What VCs Actually Look for in AI Founders

Everyone wants to find the next great AI founder. Most people are looking in the wrong places.

At HumanX 2025, investors from Strange Ventures, Theory Ventures, and Radical sat down to talk about what separates the founders they back from the ones they pass on. The conversation was refreshingly honest about what the selection process actually involves — and it's more complicated than a impressive GitHub profile.

Technical depth is table stakes. Everything else is where it gets interesting.

Tara Tan from Strange Ventures was direct: deep technical knowledge isn't optional. But the diligence process that matters most happens before a term sheet is anywhere near the table. That means bringing in potential customers early — not to validate the pitch deck, but to stress-test whether the idea actually solves something real. Plenty of technically brilliant founders have built things nobody wanted. The ones worth backing know the difference.

Tomasz Tunguz from Theory Ventures added a dimension that doesn't show up on any cap table: coachability. In a sector moving as fast as AI, a founder who can't integrate feedback or acknowledge their own blind spots is a liability. Not because investors want to control the company, but because rapid growth has a way of revealing every crack in a founder's operating system. Self-awareness isn't a soft skill — it's a survival mechanism.

The playbook is being rewritten in real time.

Tunguz was candid about what the pace of change actually demands: rapid iteration cycles, a tolerance for being wrong, and the discipline to course-correct without losing momentum. This isn't aspirational. It's the minimum viable operating mode for anyone trying to build a durable company in AI right now.

Rob Toews from Radical shared a different angle — his firm often incubates companies when they find exceptional founders or identify significant market gaps before a company exists to fill them. The prerequisite for that bet? Commercial validation. Not a survey. Not a deck. Actual engagement with industry stakeholders who would eventually become customers or partners. It's a higher bar than most early-stage investors set, and that's exactly the point.

The next great AI founder might not look like the last one.

Elisa Cohn offered what might have been the most under-appreciated insight of the session: the best AI founders aren't necessarily coming from the same talent pools they always have. Her interviewing approach deliberately surfaces the emotional and psychological dimensions of a founder's story — the parts that reveal whether someone can navigate ambiguity, rebuild after failure, and keep moving when the roadmap disappears.

She's also thinking about community as infrastructure. Connecting founders with resources and creating environments where collaboration happens organically isn't just good for the ecosystem — it's how you discover people who would otherwise be overlooked.

The honest constraints nobody wants to talk about.

Toews didn't sugarcoat the structural challenges: advanced AI development requires rare talent and expensive compute. Radical has a dedicated talent team specifically to help portfolio companies solve the first problem. The second one doesn't have an easy answer yet.

What the whole conversation pointed toward is this: the AI industry is entering a phase where the quality of founders matters more than ever, because the technology itself has become accessible enough that execution is the differentiator. Agentic AI is opening up new categories across industries, but the leaders who figure out how to deploy it effectively — not just build it — will be the ones who define what the next generation of AI companies looks like.

The talent is out there. It just takes a different kind of eye to find it.

Conversations like this one are exactly why HumanX exists. Join us April 6–9, 2026 at the Moscone Center in San Francisco for HumanX San Francisco or September 22-24 at The RAI in Amsterdam for HumanX Amsterdam.