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

Aug 31, 2025 Episode Page ↗
Overview

Howie Liu, co-founder & CEO of Airtable, discusses reinventing his decade-old business for the AI era. He shares how CEOs are becoming ICs, restructuring into "fast" and "slow" thinking groups, and the critical skills product teams need to thrive in this rapid transition.

At a Glance
28 Insights
1h 40m Duration
12 Topics
5 Concepts

Deep Dive Analysis

Introduction to Howie Liu and Airtable's AI Transformation

Addressing the 'Airtable is Dead' Viral Tweet Controversy

The Trend of CEOs Becoming Individual Contributors (IC CEOs)

AI's Paradigm Shift and Continuous Evolution in Product Development

Airtable's Organizational Restructure: Fast-Thinking and Slow-Thinking Teams

New AI Form Factors and Airtable's Vision for App Building

Empowering Teams Through AI Tool Experimentation and Play

Developing Cross-Functional Skills for Product Managers, Engineers, Designers

When to Use 'Vibes' vs. Formal Evals for AI Product Development

Key Strategies for AI-Driven Success and Company Reinvention

Counterintuitive Startup Wisdom: Staying Close to Product Details

Advice for Aspiring Engineers and Designers in the AI Era

IC CEO

An 'IC CEO' refers to a CEO who acts as an individual contributor, getting deeply involved in coding, building, and leading initiatives directly. This approach is crucial in the AI era to stay relevant and continuously refine product-market fit by understanding product details intimately.

Fast Thinking Group

This is an organizational structure focused on rapidly shipping new, high-value AI capabilities, ideally on a near-weekly basis. The goal is to create 'jaw-dropping' value and generate top-of-funnel excitement and new use cases.

Slow Thinking Group

This group focuses on more deliberate, premeditated planning and execution, typically for foundational infrastructure or complex data systems that cannot be shipped quickly. It complements the fast-thinking group by allowing initial adoption seeds to grow into larger, more durable deployments.

Vibe Coding / App Building with AI

This concept describes the magical experience of using AI to build apps by simply describing what you want. As AI models improve, new form factors emerge, allowing users to generate complex code or full-stack applications agentically.

LLM Map Reduce

This is a capability that allows processing of large content corpuses that exceed an LLM's context window. It breaks content into chunks, performs an LLM call on each chunk, and then aggregates those results with another LLM call.

?
How should CEOs adapt to the AI era?

CEOs should become 'IC CEOs' again, getting into the code, building things, and leading initiatives themselves, as AI necessitates being deeply involved in product details to continuously refine product-market fit.

?
How can companies accelerate AI feature development?

Companies can restructure into 'fast-thinking' teams focused on shipping new, high-value AI capabilities weekly, complemented by 'slow-thinking' teams for deliberate, foundational infrastructure work.

?
What skills do product managers, engineers, and designers need in the AI era?

Individuals in these roles need to become more cross-functional, developing a baseline proficiency in all three areas (product, engineering, design) to think full-stack and operate with more autonomy.

?
How should companies approach new AI product development and evaluation?

For novel product experiences, start with 'vibes' and open-ended experimentation to discover what works, then introduce formal 'evals' once a clear use case cluster has emerged.

?
What is a key counterintuitive lesson for company building?

Avoid stepping away from the product details and the core passion that founded the company, even as it scales, to maintain holistic vision and drive non-incremental innovation.

?
How can individuals learn and adapt to the AI era?

Actively play and experiment with various AI products, build personal side projects, and leverage AI tools themselves as expert tutors to bolster skills in engineering, design, and product management.

?
Why is it important for leaders to use AI tools themselves?

Using AI tools hourly and aggressively, even if 'wasteful' in terms of inference cost, provides invaluable strategic insights and a deep understanding of what's possible, far outweighing the cost.

1. Re-found Mission AI-Native

If your existing company cannot execute its mission with a fully AI-native approach, consider selling it and starting a new, AI-native version of that mission.

2. Clean Slate AI Mission

For existing companies, take a clean-slate approach to your mission by imagining how you would execute it using a fully AI-native approach if starting from scratch today.

3. CEOs: Become ICs Again

CEOs should revert to an “ICCO” (Individual Contributor CEO) role by getting hands-on, building, and leading initiatives, especially in the AI era, to understand product details intimately.

4. Structure for Fast & Slow AI

Restructure your company into a “fast-thinking group” for rapid AI capability development and a “slow-thinking group” for deliberate, foundational work, to accelerate AI investments and ensure durable growth.

5. Embrace AI-Native Urgency

Adopt the intensity and urgency of an AI-native company, constantly asking if you are executing as fast and leveraging new capabilities as effectively as they are.

6. Aggressively Invest in AI Compute

Aggressively invest in AI compute cycles for high-value problems, even if seemingly costly, because the strategic insights gained can provide invaluable returns.

7. Break Down Role Silos

Break down role silos across all departments (e.g., marketing, sales) by encouraging individuals to become more “full stack” and outcome-oriented, reducing dependencies and increasing self-sufficiency.

8. Cultivate Hybrid Unicorn Skills

Encourage product managers, engineers, and designers to develop cross-functional skills, becoming “hybrid unicorn types” who can understand and contribute to adjacent disciplines.

