Howie Liu on the IC CEO, Fast and Slow Thinking Teams, and Airtable's AI-Native Refounding

Howie Liu on the IC CEO, Fast and Slow Thinking Teams, and Airtable's AI-Native Refounding

transcriptceo-operating-modeorg-designai-nativeairtablelenny

Howie Liu on the IC CEO, Fast and Slow Thinking Teams, and Airtable’s AI-Native Refounding

Howie Liu, co-founder and CEO of Airtable, on the shift back to individual-contributor product depth in the AI era, a novel org structure built around Kahneman’s fast and slow thinking, and a forcing-function thought experiment for AI-native reinvention. Recorded on Lenny’s Podcast.

Key ideas

  • IC CEO — the chief taste-maker: In the AI era, a CEO cannot maintain genuine product taste by reviewing finished output. You must “taste the soup” by participating in making it — running models via API, building prototypes, using AI tools hourly. Howie cut his standing one-on-one roster to free time for hands-on building and alpha-informed, urgency-driven meetings.

  • Fast and slow thinking teams: Howie restructured Airtable’s EPD org into two complementary groups inspired explicitly by Kahneman: a fast thinking group (AI platform) shipping jaw-dropping capabilities near-weekly, and a slow thinking group handling infrastructure-scale work like HyperDB. The two are symbiotic — fast creates top-of-funnel excitement; slow converts adoption seeds into durable deployments.

  • AI-native refounding thought experiment: “If you were literally founding a new company from scratch with the same mission, how would you execute on that mission using a fully AI-native approach? If you can’t, then you should find a buyer.” Airtable’s answer: its no-code Lego kit functions as a reliable DSL for AI agents — constrained primitives are a moat, not a limitation.

  • Role collapsing: Every EPD role needs a minimum viable baseline in the other two — T-shaped, not flat. Howie extends this to marketing, AEs, and SEs: outcome-orientation requires removing internal dependency chains. Handoffs are latency.

  • Vibes-first evals: Don’t start with evals. Evals constrain the solution space before you know what success looks like. Start with open-ended exploration (“vibes”), crystallise use cases, then converge on evals programmatically. Resonates with the Double Diamond diverge-then-converge methodology.

Overview

Howie opens by addressing the “Airtable is dead” tweet controversy — the data was incorrect “by a strong multiple,” and the viral spread illustrated how incentives on social media favour sensation over accuracy.

The main conversation covers the IC CEO shift: as AI makes product quality inseparable from model behaviour, you cannot taste product quality without running the models yourself. Howie describes his barbell approach to time — cut standing one-on-ones; replace with urgency-driven alpha-informed meetings and high-quality in-person catch-ups.

On the EPD reorg, Howie walks through Airtable’s history of org structures (feature-area → pillar-based → fast/slow split). The Kahneman framing is explicit and intentional — the team can internalise the mental model, not just the org chart. The fast group measures itself against AI-native competitors like Cursor: “how would a cursor or windsurf execute, and are we executing as fast as them?”

The role-collapsing section extends the IC CEO insight to every function. The underlying logic is outcome-orientation: if your outcome is to close deals, and you can collapse the SE and marketing dependencies, you should.

On evals, Howie adds a temporal dimension to the conventional wisdom: vibes-first in the discovery phase; evals once use cases crystallise. His personal LLM map-reduce practice — the most expensive inference user at Airtable by his own account — illustrates the CEO-as-power-user thesis: lead by example, not directive.

The lightning round covers the Three-Body Problem trilogy, The Studio (TV show), Runway (AI video), Japanese artisanal clothing (Self Edge, Studio D’Artisan, Whitesville/Toyo Manufacturing), and Paul Conti’s humility-and-gratitude framework drawn from his Huberman Lab appearance.