Founding an AI startup: five lessons from a $200M exit

Executive overview

Customers don't care about AI — they care about outcomes. Joseph Lee sold his first AI company for $200M and draws a direct line between that lesson and how he built his next one.

Technology is a medium, not a value proposition. Start with the problem, not the model.

Validate before you build

• Called ~500 prospects in one vertical before landing first customers
• Pitched with a one-pager and a live demo, no product — just an algorithm and an end-goal vision
• Offered three months free; customers paid what they felt was fair
• 3 of 4 pilots converted to paying customers
• Took three years to generate meaningful revenue; five-times growth followed each year after

Find the hair-on-fire problem

• Start every customer conversation with: "What does success look like this year?"
• Compile candidate pain points, then stress-test them across new conversations
• After 10 customers, a list of five narrows to three; after 50–100, one common problem dominates
• Don't deliver output through your buggy new product — use tools the customer already trusts (e.g. Looker, their existing support platform) to prove value first

Sell yourself, not the product

• Early-stage founders can't out-build entrenched products; compete on founder commitment instead
• Solve the customer's problem with whatever is available, even before the product exists
• Once customers see value, then build the dedicated product

AI as a trend vs. AI as a solution

• Raising money on AI hype alone buys ~two years of runway — then the value gap becomes fatal
• Generative AI is one wave in a longer cycle; the trend always changes
• The durable question: what specific value does this technology unlock for this customer?

Choosing a co-founder

• Treat co-founder fit like a marriage: mismatched values compound over time
• Key alignment check: technology-first vs. customer-first mindset
• Complementary skills matter more than shared strengths — each founder should need the other