9. Baseline Proficiency in All Roles

Aim to be at least “minimally good” at all three core product development roles (PM, engineering, design), even if specializing in one, to foster a multidisciplinary approach.

10. Embrace Growth Mindset in AI

Individuals across all roles (PM, engineering, design) should embrace a growth mindset and proactively learn new skills to become more effective and versatile in the AI era.

11. Stay Connected to Product Details

As a founder or leader, prioritize staying connected to the product details and the core work you love, even as responsibilities expand, to maintain passion and drive.

12. Explore AI Products & Side Projects

Continuously use a wide range of AI products (even outside your company’s offerings) and create small side projects to gain experiential understanding of new capabilities and form factors.

13. Experience New AI Releases

Stay constantly abreast of new AI product releases and capabilities by experiencing them directly, as the pace of innovation is weekly, not yearly.

14. Approach AI with Play & Curiosity

Encourage a mindset of “play” and curiosity when exploring AI tools, rather than just checking a box, to foster deeper learning and discovery.

15. Build Fun & Useful Projects

Choose personal or work projects that are both useful and enjoyable to build, as this intrinsic motivation will drive deeper learning and skill development in AI tools.

16. Build Your Own Projects

To truly learn and develop product sensibilities, actively engage in building your own projects through trial and error, rather than just observing others’ work.

17. PMs: Become Hybrid Prototypers

Product Managers should develop hybrid skills, becoming prototypers with strong design sensibilities to thrive in the AI era.

18. Prioritize Interactive AI Demos

Prioritize interactive prototypes and demos over static decks or PRDs when evaluating AI product ideas, as the real proof and feel of an AI product comes from direct interaction.

19. Embrace AI Experimentation & Iteration

Shift from deterministic, timeline-driven execution to a model of constant experimentation and iteration, especially for AI product development.

20. Start with Vibes, Not Evals

For novel AI product experiences, prioritize “vibes” (open-ended, broad testing) over formal evals initially, to discover what works and understand the range of possibilities before narrowing down.

21. Use Evals for Iterative Improvement

Once a product’s basic form factor and core use cases are established, use evals for iterative improvement and empirical testing (like A/B testing) to refine performance.

22. Ship Rapidly, Cohesively

Ship major new product capabilities on a near-weekly basis, aiming for a cohesive product experience that moves at a breakneck pace, like AI-native companies.

23. Block Time for AI Play

Encourage employees to block out dedicated time (a day or a week) to play with various AI products relevant to the company’s domain, fostering curiosity and learning.

24. Reduce Standing 1-on-1s

Reduce standing one-on-one meetings to free up time for more timely, urgency-driven, and insight-informed discussions, fostering agility and responsiveness.

25. Prioritize Urgency-Driven Meetings

Prioritize meetings that are urgency-driven and informed by timely, high-value insights (“real alpha”) to maximize their impact and avoid unproductive discussions.

26. Sell If No AI Advantage

If your existing product or business doesn’t offer a clear advantage for an AI-native approach to your mission, consider selling the company and re-founding the mission with a fresh, AI-native start.

27. Use AI as an Expert Tutor

Leverage AI tools like ChatGPT as an “expert tutor” for learning new skills (e.g., software engineering, design), asking open-ended questions to understand how to build things.

28. Cultivate Humility & Gratitude

Approach life and work with a spirit of humility and gratitude, as this mindset fosters open-mindedness, attracts opportunities, and can become a self-fulfilling prophecy for positive outcomes.

Regardless, I think that now we're entering this moment where like every, certainly every software product in my opinion has to be refounded because like AI is such a paradigm shift.

Howie Liu

I'm proud to say like I am, I'm pretty sure I'm still the, I just checked this recently, but like I take pride in being the number one most expensive in inference cost user of Airtable AI, not just within our own company, but I think for a long time I was globally across all our customers as well.

Howie Liu

If you want to cancel all your meetings for like a day or for an entire week and just go play around with every AI product that you think could be relevant to Airtable, go do it.

Howie Liu

I think for a completely novel product experience or form factor, you should actually not start with evals and you start with vibes.

Howie Liu

Don't step away from the details that both you love.

Howie Liu

Everyone can learn how to be a versatile, you know, kind of unicorn, like product engineer, designer hybrid in the AI native era. And, and like the only thing stopping you is like just going out and doing it.

Howie Liu

Howie Liu's AI Learning & Experimentation Protocol for Employees

Howie Liu
  1. Cancel all meetings for a day or an entire week to create dedicated time.
  2. Go play around with every AI product that you think could be relevant to the company.
  3. Approach experimentation with curiosity and a spirit of exploration, not just to check a box.
  4. Share findings, links, and screenshots of what you're doing and learning with others.
  5. Prioritize building actual interactive demos and prototypes over writing extensive documents or PRDs.
  6. Focus on continuous experimentation and iteration rather than rigid, deterministic timelines for execution.
700 million
ChatGPT users MAUs or weekly active users, representing 10% of humans on Earth.
under three years
ChatGPT growth period to 700 million users Timeframe for achieving massive user scale.
a dollar plus
OpenAI's deep research API cost Cost per research call, considered trivial compared to potential strategic value.