Most founders think building a great AI model is enough. They're wrong. I see the pitches every single day. Brilliant minds, beautiful slide decks, models trained on pristine datasets. They all say the same thing: our algorithm is 99.8% accurate. And I always ask the same question: so what?
I’ve spent more than a decade in Silicon Valley. I’ve built and sold two companies (you might have heard of RemoteTeam, which was acquired by Gusto. I’ve also written checks for over 200 startups, including some of the AI companies that are now household names like Anthropic and OpenAI. I’ve seen hype cycles come and go. But the current wave of AI crashing into healthcare is something else entirely. It’s a tidal wave. And it’s washing away a lot of well-funded startups who make one fatal mistake: they forget who they’re building for.
They think they’re building for a future where AI replaces doctors. They’re not. They’re building for a present where their AI has to be trusted by a skeptical, overworked, and legally exposed human being. If you can’t build that trust, your 99.8% accuracy means absolutely nothing.
The God Complex of the AI Founder
There’s a certain arrogance that comes with building cutting-edge tech. I get it. When you can manipulate billions of data points to find patterns invisible to the human eye, you start to think you have all the answers. This is a death sentence in healthcare.
I remember a team that came to me with a diagnostic AI for a rare type of cancer. Their model was technically brilliant. On their curated data, it outperformed a panel of top oncologists. They were convinced they were going to change the world. They’d raised a $10 million seed round based on that demo. But when they tried to pilot it in a real hospital, the whole thing fell apart in weeks.
Why? The model was trained on perfectly cropped, high-resolution images. Real-world hospital scans are messy. They have artifacts. Patients have multiple conditions, or comorbidities, that the model had never seen. The AI started spitting out false positives, and the doctors immediately lost faith. They saw it not as a helpful tool, but as another alarm bell they had to waste time silencing. The founders were so focused on the elegance of their algorithm that they never spent a day shadowing a radiologist to see what their actual workflow looked like. They never understood the chaos of the clinical environment. Their beautiful model was a Ferrari in a swamp.
This isn’t just about bad data. It’s about a fundamental misunderstanding of the user. Doctors are not just data-processors. Their brains are running a complex, multi-threaded analysis that involves the patient’s history, their tone of voice, their family context, and a dozen other things that don’t show up in an EMR. To build a tool they’ll trust, you have to respect that complexity, not try to bulldoze it with an algorithm.
Your Real Customer Isn't the Doctor, It's the Hospital's Lawyer
Let’s say you build a clinically useful tool. You’ve spent time in the trenches, you understand the workflow, and your AI provides real value. Congratulations. Now you get to the real boss battle: the hospital’s legal and compliance department.
Early in my career, I learned a hard lesson: the person using the product is rarely the person who buys it. With my first company, MovieLaLa, we had to sell to movie studios, not just movie fans. At RemoteTeam, we had to sell to CFOs and heads of HR, not just the employees who loved the platform. In healthcare, this is magnified by a factor of a thousand. The doctor might love your tool, but the person who signs the check is thinking about one thing above all else: liability.
What happens when your AI makes a mistake? Who gets sued? The doctor? The hospital? You? The answer is everyone. I’ve seen more than one promising AI deal die a slow death in the nine-month-long due diligence process of a hospital’s legal team. They will grill you on HIPAA compliance, on your data security protocols, on your FDA clearance strategy. Do you need a 510(k)? Are you a Class I, II, or III device? If you can’t answer these questions fluently, you’re done.
One of the smartest things I ever saw a founder do was hire a part-time Chief Medical Officer who had previously been on a hospital’s technology procurement committee. He didn’t just give them clinical insights; he gave them the playbook for navigating the internal politics and legal hurdles of their target customers. He helped them build a bulletproof security dossier and a clear liability framework. They stopped selling a product and started selling a solution to the hospital’s risk problem. That’s when they started closing deals.
The Counterintuitive Path to Trust: Start with the Dumbest Task
So how do you get started? How do you get that first foot in the door and begin the long process of building trust? It’s not by trying to solve the most complex diagnostic challenge. It’s by solving the most annoying, mundane, and time-consuming one.
Doctors spend, on average, more than 15 hours per week on paperwork and administrative tasks. It’s the part of the job they all hate. It’s low-risk, high-volume work that is a perfect entry point for an AI. Instead of building an AI to detect cancer, build an AI that automates the pre-authorization paperwork for a CT scan. Instead of predicting heart attacks, build an AI that transcribes a doctor’s spoken notes into a perfectly formatted EMR entry.
I invested in a company that did exactly this. They built an AI for mental health professionals. But they didn’t start with AI therapy bots. They started by automating the incredibly painful process of writing clinical notes for insurance reimbursement. Therapists loved it. It saved them 5-10 hours a week. The hospital loved it because it improved billing accuracy and compliance. It was a simple, unglamorous problem.
