The Refounding Test
Ask how you'd rebuild AI-native from scratch — then decide whether your legacy asset helps or you should sell.
- Difficulty
- Advanced
- Time to result
- ~months to results
- Steps
- 4
- Confidence
- 92%
A decision framework for any pre-AI company facing the AI paradigm shift. Rather than bolting AI features onto an existing product, take a clean-slate view: if you were founding a new company today with the same mission, how would you execute fully AI-native? Then honestly assess whether your existing product gives you unfair building blocks — or whether the legacy asset actually makes you worse off than starting fresh. If the honest answer is that you can't do it better with what you have, sell the company and start the next incarnation.
Origin
Howie Liu's framework, articulated as the reasoning behind Airtable's decision to make its whole product AI-centric rather than sell. He frames AI as requiring a 'refounding' of every software product, distinct from the one-time desktop→mobile or on-prem→cloud shifts.
Core principles
- 01Every software product has to be refounded because AI is a paradigm shift, not a one-time predictable form-factor change.
- 02You cannot fool yourself by throwing AI onto the marketing site and adding a couple of features and calling it a day.
- 03The honest question is whether your existing building blocks are an unfair advantage or a legacy liability.
- 04If you're better off starting from scratch, the mission-respecting move is to sell and go build the next incarnation.
How to run it
- 1
Take the clean-slate view
Ask: 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? Answer without reference to your current product.
Pro tip Do this as a genuine blank-page exercise — the point is to escape the gravity of your existing roadmap.
Watch out Adding AI to the landing page and shipping a couple of AI features is not refounding; it's the trap this test exists to expose.
- 2
Inventory your building blocks
Enumerate the assets from your existing product that a from-scratch AI-native version could leverage. For Airtable these were reliable no-code primitives (real-time collaborative CRUD, view types, layout engine, automations) that an agent can assemble instead of writing everything from raw code.
Pro tip The strongest building blocks act like a domain-specific language the agent manipulates, avoiding the unreliability and context-collapse of generating an entire app from scratch.
- 3
Make the honest go/no-go call
Decide whether you're genuinely better off executing the AI-native vision with your existing pieces, or whether the legacy asset leaves you worse off than a fresh start. This requires real introspection, not motivated reasoning.
Pro tip Narrow the mission so the comparison is concrete — Airtable's is 'democratize software creation for business apps,' not consumer games.
Watch out Founders are biased to justify keeping the company. Force yourself to argue the 'you'd be worse off' case honestly.
- 4
If the answer is no, sell and restart
If you can't credibly claim your existing pieces give you a better shot, find a buyer for the company. If you truly care about the mission, go start its next incarnation unencumbered.
Watch out Refusing to sell out of ego, when a clean start would serve the mission better, wastes years and the team's momentum.
In the wild
Liu applied the test to Airtable and concluded its no-code components let an agent assemble reliable business apps far better than an agent writing every app from raw SQL/HTML/JS — where vibe-coded business apps hit bugs, security issues, and context collapse as complexity grows.
→ Airtable made its entire product AI-centric, with its conversational agent Omni as the default way to do everything, and the classic app UI reframed as an artifact the agent tool-uses. Liu states he wouldn't run the company in its current form if he didn't believe the existing pieces gave a better shot.
Common mistakes
Bolting AI on instead of refounding
Adding a chatbot sidebar and a marketing-site mention leaves the core product experience unchanged, so you capture none of the new form factors AI makes possible and lose to companies that rebuilt from the ground up.
Keeping the company for ego reasons
If the legacy asset is genuinely a liability, refusing to sell burns years pursuing a version that's structurally behind a from-scratch competitor. The mission is better served by selling and restarting.
Is it for you?
Best for
A founder/CEO of a decade-old software company deciding how aggressively to rebuild for AI, or whether to sell.
Not ideal for
Brand-new AI-native startups with no legacy product — they have nothing to refound and should just build.
From the transcript
“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…”
“you can't fool yourself and just say like, "Okay, I'm going to throw in some AI stuff on the landing on the marketing site, you…”
“if you really care about this mission, like go and start the next carnation of it”
From the episode
How we restructured Airtable’s entire org for AI
Howie Liu (co-founder and CEO)