If you type "write me a business plan for a café in a small town" into ChatGPT or Claude, you will receive twelve well-organized pages in under a minute: executive summary, market analysis, SWOT, three-year projections. The document appears complete.
In reality, most of its content is fiction.
The market sizes have no source. The competitor list is generic. The revenue projections follow a smooth upward curve, because that is what business plans in the training data look like. An experienced banker will identify the weaknesses in five minutes.
We teach a business launch program for new entrepreneurs, in which business plans play a central role, and we have tested these tools extensively in recent months. Used differently, the same tools are very effective business-plan assistants. The difference lies entirely in what you ask them to do.
Where does AI actually help in a business plan?
The first contribution is structure. Most people writing a business plan have never written one before. They do not know what a lender expects to see, in what order, or why. An AI assistant can lay out the sections, explain the purpose of each one and point out what is usually missing. That alone saves several days.
The second contribution, and the most valuable, is questions. Describe your idea and ask the assistant to interview you, one question at a time. Who is the customer? What do they pay now, and to whom? What happens in your slow months? Who keeps the books? What happens if your partner leaves?
Many of these questions are uncomfortable to answer, which is precisely what makes them useful.
Drafting comes third, and only after the first two steps. Once you have provided your real answers and your real numbers, the assistant can turn rough notes into clear paragraphs. This is an appropriate task for it. You remain the author; the assistant acts as a typist with a good command of grammar.
It is also useful for one-page tools. If you ask it to help you fill in a Business Model Canvas from your notes, you will quickly see which boxes you cannot complete. Those empty boxes define the work that remains to be done.
What is the main risk of using AI for a business plan?
The main risk is invented numbers. These models are very good at sounding certain. Ask for the size of the market for guided canoe trips in the Outaouais and you will receive a figure, perhaps with a growth rate attached, delivered with complete confidence, and sometimes with a source that does not exist.
This point is often misunderstood. The model does not look up a number. It produces text that resembles the kind of sentence that contains a number. The result is sometimes close to reality and sometimes far from it, and the tone gives no indication of which.
ChatGPT can now browse the web, which helps somewhat, but it still blends what it found with what it generated, and it does not always indicate which is which.
The rule we recommend is therefore simple: every number in the plan comes from you or from a source you have opened and read yourself. Quotes from your suppliers. Your rent. Your own count of cars in the parking lot on a Saturday. Statistics Canada, the Institut de la statistique du Québec, your regional tourism office, your municipality. If a figure cannot be traced to a real document or a real conversation, it does not belong in the plan.
AI is a good sparring partner, but it should never serve as your accountant.
How reliable are AI-generated spreadsheets?
The assistants can build a cash-flow table, and the paid version of ChatGPT can apply formulas to a file you upload. This is convenient, but every formula must be checked. In a generated spreadsheet, one month's total can silently omit an expense line, and nothing appears wrong until the figures are added up by hand. We have observed this situation first-hand.
How can AI test the strength of your business plan?
This is the use we recommend most, and the one that is least often tried.
Once your plan is written with your own numbers, paste it into the model and ask it to act as a sceptical lender, a competitor, or an investor who has seen a hundred plans like yours this year. Ask it to identify the three weakest assumptions and explain why.
The model will find issues, sometimes obvious ones that had been set aside. Seasonality is a classic example: a plan showing the same revenue in February as in July, in a region that depends on summer tourism. Another is the owner's salary set at zero for three years, with no explanation of how the owner will cover living expenses.
Then ask a second model to perform the same exercise. Different models identify different weaknesses.
Not every criticism will be relevant, and some will be generic. Nevertheless, reading an objective critique before the bank does is very valuable.
What is a step-by-step workflow for a business plan with AI?
Here is the approach we currently recommend:
- Start from a recognized template, such as the BDC business plan template, so that the structure matches what Canadian lenders expect.
- Ask the AI to interview you, section by section. Answer in your own words, even if the answers are rough.
- Collect the numbers yourself: quotes, leases, prices, and public statistics with the link noted next to each one.
- Let the AI turn your notes into clear prose, then rewrite anything that does not reflect your own voice.
- Ask two different models to critique the plan thoroughly.
- Correct the genuine weaknesses and set aside the generic comments.
- Have the plan read by a person who knows your sector.
This last step matters more than the other six. That person may be an accountant, an advisor at your local SADC or economic development office, or someone who has run a business like yours. AI does not replace that person. It allows you to arrive at the meeting better prepared, with better questions.
Why does this matter for small businesses in the regions?
In rural Quebec, many businesses start small and informally. The plan is written when a loan, a grant or a partner is involved, usually in haste during the week before a deadline. That is precisely when a polished AI-generated version is most tempting.
We advise against relying on it. A plan full of invented numbers may secure a meeting and then cost you the funding when someone checks the figures. More importantly, it can convince you. The purpose of a business plan is to think clearly about your own business before spending money. If a machine does the thinking, the most useful part of the exercise is lost.
AI is a valuable tool, provided it remains on the side of the table that asks the questions. If you would like a second perspective on the structure, the research or the financing, our business plan help service is designed for that purpose, working with your numbers and verified sources.