Let's get one thing straight: your friendly neighborhood therapy chatbot is a lie.
I don't mean it's a scam, exactly. But the promise that it can replace a real, human therapist? That's a dangerous fantasy. I've spent the better part of my life building and investing in technology in Silicon Valley. I’ve seen hype cycles come and go. I was on the front lines when my first company, MovieLaLa, was acquired by Gfycat, and again when RemoteTeam was bought by Gusto. I’ve been lucky enough to be an early investor in companies that are defining the future of AI, like Anthropic, OpenAI, and Scale AI. I say all this not to brag, but to give you context for what I'm about to say: I am a deep believer in the power of AI. But I’m also a realist.
And the reality is, the current obsession with AI-powered therapy bots is a massive distraction from the real, and far more powerful, applications of AI in healthcare. Most founders in this space think building a clever model is enough. They're wrong. I want to share the uncomfortable truth about what it really takes to succeed in the brutal, regulated world of healthcare AI, and it has very little to do with creating a digital shoulder to cry on.
The Seductive, Flawed Promise of AI Therapy
I get the appeal. I really do. A few years ago, a founder pitched me an AI therapy app. The UI was slick. The chatbot was responsive, empathetic, and available 24/7. It was a fraction of the cost of traditional therapy. On paper, it was a unicorn in the making. It promised to solve the three biggest barriers to mental healthcare: cost, access, and stigma. I was almost sold.
But then I started asking the hard questions. The questions that go beyond the demo. What happens when the bot’s advice leads to a tragic outcome? Who carries the liability insurance for an algorithm? How do you ensure patient data is truly private when it’s being used to train a commercial model? The founder’s answers were… unsatisfying. They were the answers of a technologist, not a healthcare provider. They talked about servers and encryption, not about duty of care and clinical validation.
Here’s the thing: these chatbots are just incredibly sophisticated pattern-matchers. They’ve ingested the entire internet’s worth of text on psychology, from academic papers to Reddit forums. They can mimic empathy with terrifying accuracy. But they don’t understand you. They can’t grasp the nuance of your life, your relationships, your history. They have data, but they have zero wisdom. I learned this lesson the hard way at one of my early startups. We were obsessed with data, gathering millions of data points on user behavior. We thought we could optimize our way to success. But our breakthrough came only when we started ignoring the spreadsheets and actually talking to our users, hearing their stories. Data tells you what is happening. Stories and human connection tell you why. A chatbot can only ever access the 'what'.
Why a Chatbot Can't Replace a Human
This isn't just a philosophical argument. There are deep, structural reasons why an AI cannot and should not replace a human therapist.
The first is the liability nightmare. In medicine, there's a clear chain of responsibility. If a doctor makes a mistake, they are accountable. It’s a system built on centuries of legal and ethical precedent. Who is accountable for an AI? The company that built it? The engineer who wrote the code? The user who followed its advice? This isn't a theoretical problem. The moment an AI therapy app is linked to a single adverse event, the entire industry will be buried in lawsuits and regulatory scrutiny that will make the GDPR look like a parking ticket. I’ve seen what it takes to get even simple diagnostic AI through the FDA. It’s a multi-year, multi-million dollar slog. The idea of getting an open-ended, conversational AI approved for unsupervised therapeutic use is, frankly, laughable.
Second is what I call the
"last mile" problem. A human therapist does more than just listen. They read your body language. They hear the hesitation in your voice. They pick up on the things you aren't saying. They build a relationship of trust over time. This is the ‘last mile’ of communication, and it’s where the real therapeutic work happens. An AI, by its very nature, is blind to this. It’s a black box communicating with another black box. It has no skin in the game. It doesn’t truly care. And in therapy, that caring is everything.
The Real Revolution: AI in the Background
So if AI therapy bots are a red herring, where is the real revolution? It’s not in the patient-facing applications. It’s in the boring, unsexy, backend systems that power healthcare. This is where I’ve focused my own investments in the space, and it’s where I see the most incredible potential.
