ChatGPT can help structure a business plan, challenge assumptions and turn founder notes into a readable draft. It should not invent market size, customer demand, competitor facts or financial forecasts. Start with verified business context, use separate prompts for each section, label assumptions clearly, then validate every external fact and number before the plan is shared with a lender, investor or partner.
A business plan becomes dangerous when polished language makes weak assumptions look certain.
That is the main risk of using ChatGPT for planning. The tool is very good at structure. It can turn scattered notes into sections, expose gaps and help a founder ask better questions. It can also produce confident-looking market claims, forecasts and competitor descriptions that were never verified.
The useful workflow is therefore not “write my business plan”.
It is facts first, section by section, with an evidence check at the end.
This guide restores an older IDJoy resource that previously appeared in Google under a query-string blog URL. The prompts have been rewritten for 2026 so the page remains useful instead of preserving stale figures and unsupported claims.
Before you prompt ChatGPT, collect the facts
Give the model information you actually know. Do not ask it to guess the missing pieces yet.
Prepare these notes:
- Business name and location.
- What you sell and what is included.
- Who pays for it.
- The problem the customer is trying to solve.
- Current prices or intended price range.
- Known costs.
- Current sales, enquiries or customer numbers if the business already trades.
- Competitors you have personally identified.
- Evidence you already have, such as interviews, quotes, invoices, website analytics or customer feedback.
- The purpose of the plan: internal planning, finance application, investor discussion, grant application or team alignment.
Anything you do not know should be marked unknown. That single habit makes the later prompts much safer.
Prompt 1: Build the planning brief
Use this before asking for a full document.
Act as a business-plan editor, not a source of invented facts.
I am building a business plan for a South African business.
Business facts:
- Business name: [name]
- Location: [city/province]
- Product or service: [offer]
- Target customer: [customer]
- Current stage: [idea / pre-revenue / trading / growing]
- Current price or expected price: [amount]
- Known monthly costs: [amount or list]
- Evidence of demand: [what you have]
- Competitors I have verified: [list]
- Purpose of this plan: [purpose]
First, organise these facts into a planning brief.
Then list the missing information that materially affects the plan.
Do not fill missing facts with estimates unless I explicitly ask for a labelled scenario.The output should make the gaps visible before they get buried inside paragraphs.
Prompt 2: Define the problem and offer
A business plan is easier to trust when the reader can see exactly why the business should exist.
Using only the facts in my planning brief, write a problem-and-offer section.
Explain:
1. The customer problem.
2. Who experiences it.
3. What the business sells.
4. Why the offer is a practical response to the problem.
5. What evidence currently supports the need.
6. What still needs validation.
Separate verified facts from assumptions.
Avoid claims such as "huge demand", "fast-growing market" or "industry-leading" unless evidence is supplied.If the section feels weak, that is useful feedback. The answer is usually more customer research, not stronger adjectives.
Prompt 3: Build the customer profile
Create a customer profile from the information below.
For each customer segment, show:
- Buyer or decision-maker.
- Trigger that makes them start looking.
- Main problem.
- Existing alternatives.
- Questions they need answered before buying.
- Evidence I have for each point.
- Assumptions that still need testing.
Do not invent demographics or buying behaviour that I did not provide.A customer section should help a sales or marketing decision. A fictional persona with a name, age and coffee preference usually does not.
Prompt 4: Turn competitor research into a useful comparison
Do the research first. Visit the competitors’ websites, check their public offer and save the source links.
Then use:
I have researched these competitors. Use only the notes I provide.
[Paste competitor notes and links]
Create a comparison covering:
- Target customer.
- Core offer.
- Public pricing where available.
- Proof or trust signals.
- Sales path or enquiry method.
- Clear strengths.
- Gaps or unanswered customer questions.
Do not claim a competitor is weak, expensive or low quality unless the evidence supports it.
End with three positioning options for my business that do not depend on attacking competitors.This turns competitor research into a positioning decision instead of a list of company names.
Prompt 5: Draft the route to market
Using my verified customer profile and offer, build a simple route-to-market plan.
For each channel, explain:
- Why the customer is likely to use it.
- What content or asset is needed.
- The action we want the customer to take.
