Twelve ready-to-run prompts for the writing a small business repeats every week — marketing drafts, customer replies, and operations. Copy the prompt, fill the bracketed fields with your real details, and review every output before it goes anywhere.
This page is the copy-paste companion to our 12 ChatGPT prompts for small-business work guide. That guide teaches the system — how to brief the model, what to supply, and how to review. This page is the library: twelve prompts organized by the job they do, each with an illustrative example (not a measured outcome) showing the kind of before/after difference the prompt is built to produce. The examples are demonstrations of the technique, not test results.
All twelve work in ChatGPT and adapt cleanly to other chat models. Replace every bracketed field with accurate details you are allowed to use, and never send or publish AI output without the review step at the end of this page.
Marketing prompts: get found and get chosen
Marketing writing repeats: the offer description, the social post, the email, the follow-up. These four prompts handle the most repeated pieces.
1. Offer description. Turn a feature list into copy a customer can understand.
Copy-paste prompt: “Rewrite this offer for [audience] in plain language a busy reader understands in 30 seconds. Preserve these facts exactly and label anything missing as [missing]: [facts]. Keep it under 120 words. No hype, no fake urgency, no income promises.”
Illustrative example (not a measured outcome). Before: a bakery’s “Artisanal slow-fermentation sourdough process utilizing heritage grain blends” that confused takeaway customers. After running the prompt with the audience “weekday lunch buyers,” the draft read like a short, plain description of what you get, how long it takes, and what it costs. The technique is the audience constraint and the fact-preservation rule — not a guaranteed sales lift.
2. Social post drafts. Five honest hooks from one approved message.
Copy-paste prompt: “Write five short social posts about [offer or news] for [audience]. Each must open with a specific, true detail from the source below, avoid hype and fear, and end with one clear action. Use only this source: [source].”
Illustrative example (not a measured outcome). Before: a plumber posted generic “call us for great service!” updates that said nothing. After the prompt with the source “same-day slots now available Tuesdays and Thursdays,” the drafts led with the specific detail instead of the generic claim. Again: this illustrates how the prompt changes the writing — engagement results were never measured.
3. Email newsletter draft. One useful email, not a wall of announcements.
Copy-paste prompt: “Draft a short newsletter email for [audience] with exactly one topic: [topic]. Structure: one useful paragraph, one detail, one call to action. Use only [source]. Keep it under 180 words. Do not invent discounts, dates, or claims.”
Illustrative example (not a measured outcome). Before: a monthly “everything we did” email nobody finished. After the prompt, the draft covered one workshop date, one thing attendees would learn, and one signup link. The single-topic constraint is the whole trick — it demonstrates structure, not open rates.
4. Customer review reply drafts. Respond well without sounding robotic.
Copy-paste prompt: “Draft a reply to this customer review in our voice: [describe voice]. Thank them for [specific detail from the review], address the [concern or compliment] directly, and invite [next step]. Do not offer compensation or make promises outside this policy: [policy]. Review text: [review].”
Illustrative example (not a measured outcome). Before: identical “Thanks for your feedback!” replies on every review. After the prompt with a stated policy (“no public discount offers”), drafts referenced the actual detail the customer mentioned. This shows how policy constraints shape output — it is not a claim about review scores improving.
Customer reply prompts: answer fast without answering wrong
Replies are where AI saves real time — and where unchecked output does real damage. These three prompts build the safety rails in.
5. Support answer from approved notes. Consistent answers from real sources.
Copy-paste prompt: “Answer this customer question using ONLY the approved notes below. If the answer is not in the notes, say ‘this needs checking with the team’ and suggest what to ask next. Keep it under 100 words. Question: [question]. Notes: [notes].”
Illustrative example (not a measured outcome). Before: staff improvised different answers about a return window, and two customers were told different things. After the prompt with the approved notes pasted in, drafts quoted one consistent policy. The “only the notes” rule is doing the work here — this is a demonstration, not a measured support-time saving.
6. Complaint response. Calm, specific, and within policy.
Copy-paste prompt: “Draft a calm reply to this complaint. Acknowledge the specific problem in one sentence, state what we will do per this policy: [policy], and give the next step with a timeframe. Do not admit fault beyond what the policy covers and do not invent causes. Complaint: [text].”
Illustrative example (not a measured outcome). Before: defensive replies that argued with the customer. After the prompt with a 48-hour response policy, the draft led with acknowledgement and a concrete next step. A human still approves every reply — the example illustrates tone structure, not complaint-resolution rates.
7. FAQ builder. Turn real support experience into help content.
Copy-paste prompt: “Turn these verified support notes into FAQ entries. One question per entry, answers under 60 words, use only the notes, and flag any contradictions between notes: [notes].”
Illustrative example (not a measured outcome). Before: the same three questions answered by phone daily. After the prompt, the notes became a draft FAQ page the owner edited in one sitting. The output is a starting draft, not a finished help center — contradiction-flagging is the feature being demonstrated.
Operations prompts: run the business more smoothly
Behind-the-scenes writing — SOPs, meeting notes, job posts — eats owner hours. These three prompts take the first draft.
8. SOP from a messy process. Document how work actually gets done.
Copy-paste prompt: “Organize this process into numbered steps with the owner of each step, required inputs, and what to check when it fails: [describe the process as you do it today]. Mark any step where you are unsure of the current practice.”