Think about AI in radiology. I’m an investor in a company that’s using AI to read medical scans. Their system can detect tumors with a higher degree of accuracy than a human radiologist, and it can do it in seconds. This doesn’t replace the radiologist. It supercharges them. It acts as a second, or third, or fourth pair of eyes, catching things a human might miss after a long and stressful shift. It frees up the radiologist to focus on the most complex cases and to consult with other physicians. It’s not about replacing the human; it’s about augmenting them. This is a theme I keep coming back to. We saw it at RemoteTeam – our tools didn't replace managers, they just automated the tedious parts of their jobs so they could focus on actually leading their teams.
Or consider healthcare automation. The amount of paperwork and administrative overhead in the US healthcare system is staggering. It’s a soul-crushing burden on doctors and nurses, and it’s a huge driver of cost. I’ve invested in companies that are using AI to automate everything from patient scheduling to insurance claims processing. This is not glamorous work. You’re not going to see it on the cover of Wired. But it has a far greater impact on the quality and affordability of care than any therapy bot ever will. We’re talking about giving doctors back hours in their day. Hours they can spend with patients. Hours they can spend with their families. That’s a real, tangible improvement in mental health, for both the provider and the patient.
Finally, let's talk about clinical AI. This is the holy grail. Using AI to analyze massive datasets of clinical trial data, genomic data, and real-world patient data to discover new drug targets, to personalize treatment plans, and to predict disease outbreaks before they happen. This is where companies like the ones I've invested in, such as Scale AI and Hugging Face, are making a huge difference by providing the foundational infrastructure for these models. The potential here is almost limitless. But it requires deep domain expertise. It requires navigating a labyrinth of privacy regulations. And it requires a level of scientific rigor that is completely absent from the world of consumer-grade chatbots. For a deeper dive into how we're building the infrastructure for this, you can read my post on the new AI-powered economy.
My Uncomfortable Advice for Founders
So, if you’re a founder looking to break into healthcare AI, here is my advice. And I’ll be blunt, because no one was blunt with me when I was starting out, and I wish they had been.
First, forget about building a patient-facing app. The regulatory hurdles are immense, the liability is terrifying, and the clinical validation process will drain your will to live. I honestly had no idea what I was doing when I first dipped my toes into regulated industries. I thought a good product would win. It won't. Not on its own.
Instead, focus on the backend. Find a boring, repetitive, expensive problem within the existing healthcare system and solve it. Become an expert in the messy, complicated reality of medical billing, or clinical trial recruitment, or supply chain logistics. That’s where the real opportunity is. That’s where you can build a defensible, high-margin business.
Second, partner with clinicians from day one. Don’t just hire them as consultants. Make them co-founders. Give them equity. Embed them in your team. You need their expertise at the core of your company, not as an afterthought. You can’t build a product for doctors without having doctors in the room. It sounds obvious, but you’d be shocked how many founders miss this.
Third, be prepared for a long, hard slog. This is not a get-rich-quick scheme. The sales cycles are long. The regulatory approvals are slow. The incumbents are entrenched. But if you have the patience and the persistence to see it through, you can build a company that not only generates massive returns, but also makes a real difference in people’s lives. If you want to understand the kind of mindset required, I wrote about it here: how I became the #1 angel investor.
The Path Forward
Look, I’m not an AI pessimist. Far from it. I believe AI will transform healthcare in ways we can barely imagine. But it won’t happen the way the breathless headlines suggest. It won’t be about shiny chatbots and virtual nurses. It will be about the quiet, relentless optimization of the systems that underpin the entire industry.
The counterintuitive truth is that the best way for AI to improve our mental health is to get out of our heads and into the hospital. It's to focus on the systems, not the symptoms. It's to augment the capabilities of our human healthcare providers, not to replace them. The real work is hard, it's complicated, and it's anything but glamorous. But for the founders who are willing to take on that challenge, the rewards will be immeasurable. It’s not about the next cool app; it’s about building the future of medicine itself. And that’s a mission worth dedicating your life to.
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 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 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.