- How we will measure whether the channel works.
- What assumption we are testing.
Prioritise a small number of channels. Do not recommend every social network by default.For a local service business, the right answer may be a website, Google Business Profile, referrals and targeted outreach. For another business it may be partnerships or a marketplace. The plan should follow the buyer, not a generic marketing checklist.
Prompt 6: Build the operating model
Turn these delivery notes into an operating model:
[Paste how the business sells, delivers, invoices, supports customers and uses suppliers]
Show:
1. The steps from enquiry to payment.
2. Who owns each step.
3. Tools or suppliers required.
4. Capacity constraints.
5. Risks that could interrupt delivery.
6. What needs to be documented before the business grows.
Do not invent staff, systems or supplier agreements.This section often reveals whether the plan can actually be delivered with the current team.
Prompt 7: Build financial scenarios without fake certainty
ChatGPT can help organise arithmetic and assumptions. It should not decide what your market will pay or promise a growth rate.
Build three labelled financial scenarios from the figures below: conservative, working case and stretch case.
Known inputs:
- Selling price: [amount]
- Variable cost per sale: [amount]
- Fixed monthly costs: [amount]
- Current monthly sales: [amount, if applicable]
- Current enquiries: [amount, if applicable]
- Current close rate: [amount, if verified]
For every number that is not directly supplied by me:
- mark it as an assumption,
- explain why it is needed,
- do not present it as a forecast.
Show the arithmetic clearly so I can replace assumptions later.A scenario is a decision tool. It is not evidence that the business will hit the number.
Prompt 8: Stress-test the plan
Act as a skeptical reviewer of this business plan.
Identify:
- Claims with no supporting evidence.
- Numbers that are assumptions rather than facts.
- Customer behaviour we have not validated.
- Operational bottlenecks.
- Dependencies on one customer, supplier, platform or founder.
- Legal, tax, employment or compliance questions that need professional advice.
- Sections that sound persuasive but do not actually answer the reader's question.
Rank the issues by how much they could change the decision to fund, launch or expand the business.This is one of the highest-value uses of AI in planning. Ask it to find holes rather than praise the document.
Prompt 9: Write the executive summary last
The executive summary should come after the plan is stable.
Write an executive summary from the completed plan below.
Use only information already contained in the plan.
Cover:
- What the business does.
- Who it serves.
- The problem and offer.
- Evidence of demand.
- How the business makes money.
- Current stage and traction.
- Main next milestone.
- Funding or support requested, if applicable.
Keep assumptions labelled and do not introduce new statistics or claims.If the summary needs a fact that is not in the plan, add the evidence to the plan first.
A business-plan evidence check
Before the document leaves your desk, mark each important statement as one of four things:
| Label | Meaning |
|---|---|
| Verified | You can point to a reliable source, record or direct evidence. |
| Observed | You have seen it in your own customers, enquiries or operations. |
| Assumption | It is a working belief that still needs testing. |
| Scenario | It is a deliberately modelled possibility, not a prediction. |
That simple classification prevents a common AI failure: an assumption quietly turning into a “fact” after several rounds of rewriting.
Where SWOT fits
A SWOT can help organise internal strengths and weaknesses alongside external opportunities and threats, but it should not replace customer and market evidence.
Use the restored AI SWOT analysis prompt guide after the first planning brief. It works best when the model has real business context instead of being asked to generate a generic matrix from the company name alone.
What to verify outside ChatGPT
Do not submit an AI-written plan without checking the parts that can change a real financial or legal decision.
Verify:
- Market statistics against the original source.
- Competitor information against current public pages.
- Tax and regulatory requirements with the relevant authority or qualified adviser.
- Loan, grant or investor requirements against the organisation receiving the plan.
- Financial assumptions against your actual costs, quotes and sales data.
- Customer demand through interviews, enquiries, pre-orders, sales or another observable signal.
AI can make the document easier to think through. It cannot turn missing evidence into evidence.
The better use of ChatGPT for business planning
Do not ask it to make your business sound investable.
Ask it to expose what you still need to prove.
That produces a less dramatic document, but a more useful one. The plan becomes a record of what is known, what is assumed and what the business needs to test next.