ChatGPT Prompts for Small Business: How to Get Quotes, Emails and Reports Right the First Time

You ask ChatGPT to tidy up a quote, reply to a customer or summarise a meeting. The answer sounds polished, but it has guessed a date, missed an important condition or put the information in the wrong format.

The fix is usually not a clever trick. It’s giving ChatGPT a clear brief, the right facts and instructions about what to do when information is missing. Here’s a simple framework, plus prompts you can copy and adapt.

A simple five-part prompt framework

A useful prompt gives ChatGPT enough direction to produce a workable first draft. Use these five parts:

1. Role: What kind of helper should it act as? For example, “a careful small-business administrator”.
2. Context: What does it need to know about your business, customer, audience or situation?
3. Task: What do you want it to do? Use a specific verb such as draft, summarise, organise or check.
4. Format: What should the result look like? A table, short email, checklist or report?
5. Example: If you have a format or style you already like, show a short example. Remove private details first.

Then add safeguards: what it must not guess, what it should flag and what a person needs to check.

A reusable prompt starter

Act as [role]. Context: [relevant business and audience details].
Task: [what you need done].
Use only the information I provide. Do not invent names, prices, dates, terms or commitments. If something is missing or unclear, mark it “[To confirm]” and list a question.
Format the result as [format]. Use [tone and length].
Example of the format or style I want: [optional example].
Information to use: [paste relevant, de-identified details].

OpenAI’s prompting guidance also recommends being clear and specific, providing context and refining your instructions when the first answer misses the mark. Think of the first response as a draft, not a final decision.

Copy-and-paste prompts for everyday work

Replace the square-bracketed sections with your information. You don’t need to fill every field. If a detail isn’t known, leave it out and ask ChatGPT to flag it.

1. Drafting a customer quote

A quote needs more than tidy wording. It needs the correct scope, quantities, prices and conditions. Give ChatGPT the details you’ve already confirmed, and let it organise them without setting your price for you.

Act as a careful quoting assistant for a [type of business]. Draft a customer quote using only the facts below.

Include: quote number, issue date, validity period, itemised work or products, quantities, supplied prices in AUD, subtotal, GST treatment if provided, total, assumptions, exclusions, estimated timing, payment terms and next steps.

Do not invent or change any price, quantity, date, GST treatment or business term. Mark anything missing as “[To confirm]”. Return a clear customer-facing quote, followed by a short checklist of details I need to verify.

Confirmed information: [scope, quantities, prices, timing and approved terms].

Check the arithmetic and tax treatment yourself. Confirm that the scope and exclusions match what you agreed with the customer before sending.

2. Replying to a customer email

A good email prompt explains what the reply should achieve and what you’re authorised to promise. That helps keep a friendly draft from turning into an accidental discount or delivery guarantee.

Act as a customer-service assistant for [business type]. Draft a reply to the customer’s message below.

Goal of the reply: [answer a question / confirm next steps / acknowledge a problem]. Use a [warm, professional] tone and keep it under [word count] words. Include a subject line and one clear next step.

Use only the approved facts below. Do not promise a refund, discount, delivery date or other commitment unless it is explicitly approved. If a detail needs checking, flag it before the draft and use “[To confirm]” in the email.

Customer message: [remove personal details and paste relevant text].
Approved facts and policies: [details].

Read the draft against the original message. Check the recipient, names, dates, attachments and any promises before sending.

3. Preparing a staff roster

ChatGPT can help organise availability and spot gaps, but don’t ask it to infer who can work or decide whether a roster meets your obligations.

Create a draft roster for [week and location].

Staff availability: [use initials or staff IDs and list confirmed availability].
Shifts and coverage needed: [dates, times and roles].
Rules to follow: [required skills, agreed hours, breaks, maximum hours and other relevant constraints].

Return a table with date, shift, role, assigned staff ID and hours. After the table, list uncovered shifts, possible rule conflicts, total hours by staff member and any assumptions. Do not guess availability or qualifications. Mark missing information “[To confirm]”. This is a draft for a manager to review, not an approved roster.

Check the roster against your records, relevant agreements and applicable requirements. Confirm it with the people involved before publishing it.

4. Turning meeting notes into minutes or a short report

Summaries can sound certain even when the notes are vague. Ask ChatGPT to separate decisions from suggestions and leave unknown owners or deadlines blank for review.

Act as a careful meeting administrator. Turn the notes below into concise minutes.

Meeting: [name and date]. Attendees: [names or roles, if needed].

Use these sections: summary, decisions, action items, risks or issues, unresolved questions and next meeting. For each action, include the owner and due date only if they are stated in the notes. Mark missing details “[To confirm]”. Separate confirmed decisions from suggestions. Do not add information that isn’t in the notes.

Notes: [paste relevant, de-identified notes].

Check the minutes against the original notes, especially decisions, names, action owners and due dates. If you need a more formal report, add the intended audience and purpose, and ask for a short executive summary followed by findings and actions.

Common prompting mistakes and how to fix them

Instead of…	Try…
“Write me a quote.”	Specify the scope, supplied prices, required sections and what to flag as missing.
“Reply to this customer.”	State the goal, tone, approved facts and commitments it must not make.
Pasting a long bundle of background information.	Include only the details needed for this task, in labelled sections.
Asking for “a good report”.	Name the readers, purpose, sections and preferred length.
Treating a polished response as verified.	Ask for assumptions and gaps, then check the facts against your records.

A useful follow-up prompt is: “Check your draft against the information I provided. List any unsupported details, missing facts or possible conflicts. Do not silently fill gaps.” This gives you another review step, but it doesn’t replace your own checks.

Before you send or use the result

Take a moment to check:

• Privacy: Remove customer and staff details that aren’t needed. Use labels such as “Customer A” or staff initials instead of names. Don’t paste passwords, payment details or confidential records into a tool your business hasn’t approved. Check your account settings and your business’s data rules. OpenAI explains its ChatGPT data controls. For general Australian guidance, see business.gov.au’s information on protecting customer information.
• People: Have someone responsible review drafts before they go to customers or affect staff.
• Prices and calculations: Check amounts, quantities, GST treatment, totals and payment terms against your records.
• Dates and commitments: Confirm quote expiry dates, delivery times, roster dates and promised follow-ups.
• Meaning and tone: Make sure the draft reflects what was agreed and sounds like your business.

ChatGPT can organise information and help you get past a blank page. It can also misunderstand instructions or produce details that weren’t supplied. Treat its output as a draft to check, particularly when it affects money, customers or staff. The Australian Government’s AI guidance for businesses also recommends testing AI tools, keeping people accountable and checking outputs before relying on them.

Start with one prompt your team uses often

Choose a repeat task, such as preparing a quote or summarising a meeting. Try the prompt on a few low-risk examples, make a note of what needs correcting, then improve the instructions. A saved prompt that fits your actual process is more useful than a long list of generic templates.

If you’d like help shaping prompts into a practical workflow for your team, you can explore AI Driven Biz’s services. There’s no need to overhaul everything at once.