Overview of Why every company now needs to think and operate like a lab team
In this conversation, Lenny Rachitsky speaks with Josh Woodward, head of Google Labs, the Gemini app, and AI Studio, about why fast-moving AI forces every company to adopt a “labs” mindset: stay at the frontier, run lots of experiments, move quickly on what becomes possible, and be willing to kill ideas that don’t resonate. Josh shares how Google Labs finds ideas, evaluates early product-market fit, structures teams, and balances zero-to-one exploration with scaling products like Gemini. The episode is especially useful for founders, product leaders, and anyone building in AI who wants practical guidance on experimentation, team design, and what skills matter most now.
Key Takeaways
- Every company should have a frontier team whose job is to monitor and test the latest models, tools, and emerging user behaviors.
- Great ideas rarely come from formal brainstorms or design sprints; they usually emerge from obsessive builders tinkering with new tech.
- The best early signal of product-market fit is emotional: when users’ eyes light up, lean in, and want more.
- Knowing when to kill an idea often comes from the team first, not the leader.
- AI is changing planning cycles: many teams now operate on rolling ~6-month horizons, with meaningful progress expected in 50–100 days.
- Strong teams still need specialties; “everyone is a builder” is oversold.
- The most valuable people in this era combine curiosity, fast unlearning, collaboration, and the ability to sustain intensity over time.
How Great Ideas Actually Emerge
Where ideas come from
Josh says the best ideas usually come from:
- People constantly playing with the latest models and tools
- Side projects and “weird” passion experiments
- Informal moments, not scheduled brainstorms
- A culture that encourages curiosity and obsession
What doesn’t work well
- Design sprints as the primary idea engine
- Trying to “microwave” innovation
- Forcing ideas through conventional corporate processes too early
Useful practice: track what’s “almost possible”
One of the most practical tactics discussed is keeping a list of things that are almost possible with current AI tech. When a capability crosses the line from “almost” to “now,” it may become a strong candidate for a new product or feature.
How Google Labs Evaluates Product-Market Fit
Early signal: look at people’s eyes
Josh emphasizes that in early demos, the most important metric is often human reaction:
- Do people lean in?
- Do their eyes light up?
- Are they surprised in a useful way?
What matters beyond novelty
A dazzling demo is not enough. The team also looks for:
- A real user problem
- A behavior shift that feels durable
- A willingness to pay
- A product that users “can’t live without”
Don’t fall in love with the solution
He warns against becoming attached to a specific product too early. Teams often need multiple pivots before finding the real thing.
When It’s Time to Kill an Idea
Josh says the team usually knows before the leader does.
Common signs to stop
- Passion is fading
- The team has tried everything and it still isn’t working
- User reactions are weak
- Early data and live feedback don’t support launching
Leadership lesson
The best leaders create a culture where people can comfortably say:
- “This isn’t working”
- “We should not launch this”
- “We need to pivot”
He shared an example where a Gemini feature was pulled from launch after the team observed poor user response and lack of excitement.
How Google Labs Is Structured
Labs is not just “R&D”
Josh explains that a successful labs function needs:
- Independence from existing business units
- A clear mandate
- A path to connect breakthrough ideas back into the company
The three-part model for a frontier team
- People at the frontier
They stay plugged into model releases, new capabilities, and what’s viral in the ecosystem. - An environment that lets them build
The team needs room to experiment quickly and reduce friction. - A bridge to the core company
Otherwise the lab becomes a disconnected toy factory.
Labs also challenges company processes
Google Labs is used not only for products, but for testing new ways of working across Google, such as builder career ladders and new collaboration models.
Skills That Are Becoming More Valuable
Rising in value
- Unlearning rate: How quickly someone can learn something and then discard it when evidence changes
- Explosive endurance: The ability to go very hard without burning out long-term
- Trust-building: Especially important in cross-functional and AI-assisted teams
- Collaboration with both people and agents
- Curiosity and energy: People who build constantly and want to solve real problems
Still important: specialization
Josh pushes back on the idea that “everyone is a builder now.” Instead:
- PMs, designers, engineers, strategy, ops, BD, and marketing still have distinct strengths
- The best teams preserve specialties while increasing build fluency
- For early-career PMs, vibe coding should support becoming a stronger PM, not replace the craft
How Teaming and Roles Are Changing
Smaller, more fluid teams
Labs teams are getting smaller and more modular:
- Earlier: 5–7 people
- Now: often 2–4 people on a project
PMs adapt fastest
Josh observes that PMs often adapt quickly to AI-era workflows because their role already requires:
- Simulating different stakeholder perspectives
- Coordinating across functions
- Translating ideas into execution
But every function can contribute
The best builders can come from PM, engineering, design, strategy, ops, or BD — as long as they have strong curiosity and a bias toward making things.
Planning in the AI Era
Planning horizons are shortening
Josh says Labs often plans around:
- A multi-year vision
- A rolling ~6-month roadmap
- Meaningful milestones in 50–100 day cycles
Why this matters
AI changes quickly enough that:
- Something seen two days ago can become a new project
- Companies need room to react to sudden capability shifts
- Planning must stay flexible without becoming chaotic
Rhythm matters
Labs uses “seasons” to manage intensity:
- Some periods are full sprint mode
- Others are exploration/hacking mode after major launches
- The team explicitly names the mode so expectations are clear
What’s Underhyped and Overhyped in AI
Underhyped
Principles and values
- Josh thinks the industry talks too little about what kind of future we want to build
- Products encode values, so product decisions are also worldview decisions
Overhyped
Model benchmarks
- Benchmarks matter as proxies, but most users don’t care about Elo scores or leaderboards
- The real question is whether the product creates value in someone’s life
Consumer AI Opportunities Josh Is Watching
- Entertainment: Especially immersive, AI-native entertainment
- Messaging and chat: The space may not be “solved” after all
- Personal intelligence: Tools that are personal, proactive, and powerful
- IRL experiences: Products that help people save time, money, and create meaningful real-world memories
He sees major opportunity in consumer AI, especially where it combines intelligence with personal utility and distribution.
Google Gemini and Personal AI
Josh confirms that Google is moving toward a more personal assistant experience, though the UX is still evolving.
Direction of travel
- Fewer modes and toggles
- One prompt box that just does the work
- Stronger integration with personal data where users are comfortable
- A vision of “personal, proactive, powerful” AI
He says a lot is actively being built and should ship soon.
Memorable Examples and Internal Culture
Products that sparked excitement
- NotebookLM podcast/audio summaries: A standout “aha” moment
- Google Flow / Whisk: Creative generation with a level of control that felt new
Internal awards and rituals
Josh shares several playful team awards:
- TPU Harvester: For people who optimize and reclaim compute resources
- Golden Band-Aid: For fixing annoying paper cuts in the product
- LLM Whisperer: For people who can coax surprising behavior from models
- A surprise “thank you” meeting that celebrates teammates and makes appreciation tangible
These rituals reinforce the culture: optimize, care about details, and celebrate breakthrough work.
Final Advice for Teams Building in AI
- Put people at the frontier
- Watch what is almost possible
- Build around real user needs, not just technical novelty
- Keep the team small, curious, and high-energy
- Preserve specialties while increasing builder fluency
- Use labs to improve not just products, but how the company works
- Listen closely to user feedback loops — they may become one of the next big AI unlocks
Notable Quote
“Every company needs a set of people whose job is to be at the frontier.”
Bottom Line
This episode argues that AI has made experimentation a core company capability, not a side project. Josh Woodward’s view is that winning organizations will combine frontier awareness, rapid prototyping, strong taste, and a culture that knows when to push, pivot, or stop.