How to Use AI to Write a Business Plan Step by Step

The short version: AI can write you a full business plan draft in under two hours instead of the 40-plus hours it usually takes by hand, but only if you feed it real numbers section by section rather than one vague prompt. It will make your plan look professional in minutes; it will not tell you whether your idea is a good one, and that gap is where most AI-written plans fall apart in front of a bank manager or investor.

More on this here: How Much Does It Cost to Start a Business in the UK in 2026, and How to Fund It.

Why I stopped writing business plans the old way

I have written more business plans than I can count over the last twenty years, my own included. Five years ago, rebuilding my consultancy after a run of hard years, I sat down to write a fresh plan the way I always had: blank document, three coffees, a Sunday gone. It took me the better part of two days and I still hated the financials section by the end of it.

Last year I redid that same plan using ChatGPT and Claude side by side. First working draft in ninety minutes. Full plan, fact-checked and rewritten in my own voice, in an afternoon. That is not a hypothetical saving, that is what happened on my own laptop, and it is why I now tell every client to build their plan the same way.

What AI can do here (and what it cannot)

AI is brilliant at structure, phrasing, and speed. It will turn a messy paragraph of your thinking into a clean executive summary in seconds. It will draft a market analysis section that sounds like it was written by someone who has done this thirty times.

What it cannot do is know your market, your numbers, or your customers better than you do. Ask ChatGPT to write a business plan for a mobile dog grooming van in Leeds with no input from you and it will invent a plausible-sounding market size, a plausible-sounding competitor list, and plausible-sounding growth projections. All three might be wrong. Nobody checks. That is the uncomfortable bit nobody wants to say out loud: a lot of AI-written business plans get submitted for funding with made-up figures dressed up in confident language, and they get spotted, because bank underwriters and angel investors read dozens of these a month and the AI cadence is now unmistakable to them.

So the rule I work by: AI writes the words, you own the numbers.

Step 1: Decide who this plan is for

A plan for your own clarity looks different from a plan for a Start Up Loans application, which looks different again from one going to an investor who wants to see a path to a 10x return. Before you open any AI tool, write down in one sentence who reads this and what they need to believe by the end of it. This changes the tone you ask AI to write in far more than any prompt trick does.

Step 2: Build a messy brain-dump document first

Do not start in ChatGPT. Start in a blank note and dump everything you know in bullet points, no order, no polish: what you sell, who buys it, what you charge, what it costs you, who your three biggest competitors are, what your last twelve months of revenue looked like if you have any. This takes twenty minutes and it is the single thing that separates a useful AI-written plan from a generic one. Feed AI facts, not vibes.

Step 3: Draft it section by section, never all at once

Asking for "a full business plan for my cleaning company" gets you fluff. Building it section by section, feeding your own facts into each prompt, gets you something you can use. Here is the order I run through with clients, with the prompts I type.

Executive summary (do this one last, not first)

I know it sits at the top of every template, but write it after everything else exists. Prompt: "Using the sections below, write a one-page executive summary for a business plan. Keep it under 400 words, plain language, no jargon, written for a bank manager who reads twenty of these a week." Paste your finished sections underneath.

Business description and structure

Prompt: "I run a [type of business] in [location]. It is structured as a [sole trader / limited company / partnership]. Write the 'business description' section of a business plan covering what we do, our legal structure, and our mission in two short paragraphs, no corporate cliches."

Market analysis

This is where people get lazy and let AI invent statistics. Give it real numbers if you have them, and if you do not, ask it to show its working. Prompt: "Here is what I know about my market: [your bullet points]. Write a market analysis section and clearly flag any figure you have estimated rather than pulled from what I gave you." That single instruction, flag anything estimated, is the one most people skip and it is the one that keeps you honest.

Competitor analysis

Give AI three or four real competitor names and let it structure the comparison. Prompt: "My main competitors are [names]. Here is what I know about each of their pricing and offer: [notes]. Build a simple comparison table showing where I am cheaper, more expensive, or offering something different." This is also a good moment to check your branding stands out on paper the way it does on your website; if your logo and visuals still look homemade, it is worth a look at free AI logo generators without a watermark before your plan goes anywhere near an investor.

Products, services, and pricing

Prompt: "List my products/services with prices as I've given them, then write a short paragraph explaining my pricing strategy and why customers pay what they pay." Keep this factual, not aspirational.

Marketing and sales plan

This is where a lot of first drafts go wrong because AI defaults to generic advice like "we will use social media marketing." Push it harder. Prompt: "Write a marketing plan section based on these specific channels I use: [list them, e.g. WhatsApp broadcasts to existing customers, local Facebook groups, referral scheme]. Include a rough monthly budget split." If WhatsApp is part of how you sell, there is a decent roundup of WhatsApp business tools for small business automation worth skimming before you commit numbers to a plan.

Operations plan

Cover suppliers, premises, equipment, staffing, day-to-day running. Prompt: "Write the operations section covering suppliers, staffing, and how orders get fulfilled, based on these notes: [notes]."

