How we restructured Airtable's entire org for AI | Howie Liu (co-founder and CEO)
Valued around $12 billion, with over $100 million in free cash flow. It reorganized into fast and slow thinking teams, made its agent the default way to do everything, and now has about half of product and engineering on AI. Its components act as a domain-specific language so the agent builds more reliable business apps.
Tips
- Split the org into fast thinking and slow thinking groups▶ 21:38Airtable's slow group built HyperDB, which handles multi-hundred-million record data sets, while the fast group ships AI features almost weekly.
- Ask how you would found the company today, AI native▶ 35:50Airtable concluded its no-code components give an agent a more reliable building kit than generating every app from raw code.
- Let agents assemble reliable primitives instead of generating everything from code▶ 37:40
- Start new AI features with vibes, then add evals▶ 1:04:35Airtable tested a crawler agent that compiles lists of entities by trying varied queries first, then found it worked best on people and companies with certain parameters.
- Cut standing one-on-ones and make meetings timely and insight driven▶ 16:40Half of Airtable's EPD org is now working on AI capabilities.
- Organize around missions and outcomes, not feature surface areas▶ 19:30
- Make the AI experience primary and show what the model can do▶ 28:00
- Let users try AI products themselves instead of telling them in decks▶ 30:00ChatGPT reached roughly 700 million weekly active users in under three years.
- Do not step away from the product details you love▶ 1:22:21








