When to use it
When the idea in your head is still fuzzy and you need to compress it onto one page and see the holes. The role is founder-mentor. Result: a filled-in Lean Canvas (9 blocks) with "hypothesis, to be tested" marked wherever there's no data, plus the 3 riskiest spots.
The prompt (copy and paste)
You are a founder-mentor. Put this business idea into a Lean Canvas (one page).
IDEA: "<PASTE>". MARKET/NICHE: <briefly>. WHAT I ALREADY KNOW: <inputs>.
Fill in the blocks briefly and to the point:
1. Problem (top 3) and how it's solved today. 2. Customer segments (+ early adopters).
3. Unique value proposition. 4. Solution (top 3 features). 5. Acquisition channels.
6. Revenue streams and pricing model. 7. Cost structure. 8. Key metrics. 9. Unfair advantage.
Where there's no data, mark it "hypothesis, to be tested". At the end, the 3 riskiest spots in this plan.
Filled-in example
Idea: "a subscription service that delivers pet food on a schedule". Niche: cat and dog owners in Russia. What I know: people forget to restock food in time.
Expected AI response: 9 blocks — Problem (they forget to buy, they run to the pet shop, buying single bags is expensive); Segment (owners of 1+ pet, early adopters — busy people in big cities); UVP "the food arrives on its own, exactly when you run out"; Solution (subscription, automatic interval calculation, reminders); Channels; Revenue (markup + subscription); Costs (logistics, purchasing); Metrics (retention, average order value); Advantage (data on consumption cycles). Everywhere it's guessing — "hypothesis, to be tested". At the bottom, 3 risks (retention, logistics unit economics, purchasing).
Variations
- Compare pivots. "Produce 2 Lean Canvases for two different business models of this idea."
- Chained. Take the risky spots and run them through "Validating a business hypothesis".
- Investor précis. "Compress the canvas into a 5-sentence pitch for an investor."
Pro tips
- A Lean Canvas is a map of hypotheses, not a set of facts. The classic mistake is treating the filled-in cells as truth; the "to be tested" marker is mandatory.
- The "Problem" and "Segment" blocks are the foundation; if they're shaky, nothing else will save you. Ask the AI to press hard on exactly those.
- The AI structures things well and highlights gaps, but it can't invent your "unfair advantage" for you — that's about your real head start, not a nice-sounding phrase.