Financial projections

Here is where you must slow down. AI can build the structure of a cash flow forecast and profit and loss projection, but it should never be inventing your revenue assumptions. Prompt: "Build a simple 12-month cash flow forecast table based on these monthly figures I am giving you: [your actual sales estimates, costs, fixed expenses]." Then check every single row against your real bank statements or invoicing records. If your invoicing and bookkeeping are still a mess of spreadsheets, sorting that out first makes this whole section faster; I have gone through the options in invoicing software for freelancers and small business compared, and having clean numbers pulled straight from software rather than guessed cuts this section from hours to about twenty minutes.

Step 4: Turn AI into a skeptical reader, not just a writer

Once you have a full first draft, do this: paste the whole thing back in and prompt "Read this as a skeptical bank underwriter who has seen hundreds of these plans and rejects most of them. List every weak assumption, every gap, and every place a competitor could undercut this plan." This single step catches more genuine problems than any writing prompt does, because you are using AI to stress-test rather than flatter you. My own first draft came back with four flagged weaknesses in the financial section, all fair, all things I had glossed over.

Step 5: Rewrite the executive summary and tone by hand

AI has a house style: em-dashes everywhere, three-item lists, "it is important to note that." A bank manager or investor who reads a lot of these spots it instantly and it quietly costs you credibility. Take the final draft and rewrite the opening page yourself, in your own sentences, using your own way of talking about your business. Ten minutes of your own voice on page one buys trust for the other nineteen pages.

The real example: a florist client in Kent

A client of mine runs a florist and events business in Kent. She needed a plan for a Start Up Loans application to fund a second van and part-time staff. We sat down, dumped her real numbers into a document (average order value, wedding season versus quiet months, supplier costs down to the pound), and ran the whole thing through ChatGPT section by section over one afternoon.

The draft came back clean, but the financial projections assumed twelve steady months of revenue, which was nonsense for a florist whose business triples around Valentine's Day and Christmas and dips hard in January. We caught it because we ran the skeptical-underwriter prompt from step four, fixed the seasonality in the forecast with her real order history, and the loan was approved a few weeks later. Without that stress-test step, the plan would have gone out with a smoothed-over forecast that any lender who has seen a florist's books before would have questioned in the first meeting.

How much time this saves, in real numbers

• Traditional handwritten plan for a small business: roughly 30 to 40 hours, spread over several weeks
• AI-assisted first draft, section by section, using your own real figures: 90 minutes to 3 hours
• Fact-checking, rewriting the summary, and stress-testing the financials: another 4 to 6 hours
• Total time from blank page to lender-ready document: one to two days rather than three to four weeks

That is not a saving that requires expensive tools either. ChatGPT's free tier and Claude's free tier both handle every step above; you only need a paid plan if you are running this for multiple clients or businesses at once. There is a wider set of prompts worth having on hand for the rest of your business admin too, not just the plan itself, in 50 ChatGPT prompts for small business owners that save an hour a day.

Where AI runs out of road

If you are raising serious money, six figures plus from angel investors or venture capital, or your business model is complex (multiple revenue streams, licensing, international expansion), a self-written AI plan is a starting point, not the finished article. This is the point where I tell clients to bring in a human who has read investor term sheets and knows what a specific fund is looking for. If that is where you are, it is worth reading through what an AI consultant for small business does at this stage, because the value there is judgment on the assumptions, not more polished sentences.

Mistakes I see constantly

• Asking for the whole plan in one prompt and accepting the first output without checking a single number
• Letting AI invent market size figures ("the UK market for artisan candles is worth £X million") without a source, then quoting that figure to a lender as fact
• Copying the AI's default structure without checking what the specific funder or investor asks for in their guidelines
• Skipping the skeptical-reader stress test because the draft already "sounds" convincing
• Leaving the AI's tone untouched on the executive summary, the one page every reader gives full attention to

Frequently asked questions

Can I use ChatGPT to write an entire business plan from one prompt?

You can, but the output will be generic and often includes invented statistics presented as fact. Section by section prompting, feeding your own real numbers into each part, produces a plan you can stand behind in front of a lender.

Will a bank or investor know if my business plan was written by AI?

Often, yes. Repeated phrasing patterns, smoothed-over seasonality in financials, and generic competitor sections are common tells that experienced lenders and investors now recognise, so rewriting the executive summary yourself and fact-checking every figure matters more than it did two years ago.

How long does it take to write a business plan with AI?

A working first draft can take 90 minutes to 3 hours if you prepare your own facts and figures beforehand and build the plan section by section, compared with 30 to 40 hours writing entirely by hand.

What is the biggest mistake people make using AI for business plans?

Trusting the financial projections without checking every assumption against real numbers, especially seasonality, which AI tends to smooth into an unrealistic flat monthly average unless you correct it with your actual order or sales history.

Primary sources

• Anthropic's own documentation
• the NIST AI Risk Management Framework