trained on a perfect, curated dataset that looks nothing like the chaotic data streams in your hospital. It doesn’t tell you that your top radiologists will quit if you force them to use a tool that disrupts their workflow and makes them feel like a cog in a machine. It doesn’t tell you that the real work isn’t buying the AI, it’s changing the entire culture of your organization to support it.
Most founders and hospital execs think building or buying a great AI model is the finish line. They’re wrong. It’s the starting gun. And they’re running a race they’re not prepared for.
The Uncomfortable Truth About Healthcare AI
I learned this the hard way. One of my early investments was in a brilliant team out of Stanford that had developed a predictive analytics tool for sepsis. The algorithm was flawless. In trials, it could predict the onset of sepsis hours earlier than the best doctors. We were sure we had a winner. We were going to save lives and make a fortune.
We were wrong.
We got into a pilot program with a major hospital system. And it was a disaster. The doctors hated it. The alerts were firing all the time, creating a classic case of alert fatigue. The data from the hospital’s EMR was a disaster—inconsistent, incomplete, and formatted in a dozen different ways. The IT team was stretched thin and couldn’t give us the support we needed. After six months, the hospital pulled the plug.
We had the best algorithm in the world, and it didn’t matter. We had failed to understand the human context. We hadn’t spent enough time with the nurses and doctors on the front lines. We hadn’t designed the tool to fit into their world. We had tried to force the world to fit our tool.
This is the core of the $1.2 billion mistake. We fall in love with the technology and forget about the people. We focus on the algorithm and ignore the workflow. We think data is something you can just plug in, not something that needs to be cleaned, structured, and constantly maintained.
The Playbook They Don't Teach You in Business School
When we built RemoteTeam, which was later acquired by Gusto, we weren’t just building a software platform. We were building a new way for companies to operate. We had to think about everything from payroll and compliance to company culture and team communication. We succeeded because we were obsessed with the user’s entire workflow. We didn’t just throw a tool at them; we gave them a complete system.
That’s the mindset you need to bring to AI in healthcare. You’re not just installing a piece of software. You’re performing surgery on the hospital’s central nervous system. You have to be meticulous. You have to be empathetic. You have to be relentless.
Here’s what that looks like in practice:
- Shadowing: Before you write a single line of code or sign a single purchase order, your team needs to spend weeks, even months, shadowing the clinicians who will be using the tool. Eat in the hospital cafeteria. Sit in on their meetings. Understand their frustrations, their shortcuts, their moments of joy. You can’t solve a problem you don’t deeply understand.
- Co-development: Don’t build in a silo. Your clinical users should be part of the development team from day one. They should be testing prototypes, giving feedback, and helping you make the thousands of small decisions that will ultimately determine whether the tool gets used or ignored.
- Data Infrastructure: Stop thinking of data as a given. It’s not. You need a dedicated team to clean, structure, and maintain your data. This is not a one-time project; it’s an ongoing process. If you’re not willing to invest in your data infrastructure, you have no business investing in AI.
It’s Not About the Model, It’s About the Machine
Everyone gets excited about the AI model. The neural network. The deep learning algorithm. It’s the sexy part. But the model is just one small part of a much larger machine. The machine is the entire system of people, processes, and technology that delivers the result.
I’ve seen companies with mediocre models succeed because they built a great machine. And I’ve seen companies with brilliant models fail because they ignored the machine entirely. In healthcare, the machine is everything.
Think about it. A diagnostic AI is useless if the images it analyzes are low-quality. A predictive model is worthless if the data it’s fed is garbage. A clinical decision support tool is a liability if it doesn’t integrate seamlessly into the doctor’s workflow and present information in a way that is clear, concise, and actionable.
Building the machine is the hard part. It’s the unglamorous work of data cleaning, workflow mapping, and user training. It’s the political work of getting buy-in from all the different stakeholders. It’s the long, slow, frustrating work of changing a culture. But it’s the only work that matters.
Stop Buying AI, Start Building Systems
The solution isn’t to stop investing in AI. The solution is to stop thinking of it as a plug-and-play technology. It’s not a magic wand you can wave to solve all your problems. It’s a powerful, complex tool that requires a deep and sustained investment in people, processes, and infrastructure.
So, the next time a vendor comes to you with a slick demo and a promise of "transformative" results, ask them the hard questions. Ask them about their implementation process. Ask them how they’ll integrate with your existing workflow. Ask them how they’ll help you clean your data and train your people. Ask them how they’ll help you build the machine.
If they don’t have good answers, walk away. Because the most expensive AI in the world is the one that nobody uses. Don’t make the $1.2 billion mistake. Build the system, not just the algorithm. Your patients, your doctors, and your bottom line will thank you for 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.
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.