Illustrative example (not a measured outcome). Before: the opening routine lived in the owner’s head, so cover shifts guessed. After the prompt, the dictated routine became a numbered checklist with owners. It demonstrates documentation, not a measured drop in errors.
9. Meeting notes to action list. Decisions, owners, and dates — nothing invented.
Copy-paste prompt: “From these notes, list decisions, owners, due dates, and open questions. Write ‘not assigned’ for anything missing. Do not invent commitments. Notes: [notes].”
Illustrative example (not a measured outcome). Before: meeting notes ended with “great discussion” and nothing happened. After the prompt, the notes became an owner-and-date list. The “not assigned” rule is the point — no follow-through rate was measured.
10. Job post draft. Clear posts attract the right applicants.
Copy-paste prompt: “Draft a job post for [role] at a small business. Include: what the person actually does day to day, must-have vs. nice-to-have skills, schedule and pay range [range], and how to apply. Plain language, no ‘rockstar’ or ‘fast-paced environment’ clichés. Facts: [facts].”
Illustrative example (not a measured outcome). Before: a vague “looking for help” post that drew unqualified applicants. After the prompt with a real pay range and schedule, the draft filtered readers before they applied. It illustrates clarity, not hiring outcomes.
Planning prompts: think before you spend
Two prompts for decisions — the kind where a structured draft beats a blank page.
11. Simple cost comparison. Compare options on the same facts.
Copy-paste prompt: “Compare [options] for [decision] using only these criteria and the facts I supply: [criteria and facts]. Separate what the facts show from what still needs checking. End with the three questions I should answer before deciding.”
Illustrative example (not a measured outcome). Before: choosing between two point-of-sale systems by gut feel and a sales call. After the prompt with the owner’s fee facts, the draft separated knowns from unknowns and listed three questions. It demonstrates structured comparison, not a recommendation of either system.
12. Quarterly focus plan. One page, three priorities.
Copy-paste prompt: “Based on these notes about my business, propose a one-page quarterly plan: three priorities maximum, each with one measurable check, the main risk, and what we will NOT do this quarter. Notes: [notes]. Challenge anything that looks like wishful thinking.”
Illustrative example (not a measured outcome). Before: a twelve-item wish list called a plan. After the prompt, the draft held three priorities and an explicit not-doing list. The “challenge wishful thinking” instruction is doing the work — this is an illustration, not a business result.
How to adapt these prompts to your business
A library prompt is a starting point, not a finished tool. Make each one yours in five steps:
- Start with your most repeated task. Do not adopt all twelve. Pick the writing you redo weekly — the email, the reply, the post — and adapt that one prompt first.
- Replace every bracket with your real material. Real audience descriptions, real policies, real facts. The prompt’s quality ceiling is the quality of what you paste into the brackets.
- Add one voice example. Paste a short paragraph you wrote that sounds like you, and add: “Match this tone.” One example beats ten adjectives.
- Run it three times on different inputs and keep the best output with its prompt as a pair. If the third run surprises you, the prompt needs tighter constraints.
- Save the pair where the work happens — a shared doc, a team wiki, a template file. A prompt only you can find is not a business asset.
For the broader habits that make prompts reliable — choosing one outcome, gathering approved facts, removing sensitive data, and building review checklists — see the companion prompt guide, which covers the system these twelve plug into. If you are choosing which tools deserve a place in the business at all, start with best AI tools for small business.
Review rules before anything AI-written goes out
Every prompt in this library is built with review in mind, but the rules are yours to enforce. Before any output reaches a customer, a platform, or a printer:
- Check every fact against the source you supplied — names, numbers, dates, policies, prices.
- Confirm every promise. If the draft offers a discount, a deadline, or a guarantee, make sure someone with authority actually approved it.
- Read it in your customer’s shoes. Would a reasonable reader be misled by anything the draft implies but does not say?
- Check the data boundary. Make sure nothing in the prompt or output contains passwords, payment data, or private customer records you should not have pasted.
- Keep a human as the sender. The model drafts; a named person approves. That is the whole accountability model, and it does not scale away.
When a prompt repeatedly needs the same correction, fix the prompt — add the constraint you keep applying by hand. That is how a library improves: the review notes become the next version of the prompt. And when the writing starts flowing into systems rather than one-off messages, the next step is usually connecting the steps — see our small-business AI workflow for a safe way to start.
Frequently asked questions
Yes. Every prompt here is a structured brief with facts, constraints, and a review standard, so it works in any capable chat model. The underlying technique is explained in our companion prompt guide.
Replace every bracketed field with accurate, approved details from your own business. A prompt run with your real prices, policies, and customer words will outperform a generic version every time.
No. Always check facts against your source, confirm any promise or policy mentioned, and adjust the tone. AI drafts the structure; a responsible human approves the message.
Never paste passwords, payment data, private customer records, confidential contracts, or restricted business information into an AI tool unless your policy explicitly approves that tool for that data.
The companion guide teaches the prompting system and workflow habits. This page is the copy-paste library: twelve ready-to-run prompts organized by business use case, each with an illustrative example.
Assign an owner to each reusable prompt. Update source fields when offers or policies change, test on messy inputs, and retire prompts that keep producing errors.