What an AI business plan should include (and what to throw out)

AI can draft a business plan in a minute. Whether that plan is worth executing depends entirely on what's in it — and most aren't, because they're twenty pages of confident filler wrapped around zero decisions. Here's what a usable plan contains, what to delete on sight, and the one test that separates the two.

The six sections that actually matter

1. Positioning — one decision, one sentence

Who you serve, what they get, and why you specifically — sharp enough that the wrong customers rule themselves out. “Affordable coaching for everyone” is not positioning; “a fixed-price career-decision sprint for stuck mid-career professionals, run by an HR insider” is. If the positioning section could describe three other businesses, it describes none.

2. A specific customer, not a demographic

Not “millennials interested in wellness.” A person: what problem they have, what they've already tried, what they'd pay, and what objection they'll raise first. The test is whether you could roleplay this customer in a conversation. If the plan can't, neither can you — and you'll find out in your first sales conversation.

3. Pricing — with the revenue math attached

A price on its own is a guess. A usable plan shows the arithmetic behind it: price × customers per month = your income goal, and then says how many hours those customers cost you to serve. If a plan names a monthly revenue target without showing the units — how many sales, at what price, taking how long — the target is decoration.

4. A validation test you can run this week

The cheapest experiment that could prove the plan wrong: a concrete offer, put in front of a small number of right-fit people, with a real action to count. A plan without a kill-test is a plan that assumes it's correct — which is exactly the assumption that needs testing. (We've written up the 48-hour version separately.)

5. A 90-day plan, in weeks

What happens in week one, what has to be true by day 30, what you stop doing if the numbers don't show up. Ninety days is far enough to mean something and near enough to be honest about. Anything labeled “Year 3” in a plan for a business with zero customers is fiction with a heading.

6. Risks, named plainly

What's most likely to kill this — seasonality, one acquisition channel, a contract clause with your employer, your own available hours — and what you'd do about each. A plan that contains no bad news hasn't been finished.

What AI plans typically get wrong

• Generic filler. “Leverage social media to build brand awareness.” Advice that applies to every business helps no business. If a sentence would survive being pasted into a stranger's plan, cut it.
• Fantasy market-size slides. “The global wellness market is worth $5 trillion” is trivia. Your first year will be served to dozens of people, not a market. TAM numbers in a solo-business plan exist to make the document feel important.
• Five-year projections. Invented numbers with two decimal places. Nobody can see past the next quarter of a business that doesn't exist yet; delete everything beyond 12 months and demand more detail on the next 90 days instead.
• Strategy words without decisions. “Premium yet accessible,” “omnichannel approach,” “community-driven growth.” A real plan says no to things. If nothing was excluded, nothing was decided.

The test: is it specific to your constraints?

One question sorts usable AI plans from generated filler: could this plan belong to someone else? Swap your name for a stranger's. If the plan still reads fine, it isn't a plan — it's a template.

Concretely, check: Does the schedule fit your real week, or assume full-time hours you don't have? Does the revenue math end at your goal, or at a round number? Do the ideas use skills you have today, or skills you'd need a year to build? Does it tell you what not to do? (This is the standard we built LAUNCH around — it generates the plan from 12 questions about your skills, hours, budget, and goal, so the output can't be generic. Judge it by the same test.)

A plan is a starting point, not a guarantee

The best plan you'll ever hold — AI-generated or handwritten — is version one. Its job is to make your first moves deliberate instead of random: a defensible price instead of a guessed one, a named customer instead of “anyone,” a test instead of a hope. Then reality votes. Customers misunderstand the offer, the channel you counted on underperforms, something you treated as a side note turns out to be the business.

That's not the plan failing; that's the plan doing its job — giving you something concrete enough to be wrong about, so you can correct course with evidence instead of vibes. Execute, measure, revise. The founders who get burned by plans aren't the ones who wrote bad ones; they're the ones who kept following a plan after the numbers said stop.

Common questions