Airtable’s CEO, Howie Liu, led a bold AI transformation after a viral tweet challenged the company’s direction. He became the main user and builder of Airtable’s AI tools, focusing on rapid product changes and cutting out layers of management. Teams worked faster, launching new AI features weekly and slashing the time to build new products. After a year, these changes helped Airtable make over $100 million in cash flow while reducing costs and outpacing competitors with smarter, faster software.
How did Airtable use AI to achieve a $100M financial turnaround?
Airtable’s CEO implemented the “IC CEO” approach, personally leading AI development and reorganizing teams for rapid innovation. By focusing on product velocity, reducing management layers, and embedding AI across workflows, Airtable boosted free cash flow to over $100 million and drastically cut costs in 12 months.
Airtable’s 2024 re-org was triggered by a single viral tweet claiming the company had flatlined. Instead of crisis management theatrics, CEO Howie Liu turned the rumor into a full-blown AI overhaul. The outcome: 12 months later Airtable was generating more than $100 million in free cash flow and Liu himself had become the platform’s single heaviest AI user on the planet.
The ‘IC CEO’ blueprint
Liu coined the term IC CEO – Individual Contributor Chief Executive. Translation: the CEO spends most days writing code, running inference, and shipping internal AI experiments. No slide decks, no status theatre, just product velocity.
| Traditional CEO week | IC CEO week (actual calendar snapshot) |
|---|---|
| 10 board prep calls | 30 inference runs on new summarizer |
| 5 all-hands decks | 1 working prototype for auto-schema builder |
| 3 exec off-sites | 2 async reviews of AI cost dashboards |
The company now tracks leadership health with the CEO Inference Meter: Liu alone burns more tokens per week than the next 50 employees combined.
Fast vs slow thinking squads
Airtable borrowed Daniel Kahneman’s dual-system metaphor to structure its AI teams:
- Fast-thinking crews ship weekly drops: an auto-generated interface builder, sentiment tagging for 2 million records, or a one-click migration wizard.
- Slow-thinking crews place long-term infra bets: vector search architecture that cut query latency by 60 %, or a privacy layer that allows regulated customers to run LLMs on-prem.
Fast teams operate under “zero-meeting weeks” each quarter. The rule: every recurring meeting is cancelled so engineers can play with unreleased models. Experiment time is logged as a core OKR.
Skills that survived the pivot
The re-org halved middle-management roles but tripled the bar for three disciplines:
- Product managers must ship a working AI prototype before drafting any PRD.
- *Designers * are measured on prompt-to-interface conversion speed – average 48 hours from text idea to live component.
- *Engineers * own inference cost per user session; the budget line appears in every sprint review.
Airtable sunsetted traditional model evaluation suites for certain surface areas. Instead, teams use a “vibes check”: if a feature demo shocks the internal Slack channel with delight, it ships.
From ‘ask AI’ to ‘deploy AI’
A recent internal audit showed:
- 73 % of Airtable bases now run at least one AI field in production
- Median build time for a new workflow dropped from 9 days to 11 hours after Omni assistant went live
- Inference cost per automated decision fell 85 % between January and October 2024
The company’s public roadmap for 2025 includes turning every Airtable account into a sandbox for agentic AI teams – no API keys required.
Competitive heat map (mid-2025)
| Platform | AI-first features launched 2024 | Airtable’s delta |
|---|---|---|
| Notion | AI autofill blocks | Airtable auto-builds entire apps |
| Smartsheet | Generative formulas | Airtable generates connected databases |
| monday.com | AI item suggestions | Airtable agents orchestrate end-to-end workflows |
Sources: Airtable AI Platform, GAP Consulting AI update June 2025
The viral tweet is now pinned in Airtable’s internal #wins channel. Next to it: a live counter showing cumulative revenue per inference dollar. The number ticks up every minute.
What triggered Airtable’s $100 million turnaround?
A single viral tweet claiming “Airtable is dead” forced the company to rethink its entire strategy. Instead of defending the status quo, Howie Liu and his team rewired the organization around AI in 90 days, unlocking $100 million in free cash flow and setting a new playbook for how non-AI-native companies can pivot fast.
How did Airtable restructure teams for AI speed?
They copied Daniel Kahneman’s Thinking, Fast and Slow concept:
- Fast-thinking teams ship bold, weekly AI features (think new agents, experimental prompts, UI hacks).
- Slow-thinking teams work on 12-month infrastructure bets (scale, security, data pipelines).
The result: product velocity doubled while long-term architecture stayed solid.
What is an “IC CEO” and why does it matter?
IC = Individual Contributor.
Liu stopped acting like a distant exec and started coding daily, becoming Airtable’s #1 inference-cost user worldwide. His rule of thumb: “Cancel every meeting for a week and just play with the product.”
Teams report 27 % faster adoption of new AI features when they see the CEO using them live in Slack.
Which skills does Airtable now hire for?
Three roles dominate 2025 job posts:
- AI PMs who can write prompts and read model cards.
- Designers who prototype with no-code AI tools (Cursor, Lovable, v0) before Figma.
- Engineers who treat evals as part of the process, not the gate – “vibes-based shipping” is officially allowed for 0-to-1 bets.
How are competitors reacting?
The no-code AI space is crowded:
- Magical and Rows pitch pure automation.
- Notion and Coda embed AI inside docs.
- Grist and Baserow go open-source.
Yet Airtable’s edge is the combination of a visual database + low-latency AI agents, all on SOC-2-grade infrastructure. In Q2 2025 surveys, 68 % of Fortune 500 innovation teams list Airtable as their “default AI app builder.”
Key takeaway: The “ask AI” era is over. The winners in 2025 will be the companies that deploy AI as an operating system, not a chatbot – and leadership that codes beats leadership that PowerPoints every time.
