Three AI chatbots. One fictional dog grooming business. Sixty minutes of my life I'm not getting back.
I wanted to know whether the big general-purpose AI models could produce a genuinely useful business plan — the kind a first-time founder could actually follow on a Monday morning. Not a document to impress a bank manager. A plan that tells you what to do, who to sell to, and where to find them.
Spoiler: none of them managed it. But they failed in interestingly different ways.
The Test Setup
I created a simple scenario. A solo founder called Sam wants to open a mobile dog grooming business in Bristol. Budget of £5,000. No existing customer base. Needs to be earning within 60 days.
I gave each model the same prompt:
"Write me a business plan for a mobile dog grooming business in Bristol, UK. I have a £5,000 budget, no existing customers, and I need to generate revenue within 60 days. Be specific and actionable."
Then I sat back and watched what happened.
ChatGPT (GPT-4o): The Overachiever That Missed the Point
ChatGPT produced 2,400 words in about 45 seconds. It looked impressive. Executive summary, market analysis, financial projections, SWOT analysis, the lot. A proper MBA-grade document.
The problem? Almost none of it was actionable.
The market analysis cited the "UK pet grooming market valued at £5.7 billion" — a number I couldn't verify anywhere. The financial projections assumed 15 clients per week from month one with zero explanation of how Sam would find those 15 people. The SWOT analysis listed "strong social media presence" as a strength for a business that doesn't exist yet.
The customer acquisition section — the bit that actually matters — was three sentences long. "Leverage social media platforms to build brand awareness. Partner with local veterinary clinics. Offer introductory discounts to generate word of mouth." That's not a strategy. That's a wish list.
ChatGPT never once mentioned which specific platforms to use for booking, what domain to register, or where in Bristol has the highest density of dog owners. It gave Sam a document to print but nothing to execute.
Gemini (1.5 Pro): The Cautious One
Gemini took a more conservative approach. Shorter output, around 1,600 words. More caveats. It kept reminding Sam to "consult a financial adviser" and "conduct thorough market research" — essentially telling him to go and do the work it was supposed to help with.
To its credit, Gemini did mention specific platforms: it suggested Rover and Bark for finding initial clients, and recommended Wix for a website. That's more practical than ChatGPT's output. It also correctly identified Clifton, Redland, and Bishopston as affluent Bristol neighbourhoods likely to have higher demand for premium dog grooming.
But the financial projections were bizarre. It estimated startup costs at £12,000 — more than double Sam's stated budget — without offering any advice on what to cut or how to launch leaner. When your customer gives you a constraint, ignoring it isn't helpful.
Customer acquisition? Same vague platitudes. "Create engaging content on Instagram." What content? How often? What hashtags? What's the conversion path from Instagram viewer to paying customer? Silence.
Claude (3.5 Sonnet): The Honest One
Claude's response was the most structured. It asked clarifying questions first — which is actually useful behaviour — then produced around 1,800 words broken into weekly action items.
I liked that it gave Sam a week-by-week timeline for the first 60 days. Week 1: register on Bark, order business cards, set up a Google Business Profile. Week 2: leaflet drop in Clifton, first free groom for a friend's dog, post before-and-after photos. That's specific. That's executable.
But it still fell into the same trap on customer acquisition. "Build a referral programme" — what does that look like? "Engage with local Facebook groups" — which ones? How? What do you say that doesn't get you banned for self-promotion?
Claude also skipped platform recommendations entirely. No mention of booking software, payment processing, or which website builder to use. Sam would still need to spend hours researching those decisions independently.
The Comparison
| Criteria | ChatGPT (GPT-4o) | Gemini 1.5 Pro | Claude 3.5 Sonnet | Mira.AI Launch Plan |
|---|---|---|---|---|
| Specificity | 3/10 | 5/10 | 6/10 | 9/10 |
| Actionability | 2/10 | 4/10 | 7/10 | 9/10 |
| Customer strategy | 1/10 | 3/10 | 4/10 | 9/10 |
| Platform recommendations | 0/10 | 5/10 | 2/10 | 9/10 |
| Time taken | 45 sec | 30 sec | 60 sec | 30 sec |
Every chatbot produced something that looked like a business plan. None of them produced something that works like one.
Why General-Purpose AI Fails at Business Planning
The core problem is obvious once you see it. ChatGPT, Gemini, and Claude are trained to generate plausible-sounding text. They're not trained to build executable action plans for specific business types in specific locations.
When you ask ChatGPT for a business plan, it draws on thousands of business plan templates it's seen during training. That's why they all look the same. Executive summary, market analysis, SWOT, financials. It's regurgitating a format, not solving a problem.
A new founder doesn't need a document. They need answers to five questions: What am I selling? Who am I selling it to? Where do I find them? What platform do I use? What do I do in the first week?
None of the three chatbots consistently answered all five.
I spent 45 minutes testing prompts across three AI models, refining, re-prompting, stitching the best bits together. Mira.AI Launch Plan answered all five questions in 30 seconds — and included a customer acquisition strategy that none of the chatbots even attempted.
Free Tool
Get your AI launch plan in 30 seconds
Tell Mira your business idea. She builds your personalised roadmap — platform, domain, customers, and a 30-day plan.
Build My Launch Plan →What I Built Instead
After running this test across dozens of business types — not just dog grooming, but cafes, consultancies, e-commerce brands, cleaning companies, personal trainers — the pattern was always the same. General AI gives you a structure. It never gives you a strategy.
Mira.AI Launch Plan is purpose-built to answer the questions that actually matter when you're starting a business. It knows which platforms work for which business types. It recommends specific tools based on your budget. It builds a customer acquisition plan with actual targets, not vague advice about "leveraging social media."
It's not trying to write a 20-page document for a bank. It's trying to get you earning money as fast as possible.
The Verdict
If you need a formal business plan for an investor deck or a bank loan, ChatGPT will give you the right structure. Add your own numbers and you'll have something presentable.
If you need to actually launch a business and find your first customers, a general-purpose chatbot won't get you there. You'll spend more time fixing its output than you would have spent just doing the research yourself.
That's not a criticism of the technology. ChatGPT is brilliant at what it's designed for. It's just not designed for this.
Frequently Asked Questions
Can ChatGPT write a good business plan?
It can write a convincing-looking business plan. The structure will be correct, the language will be professional, and it'll include all the standard sections. But the content is generic. Market sizes are often hallucinated, customer acquisition strategies are vague, and platform recommendations are non-existent. For a formal document to show an investor, it's a decent starting point. For actually launching a business, it's not enough.
Which AI is best for creating a business plan in 2026?
For a traditional business plan document, Claude 3.5 Sonnet produces the most structured and honest output. For actionable launch planning — which platforms to use, where to find customers, what to do in week one — a purpose-built tool like Mira.AI Launch Plan outperforms all general-purpose chatbots because it's designed specifically for that job.
Is an AI business plan good enough for a bank loan?
Not without significant editing. Banks want to see realistic financial projections based on actual market research, not numbers an AI generated from its training data. Use AI to create the structure and first draft, then replace every number with data you've verified yourself. The narrative sections are usually fine. The financials need human oversight.
Ready to Go Beyond the Business Plan?
Once your launch plan is built and you're ready to start finding customers, MiraReach automates the entire prospecting process — from finding ideal customers to researching their businesses to drafting personalised outreach. It's how you turn a plan into pipeline.