How we restructured Airtable’s entire org for AI | Howie Liu (co-founder and CEO)

Howie Liu argues that an established software company should periodically ask what it would build if it were starting from scratch today. In his view, a company that cannot express its original mission through a fully AI-native approach may need to find a buyer and let someone else carry the mission forward. The episode frames this as a practical test for relevance rather than a claim that every legacy asset must be discarded.

The conversation presents Liu as an example of the IC CEO: a chief executive who returns to code, product building, and direct experimentation. He says this matters especially for a software platform such as Airtable, where architecture, interface, and product behavior are inseparable from the value proposition. In his account, scale had gradually pulled him away from those details, but generative AI changed the product landscape so radically that every software product must effectively be refounded. New model capabilities create new product forms and interaction patterns, so high-level delegation alone cannot determine the right experience or business model.

Liu's account of the shift is grounded in Airtable's early history. He describes writing both backend and frontend code, designing the real-time data architecture, and making product decisions directly while the company searched for product-market fit. He compares this period with Figma's early hands-on development and contrasts it with the later phase of scaling teams, processes, and go-to-market execution. The implication is that broad organizational tools can scale an existing product, but may be too blunt for a new product or a major change in what the product must become.

His proposed method is experiential. Leaders should use AI products directly, test the underlying models through APIs or chat interfaces, and push them toward their limits instead of judging only screenshots or recorded demos. He compares this to a chef learning unfamiliar ingredients: a person must work with the ingredients before knowing how to put them into a new dish. The limitation of this approach is that it demands sustained hands-on time from leaders, but Liu treats that time as part of understanding the solution space rather than as a distraction from executive work.

The interview opens with a widely shared claim that Airtable was financially failing. Liu says the claim was based on materially incorrect revenue and growth figures, and notes that the discussion gained reach after a popular podcast repeated it before later issuing a correction. His account illustrates a broader risk for companies: an inaccurate social-media narrative can become a proxy for debate about highly valued technology companies even after the underlying data is challenged. The exchange does not independently establish Airtable's financial figures; it records Liu's account of the episode and the later correction.

The hosts connect that episode to the incentive structure of social media. They observe that sensational posts are easier to share than careful factual analysis, while Liu acknowledges that such content works even on him as a reader. The point is not that bold product changes or provocative formats are inherently wrong; it is that distribution mechanics can amplify a claim before its evidence is checked.

The interview treats a CEO's personal use of AI as a visible signal of organizational commitment. Liu describes using Airtable AI repeatedly and even tracking inference cost, arguing that leaders need enough direct exposure to understand where the technology creates value. His example is deliberately extreme: he says he may spend hundreds of dollars running many LLM calls over sales transcripts because the resulting product, marketing, positioning, and segmentation insights can be strategically valuable.