The AI Prompt Playbook for Small Business Owners
The difference between a vague AI answer and a genuinely useful one is usually the prompt. This playbook gives you a six-part structure you can reuse for any task, plus ready-to-adapt prompts for the jobs small business owners do every week.
Key takeaways
- A strong prompt covers role, task, context, format, constraints and examples. You rarely need all six, but each one you add reduces guessing.
- Treat the first answer as a draft and refine it with short follow-up instructions.
- For bigger tasks, ask the AI to question you first before it writes anything.
- Save your best prompts in a shared library with placeholders, so the whole team gets consistent results.
AI assistants built on large language models respond to exactly what you give them. Ask for "a marketing email" and you get something generic, because the model has to fill every gap with an average guess. Give it your audience, your offer and your tone, and the output starts to sound like your business.
You do not need special tricks or secret phrases. You need a consistent way of briefing the tool, the same way you would brief a new freelancer. This playbook shows you that method and then puts it to work.
The six-part prompt structure
Think of a prompt as a short creative brief. These six elements cover almost everything an assistant needs to know:
| Element | What it tells the AI | Example phrase |
|---|---|---|
| Role | The perspective or expertise to write from | "Act as an experienced customer service lead" |
| Task | The specific job to do | "Write a reply to this complaint" |
| Context | Background facts, audience and goal | "We are a family-run bakery; the customer's cake arrived late" |
| Format | Length, structure and layout | "Under 120 words, three short paragraphs" |
| Constraints | What to avoid or must include | "Do not offer a refund; offer a 20% voucher" |
| Examples | A sample of the style or output you want | "Match the tone of this previous reply: ..." |
Context and constraints usually make the biggest difference. The model cannot know your prices, policies or customers unless you tell it.
Pro tip
Put the facts in the prompt instead of asking the AI to remember or look them up. "Our opening hours are 8am to 6pm, Monday to Saturday" is safer than "Mention our opening hours," which invites the model to make them up.
Iterate instead of starting over
Your first result is a starting point. Rather than rewriting the whole prompt, give short, specific follow-up instructions in the same conversation.
- Read the draft and name the main gap. Too long? Too formal? Missing a key point?
- Ask for one change at a time: "Cut this to 80 words" or "Make the opening less salesy."
- Point to what works: "Keep the second paragraph exactly as it is."
- Ask for options when stuck: "Give me three different subject lines for this email."
- Copy the final version into your own document and edit it by hand for the last 10 percent.
Ask the AI to ask you questions first
For tasks with many moving parts, such as a sales page or a new process document, you may not know what information to include. Let the assistant interview you.
I want you to help me write [describe the task]. Before you write anything, ask me up to 8 questions about my business, audience and goals that you need answered to do this well. Ask them one group at a time and wait for my answers.
This approach surfaces gaps early, such as your pricing, your main competitor or the action you want readers to take, and it usually produces a much stronger first draft.
Ready-to-adapt prompts for everyday tasks
Replace everything in [square brackets] with your own details. Remove names, account numbers and other sensitive data before pasting real documents into any AI tool.
1. Marketing: social media posts
Act as a social media manager for a [type of business] in [city or region]. Write 5 short posts announcing [offer or news] for [platform]. Our audience is [describe customers]. Tone: [friendly, expert, playful]. Each post under [number] words, with one clear call to action: [action]. Do not use hashtags or exaggerated claims.
2. Sales: follow-up email
Write a follow-up email to a prospect I met [where/when]. We discussed [their problem] and I proposed [your solution] at [price range]. Goal: book a 20-minute call next week. Keep it under 150 words, warm but professional, with no pressure tactics. Give me two subject line options.
3. Customer service: reply to a complaint
Act as a customer service lead. Here is a customer complaint: [paste message with personal details removed]. Our policy is [describe policy]. Write a reply that acknowledges the problem, apologizes once, explains what we will do: [your remedy], and gives a clear next step. Under 120 words. Do not promise anything beyond the remedy above.
4. Operations: turn notes into a process
Turn these rough notes into a step-by-step standard operating procedure for [task], written for a new employee on their first week: [paste notes]. Use numbered steps, flag any step where a mistake would be costly and list the tools or access needed at the top. If anything in my notes is unclear or missing, list your questions at the end instead of guessing.
5. Finance: plain-English summary
Here are my monthly figures with company names removed: [paste revenue, costs and categories]. Summarize them in plain English for a non-accountant: the three biggest changes from last month, any cost category growing faster than revenue and three questions I should ask my accountant. Show any calculations you make so I can check them.
6. Fact-checking: list claims to verify
Review the text below and list every factual claim, statistic, date, name and promise it contains. For each one, say whether it needs to be verified and what kind of source would confirm it. Do not rewrite the text. [paste text]
7. Editing: tighten a draft
Edit this draft for clarity and length. Cut it by about 30 percent, use short sentences and active voice, remove jargon and keep all facts unchanged. Then list the main changes you made so I can review them. [paste draft]
Common mistake
Trusting the fact-check prompt to do the fact-checking. AI assistants can miss errors or confidently "confirm" something false. Use prompt 6 to build your checklist, then verify each item yourself against an original source. The same applies to any arithmetic in prompt 5.
Before and after
Before: "Write a product description for our candle." The result is generic copy about "cozy ambiance" that could describe any candle.
After: "Write a 70-word product description for a hand-poured soy candle scented with fig and cedar, burn time about 40 hours, made in small batches in our workshop. Audience: gift buyers aged 30 to 50. Tone: warm and understated. Do not use the words 'cozy' or 'luxurious.'" The result now contains your real details and sounds like your brand.
Build a shared prompt library
Once a prompt works, save it. A simple shared document or spreadsheet is enough. A library stops each team member from reinventing prompts and keeps your brand voice consistent.
- Name each prompt by task, such as "Complaint reply, late delivery."
- Keep placeholders in square brackets so people know what to fill in.
- Add a short note on when to use it and who reviews the output.
- Include one good example output as a quality reference.
- Assign an owner who updates prompts when policies, prices or tools change.
A library prompt that works
- States the audience and the goal
- Includes the real facts or a slot for them
- Sets length, format and tone
- Says what to avoid
A prompt that needs work
- Is one short, vague sentence
- Asks the AI to supply facts it cannot know
- Gives no format or length
- Produces very different results each time
Your action plan
- Pick one weekly writing task and rewrite your usual prompt using the six-part structure
- Try the "ask me questions first" prompt on your next larger task
- Adapt two prompts from this playbook with your real details
- Create a shared prompt library document with names, placeholders and an owner
- Add a review step for every AI output that reaches a customer
Frequently asked questions
Do prompts work the same in every AI assistant?
The six-part structure works well across most AI assistants, because it simply gives the model clearer information. Results will still vary between tools, so test your saved prompts when you switch or when a tool is updated.
How long should a prompt be?
As long as it needs to be to remove guesswork. A quick edit may need one sentence. A customer-facing email or process document often benefits from a short paragraph of context and constraints.
Why does the AI give a different answer to the same prompt?
Large language models generate text with some built-in variation, so outputs differ between runs. Tighter constraints, examples and a fixed format make results more consistent.
Is it safe to paste business documents into a prompt?
Only if the tool is approved by your business and you have removed sensitive details such as customer names, account numbers and passwords. Check the tool's privacy settings and terms before sharing anything confidential.
Related reading
The bottom line
Good prompts are clear briefs: who, what, why, how long and what to avoid. Use the six-part structure, refine in small steps and save what works in a shared library. You will get better results in less time, and you will still be the one who checks them.