Why a $100M AI Mental Health Startup Collapsed

Woebot, the AI mental health startup that raised over $100 million, shut down in 2024. It was a shock to the industry. The company was a pioneer — they’d been building AI therapy before ChatGPT existed, they published extensively, everyone knew them. Yet they’re gone.

Xuan Zhao, CEO of Flourish AI, watched this happen with clear eyes. She wasn’t surprised. “I think it made sense,” she says. The problem wasn’t the technology or the team. It was the bet they made about how to reach customers.

The Wrong Customer

Woebot bet that healthcare integration was the path to scale. Get into hospital systems, get prescribed by therapists, become part of the clinical workflow. It’s a reasonable bet — healthcare pays for mental health services, there’s legitimacy in the healthcare system, and once you’re in, you have an installed base.

But here’s the problem: healthcare systems move slowly. They have endless stakeholders. Regulations are murky. And as Xuan points out, changing healthcare from the inside is exponentially harder than building outside it.

“They were really betting on FDA and healthcare integration, but it turns out it’s really hard to change the system, especially when it comes to healthcare,” she explains. “There are so many stakeholders and as a startup, you have to innovate not just for the product, but the business model too. You have to start from somewhere you can actually start making money and finding product-market fit.”

Woebot was trying to change two things at once: the product (new AI for mental health) and the system (healthcare distribution). That’s an impossible position for a startup with finite capital and time.

The Lesson: Start Where You Can Make Money

Woebot’s mistake wasn’t building a good product. It was assuming that being good was enough. Being good at something healthcare providers don’t know how to buy from you is almost worthless.

Flourish took a different path. Start with schools and employers — institutions that actively want to improve student and employee wellbeing, can make buying decisions quickly, and understand the value proposition immediately. Then, once you have customers, revenue, and product-market fit, explore healthcare if you want.

“For Flourish, we’re not providing Flourish as therapy,” Xuan says. “We’re providing it as mental health promotion and early intervention. And we also have the social aspect of the app that helps people connect with each other. So that hospital, they want to provide better new patient experience, patient support groups.”

Notice the reframing. She’s not trying to replace a therapist. She’s offering something adjacent to what hospitals already want to provide — better patient experience, community-building, early intervention before crisis. That’s a much easier sell than “here’s an AI therapist.”

Another Failed Bet: Ash

Another AI mental health startup, Ash (by Slingshot), raised nearly $100 million. But they made a different bet — they went directly to consumers, positioning as AI therapy.

“They are going directly to consumer. I think that could be a better strategy than studying a healthcare system,” Xuan says. But she adds a caveat: “I’m guessing the healthcare system actually may not like what they do.”

The difference between Woebot and Ash is distribution channel. Woebot bet on healthcare and lost. Ash bet on consumers and is still alive. Both massive burns of capital. But at least Ash has customers.

What Actually Works

Flourish’s strategy is B2B2C: schools and employers pay for the app, employees and students get free or low-cost access. This solves multiple problems at once.

The buyer (school or employer) is motivated by real outcomes: reduced anxiety, better engagement, lower turnover. They can make purchasing decisions quickly and they understand ROI. The consumer gets access to high-quality software without having to pay out of pocket. And Flourish gets stable revenue from institutions with ongoing budgets for mental health and wellness.

“We figured out that school is willing to pay,” Xuan explains. “School doesn’t want to see their students struggling or even reaching a crisis point. So we figured that schools are willing to pay.”

But she’s been noticing something. “Recently we’ve been realizing that there is a very big consumer market who are not in school but who has been really benefiting from Flourish. So we are exploring how to grow faster in the consumer space.”

This is the ideal position. You start where you have clear customers and revenue. Then you expand into adjacent markets (consumers) once you’ve proven the product and have the capital to invest in that channel.

The Broader Pattern

Three companies: Woebot bet on healthcare and folded. Ash bet on consumers and survives. Flourish bet on schools first, then employees, now expanding to consumers. The success isn’t about being the smartest or the best-funded. It’s about finding a customer who will actually buy.

“Technology and product doesn’t need to innovate business model,” Xuan says. “You have to start from somewhere you can actually start making money and finding product-market fit.”

This is the unsexy part of startups that founders don’t want to hear. You could have the best product in the world. If the customer can’t buy it, doesn’t have budget authority to buy it, or doesn’t understand why they should buy it, you lose.

Woebot built a phenomenal product. They just built it for a customer that couldn’t move at the speed of a startup. That’s not a product failure. It’s a distribution failure. And in startups, distribution failure is fatal.

FAQ

Why did Woebot fail if they were so well-funded?

They bet on healthcare integration but healthcare systems move slowly and have complex stakeholder approval processes. A startup can’t survive that timeline. Woebot ran out of capital before they could establish significant healthcare partnerships. The problem wasn’t the product — it was the customer choice.

Is healthcare integration impossible for AI startups?

Not impossible, but it shouldn’t be your first market. Prove the product with faster-moving customers (schools, employers, consumers) first. Once you have revenue, capital, and proof of efficacy, then pursue healthcare partnerships. Woebot tried to do it backwards.

What’s the difference between Woebot’s strategy and Flourish’s?

Woebot went direct-to-healthcare (slow stakeholders). Ash went direct-to-consumer (slow growth, high CAC). Flourish went B2B2C: schools pay upfront, students get access free, Flourish gets revenue from a buyer motivated by outcomes. It’s the fastest path to revenue and product-market fit.

Can an AI mental health startup succeed by selling to consumers directly?

Yes, but it’s harder. Consumer mental health apps have high user acquisition costs and low retention. Ash is doing it with massive funding and seems to be surviving, but their path is capital-intensive. B2B2C is more efficient because the institution is the buyer, not thousands of individual users.

Why are schools and employers better customers than healthcare systems?

Schools and employers have budget authority and make faster decisions. A school principal or HR director can approve a $100K expense. A hospital system has to go through purchasing, compliance, clinical review, etc. Same product, vastly different sales cycle. B2B2C compresses that cycle from years to months.

If Flourish is so successful, why expand to consumers at all?

They found significant consumer demand. Schools and employers are the foundation revenue stream, but there’s also an organic consumer market discovering the app. Expanding there increases TAM without cannibalizing institutional revenue — they’re different customer types with different buying power.

What should other AI mental health startups do differently?

Choose your first customer carefully. Can they move fast? Do they have budget authority? Do they understand the value? If the answer is no to any of these, pick a different customer. Don’t build a world-class product only to fail at distribution.

Could Woebot have succeeded if they’d started with schools instead of healthcare?

Probably yes. Same product, different first customer, likely different outcome. They’d have proven efficacy, established revenue, and built capital before attempting the harder healthcare play. The core lesson is that product quality is just one variable. Distribution strategy is often the decisive one.

Is AI mental health a viable business at all?

Yes, but only if you solve the distribution problem first. Direct-to-consumer (Ash) works with massive funding. B2B2C (Flourish) works faster with less capital. Direct-to-healthcare (Woebot) doesn’t work as a first market. The business model determines success more than the technology does.

What’s the timeline difference between these strategies?

Healthcare: 2-3 years to first meaningful contract. Consumer: 12-24 months to product-market fit. Schools/employers: 3-6 months to first customers. Startups typically have 24-36 months of runway. Do the math on which customer makes sense for your timeline.