Why every company now needs to think and operate like a lab team | Josh Woodward (Google Labs, Gemini, AI Studio)
Lenny's PodcastAI Summary
→ WHAT IT COVERS Josh Woodward, head of Google Labs, Gemini, and AI Studio, explains why every company now needs a dedicated frontier team operating like a labs unit. He covers where breakthrough product ideas originate, how to detect early product-market fit, which skills are rising in value, and how to structure an internal lab that avoids the three-to-four-year fizzle pattern. → KEY INSIGHTS - **Frontier Team Structure:** Every company needs a dedicated group whose sole job is staying at the frontier of AI capabilities—people with subscriptions to every new tool, tracking what just became possible before competitors do. Critically, this team cannot sit inside an existing business unit. It requires structural independence, ideally reporting to the CEO, or it will follow the historical pattern of labs fizzling out within three to four years. - **"Almost Possible" Tracking:** Maintain a live list of capabilities that are not yet viable but approaching viability. When a capability crosses the threshold—what Woodward describes as a phase shift from solid to liquid—immediately form a small team around it. Google Labs pairs this with 82 documented future predictions, knowing most will be wrong, but the discipline of holding a point of view accelerates pattern recognition when signals emerge. - **Early PMF Signal — Watch Eyes, Not Metrics:** During early prototype testing, the primary signal is physical: do people's eyes light up and do they lean forward? DAU, MAU, and D7 retention are irrelevant at this stage. Woodward also warns against falling in love with a specific product solution—teams should stay attached to the problem, because most successful products require three to five pivots before finding the right form. - **Kill Signal Comes From the Team First:** Leaders typically recognize when a project should be killed after the team already knows. The practical sign is when passion drains and the team has exhausted every approach without traction. Build a culture where team members can surface this openly. Woodward cites a recent Gemini feature that a PM recommended killing after early data came back flat—he publicly praised that call as the right behavior. - **Unlearning Rate Over Learning Rate:** When hiring for labs-style roles, Woodward prioritizes "unlearning rate"—how quickly someone can absorb a new approach and then discard it when evidence demands a change—over raw learning speed. He also looks for "explosive endurance," a term borrowed from Roger Federer's biography, meaning the capacity to sprint intensely across defined seasons while sustaining performance over multiple years without burning out. - **Labs Team Size and Lifecycle Stages:** Effective zero-to-one labs teams have shrunk from five-to-seven people to two-to-four people as AI reduces execution costs. Google Labs internally distinguishes three stages—zero-to-one, one-to-ten, and ten-to-one-hundred—each requiring different metrics and talent profiles. The zero-to-one stage uses eye-response as the metric; the ten-to-one-hundred stage shifts to conversion funnels and cost-per-acquisition, requiring a deliberate handoff in team composition. → NOTABLE MOMENT Woodward reveals that historically, large-company labs almost universally fail within three to four years—either fizzling out or producing innovations that get commercialized by outside startups instead. The third failure mode, bridging frontier work back to the core business without being dismissed as a gimmick factory, is where most labs collapse even when the first two elements are executed well. 💼 SPONSORS [{"name": "WorkOS", "url": "https://workos.com"}, {"name": "DX", "url": "https://getdx.com/lenny"}] 🏷️ AI Product Strategy, Labs Team Structure, Product-Market Fit, Hiring and Skills, Google Labs, Innovation Management
