Josh Woodward, head of Google Labs, says good ideas come from tinkering by small groups of obsessed people rather than scheduled brainstorms or design sprints.
Woodward says Labs keeps 82 predictions internally and expects about nine in 10 to be wrong, yet it still watches trends well beyond the Bay Area.
Woodward warns that corporate labs often fizzle after three to four years or fail to commercialise, so he advises giving a lab independence and a clear goal.
Why every company should run experiments like Google Labs
Lenny's thesis is that fast-changing AI models force every company to experiment before competitors or startups do. Josh Woodward agrees but says the details are hard. Ideas come from small obsessed groups, most bets will fail, and a lab survives only with independence and metrics suited to its stage.
Takeaways
Good ideas come from tinkering, not scheduled brainstorms. Josh says ideas rarely come from design sprints. They appear when small groups of obsessed people tinker on weekends or over breaks. Labs keeps a list of what is almost possible and moves once a technology crosses over.
Product market fit is more art than science. Josh says technology alone is not enough and must be matched to an interesting user problem. Teams should fall in love with a problem rather than a product, since it takes three to five pivots. Early on, the test is whether eyes light up.
The team usually knows an idea is failing. Josh says the team usually knows before the leader does. A Gemini feature was dropped after early data came in weak. Labs runs 20 or 30 experiments at once.
Expect nine in ten predictions to be wrong. Labs keeps 82 predictions in an internal document, and Josh admits most will miss. It also studies future behaviour such as immersive entertainment and tries to observe trends well outside the Bay Area.
AI shrinks teams and blurs job functions. Labs project teams fell from five to seven people to two or three, maybe four. Product managers adapt fastest to other functions because the role is horizontal, though specialties still matter. Josh hires for unlearning rate and the ability to build trust.
Consumer AI is still wide open. Josh sees room in entertainment, messaging and scarce resources such as time and money. Gemini Spark connects Google data like email and calendar, and Google is moving from modes to one prompt box. He calls benchmarks over-hyped.
Running a lab inside a company
Give the lab independence rather than nesting it under an existing business unit.
Define whether its goal is graduating products into the company or inventing new lines of business.
Use it to challenge how the company works. Labs calls this creating good trouble at Google.
Match metrics to stage. Zero to one tracks eyeballs, while 10 to 100 tracks waterfalls and CPAs.
Josh warns that corporate labs often fizzle after three to four years or fail to commercialise what they make.
Premature consolidation of subteams is a mistake, since teams sometimes must live in the mess. Hiring too fast and vanity headcount metrics are others.
Tap a name to go there · tap a line to see why they are linked