A business plan with AI: inputs, calculations, and assumption checks

You need a verifiable plan, rather than a persuasive document

An AI business plan can look finished before its data exists. A model can easily connect an attractive customer description, growing orders, and profit into one story. Start with a smaller decision instead: at which volumes does our educational service cover the specified expenses, and which data is still missing for a real launch?

We will examine a fictional presentation design studio. All amounts and volumes are invented to test the logic. They are not market prices, a demand forecast, or an investment recommendation. AI is proposed here as a structural editor and calculation assistant. No real service run was performed; the arithmetic below can be reproduced independently of the chosen AI tool.

Separate facts from working assumptions

US agency SBA guidance connects company, market, product, and sales descriptions with financial calculations in a business plan. The source helps explain the document’s contents, but does not confirm demand for our studio or establish Russian accounting rules. For the educational decision, choose a short format: service, customer problem, order acquisition method, workload, expenses, and questions to verify.

One service is designing a presentation from supplied text. The educational order price is 5000 rubles; variable expenses are 1000 rubles per order. Monthly fixed payments are 12000 for workspace, 6000 for software, and 24000 for the owner’s labor. Starting requires an equipment purchase of 30000 rubles. Every figure is labeled an assumption rather than an obtained commercial quote.

Create a register of missing evidence

For every real value, save the source, date, and conditions. An advertised price is not an agreed invoice, and a customer’s intention is not a paid order. Do not let the model fill an empty row with a ‘market average’ without a verifiable source. If information is missing, keep a question and a way to obtain the answer in the register.

A complete AI assignment

Give a text assistant this assignment: ‘Prepare an educational business plan for a fictional presentation design studio. All inputs are assumptions: order price 5000 rubles; variable payments 1000 per order; monthly workspace 12000, software 6000, owner’s labor compensation 24000; one-time equipment purchase 30000; initial cash 60000. Build scenarios of 8, 12, and 20 orders per month. Calculate revenue, variable payments, total fixed payments, and a simplified balance after those payments. Show formulas, units, the point at which fixed payments are covered, and order-count rounding. Then calculate cash movement for three months with volumes of 8, 12, and 20, customers paying in the same month, and equipment purchased before the first month. Do not include equipment again in monthly payments. Taxes, refunds, payment delays, and depreciation are unspecified: list them as missing parameters and do not invent rates. Do not call the result net profit or promise demand growth. End with an assumption register and three checks that could change the decision.’

Ask the assistant to repeat the input conditions first. An error in units or a quietly added expense item will change the entire document. If it starts with ‘the project has high potential’, return it to the data: our example does not yet allow that conclusion to be checked.

Check three scenarios manually

Fixed payments total 12000 + 6000 + 24000 = 42000 rubles. Each order leaves 5000 − 1000 = 4000 rubles after the specified variable payments. Covering fixed payments requires 42000 / 4000 = 10.5 orders. Under our assumption, an order is indivisible, so the minimum whole number is eleven. This does not mean eleven orders ensure net profit: some real expenses are absent.

Check the middle scenario in two ways: 60000 − 12000 − 42000 = 6000 and 12 × 4000 − 42000 = 6000. If the answers differ, find the error first rather than choosing the more optimistic number. Naming the column matters too: ‘balance after specified payments’ honestly reflects the limited dataset.

Calculate cash movement separately

The initial 60000 rubles falls by the 30000 equipment purchase, leaving 30000 before the first month. Eight orders in month one produce minus 10000, leaving 20000. Month two with twelve orders adds 6000, producing 26000. Month three with twenty orders adds 38000, producing 64000. All customers in this version pay in the month of work: this is a separate model condition.

Now change one condition: payment for the first month arrives only in the second. Month one has no customer receipts, but has 8000 in variable and 42000 in fixed payments. Against the starting balance of 30000, this creates a 20000 shortfall. This is how a positive business description can conceal a cash gap. We are not asserting this delay will occur: it tests sensitivity to cash receipt timing.

Find errors before formatting the final text

Check three common failures. First, equipment is counted both as a startup purchase and again as a fixed monthly payment. Second, owner compensation disappears from the totals despite being in the inputs. Third, the model claims twenty orders are feasible without knowing how long one order takes. Correct arithmetic does not prove sufficient studio capacity.

To check capacity, add a separate educational hypothesis: one order requires six hours of work. Twenty orders then need one hundred and twenty hours of design alone. Sales, negotiation, revisions, and administration are excluded from those hours. Do not use this calculation as evidence of a real person’s workload. Request their actual time records and definition of what the service includes.

Turn assumptions into the next verification plan

The final document must show which decision is being discussed, which values are confirmed, and which are not. Instead of a general ‘find customers’, choose a verifiable step: discuss presentation scope and work components with a specific potential audience, record questions, and obtain terms for a possible order. Do not declare a conversation a sale. The step may produce confirmed requirements, a refusal, or no data.

Separately clarify the expenses and payment dates mandatory in your situation. Our simplified calculation excludes taxes, depreciation, refunds, and financial obligations; attractive text cannot reconstruct these parameters. Use current rules and specialist checking for a real decision. The educational table helps reveal exactly where a missing figure changes the result.

Before the next version, save the original model and assumption date. Change one substantial parameter at a time and record why. If the price changes, scenarios must be recalculated; if a new service type appears, first describe its scope and costs. The business plan then becomes a tool for checking decisions rather than a document a model once produced and nobody understands anymore.

Sources

• SBA: business plan structure and preparation

Frequently asked questions