But here’s what happened. By solving that simple problem, the company got its software embedded in the therapist’s daily workflow. They became a trusted partner. The therapists started asking for more features. “Can you help me track patient progress?” “Can you suggest resources for this specific condition?” The company slowly and carefully expanded its offerings, building on that initial foundation of trust. They are now one of the leading platforms in the AI mental health space, and it all started with solving a boring administrative problem.
The Future is a Partnership, Not a Replacement
Forget the sci-fi headlines about AI doctors. It’s not going to happen. Not in our lifetime. The future of clinical AI is about partnership. It’s about building tools that augment, not replace, the skills of a human doctor. Your AI should be the world’s best resident, working alongside the attending physician. It can surface information, flag potential issues, and handle the administrative grunt work, freeing up the doctor to do the one thing a machine can’t: connect with a patient.
Building a successful healthcare AI company is brutally difficult. It requires a humility that many tech founders lack. It requires you to fall in love with the doctor’s problem, not your own solution. But if you can do that—if you can build a tool that a doctor genuinely trusts to make their job easier and their patients safer—you won’t just build a successful company. You’ll actually change medicine for the better.
I saw another team, brilliant engineers from a top university, who built an AI to predict sepsis in ICUs. On paper, it was incredible. It could flag patients hours before human doctors could spot the signs. They were sure they had a winner. But they made the same mistake. They never considered the 'alert fatigue' that doctors deal with. The ICU is a constant barrage of beeps, alarms, and flashing lights. The doctors, already overwhelmed, saw the AI's alerts as just more noise. The alerts weren't integrated into their existing monitoring systems, they just popped up on a separate tablet that someone had to remember to check. The project was a failure, not because the AI was wrong, but because the team never designed it to fit into the messy, chaotic reality of a doctor's life.
The Allure of the 'Big Problem'
Founders are drawn to the big, sexy problems. Curing cancer. Eradicating Alzheimer's. These are the goals that attract venture capital and media attention. But in healthcare, the big problems are also the highest-risk, most complex, and most heavily regulated. It's a brutal place to start. The odds are stacked against you. You're not just fighting a technical battle; you're fighting against decades of established medical practice, entrenched workflows, and a deep-seated professional skepticism.
I often tell founders to think like a wedge. You don't take over a market by attacking the fortress head-on. You find a small crack, a tiny opening, and you wedge yourself in. You solve a small, overlooked problem so well that you become indispensable. Then, from that position of trust, you expand. This is how the biggest companies in the world were built. Amazon started with books. Facebook started with a single college campus. In healthcare, the same principle applies. Start with the boring stuff. The paperwork. The scheduling. The transcription. The things that doctors hate doing. That's your wedge.
From Annoyance to Adoption
Let's get specific. What are these 'dumb' tasks? Here are a few examples I've seen work:
- Automated Insurance Claims: The process of submitting claims to insurance companies is a nightmare of different forms, codes, and requirements. An AI that can automatically generate and submit these claims is a godsend for any private practice.
- Patient Intake and Triage: A simple chatbot can handle the initial patient intake, asking basic questions, collecting medical history, and even scheduling appointments. This frees up the front-desk staff and makes the whole process more efficient.
- Medical Coding: Every diagnosis and procedure has a specific code. Getting these codes right is essential for billing and reimbursement. An AI that can accurately assign these codes based on a doctor's notes is incredibly valuable.
These aren't the kind of ideas that get you on the cover of a magazine. But they are the kind of ideas that get you paying customers. They solve a real, immediate pain point. And in doing so, they build the foundation of trust that you need to eventually tackle the bigger, more complex problems.
The Long Road
Building a great company in any industry is a marathon, not a sprint. In healthcare, it's an ultramarathon. The sales cycles are long. The regulatory hurdles are high. The resistance to change is immense. But the opportunity is equally massive. We are at the very beginning of a transformation that will reshape how healthcare is delivered.
For the founders who are willing to be humble, to listen to their users, to start small and build trust step-by-step, the rewards will be immense. You won't just be building a business. You'll be building a legacy. You'll be creating the tools that will empower the next generation of doctors and improve the lives of millions of patients. And that's a goal worth fighting for.
Frequently Asked Questions
Do I need technical skills to build a clinical ai that doctors actually trust: the counterintuitive guide?
Not necessarily. While technical understanding helps, the most important skills are clear thinking and the ability to break problems into smaller pieces. Many successful founders I've invested in started with zero technical background and either learned enough to be dangerous or found the right technical partner.
How do I measure success with this approach?
Pick one or two metrics that directly tie to your goal and track them weekly. Vanity metrics like page views or follower counts rarely matter. Focus on metrics that reflect real engagement or revenue impact.
How long does it take to build a clinical ai that doctors actually trust: the counterintuitive guide?
The timeline varies depending on your starting point and resources. For most founders, expect 2-4 weeks for initial setup and 2-3 months to see meaningful results. I've seen teams move faster when they focus on one thing at a time rather than trying to do everything at once.
What tools do I need to get started?
Start with the basics. You don't need expensive software or fancy tools. A spreadsheet, a note-taking app, and direct access to your customers will get you further than any enterprise platform. Add tools only when you hit a specific bottleneck.