ChatGPT Prompts for Business: Templates That Actually Deliver

15 working templates for the tasks business owners deal with every week, a formula for writing your own, and a checklist of what to verify before anything goes out to a customer

ChatGPT prompts decide whether you get a usable draft or a page of generic filler you’ll end up rewriting. Most business owners who tried ChatGPT and gave up ran into the same thing: bland, one-size-fits-all answers that know nothing about their company.

The problem is almost never the model. It’s a prompt with no context, no constraints and no clear format. So the real question behind “how do I use ChatGPT for my business?” is “how do I brief it properly?”. This article gives you a simple formula for briefing ChatGPT on any business task and 15 ready-made prompt templates for sales, marketing, analysis and your team — each with what to fill in and what to check before you use the answer.

In short

A good business prompt has six parts: a role (who’s answering), context (your company, customer and facts), the task (what to do), constraints (what to avoid, length, tone), the format (what the answer should look like) and a check (what you’ll verify before using it). The more specific you are, the less you rewrite. Always double-check facts, figures, anything legal, and anything involving personal data yourself — ChatGPT can make mistakes.

What is a prompt, and why doesn’t “write a post about our business” work?

A prompt is the request or instruction you give ChatGPT or any other AI model. Everything depends on it: the model doesn’t know your company, your customers, your prices or your tone of voice, so it fills the gaps with the most average, generic option available. Hence the bland copy.

Anthropic’s prompting guide suggests treating the model like a brilliant but brand-new employee who doesn’t yet know your norms and workflows. It also offers a simple test: show your prompt to a colleague who knows nothing about the task and ask them to follow it. If they’d be confused, the model will be too. OpenAI’s guide makes the same point in its own way: give the model a role, spell out what it should and shouldn’t do, and show it an example of the answer you want.

The first gets you a post that could belong to any company on earth. The second gets you something you can publish after five minutes of editing.

How to write ChatGPT prompts for business: a six-part formula

How do you write prompts that don’t need rewriting? We use a six-part formula. Not every task needs all six, but the more of them your prompt includes, the sharper the answer.

Two habits make a big difference. First, give an example: if you have an email or post you like, paste it in and say “match this style”. Second, invite questions: “If anything’s missing, ask me before you start.” That stops made-up details better than any “don’t invent facts” instruction.

In the templates below, anything in square brackets is yours to fill in. Swap in your details and the prompt is ready to go.

ChatGPT prompts for sales and customer communication

Business prompts are best started with customers, because that’s where ChatGPT saves time fastest. Four templates for things that come up every week: pushback on price, a meeting follow-up, a negative review and a first email to a new prospect.

1. Handling “it’s too expensive”

When to use it. A customer says your price is too high, and you don’t want to discount or send a canned reply.

You're an experienced sales manager in [your industry].
Context: we sell [service] for [price]. The customer — [who they are] — wrote: "[customer's message]".
What's included in the price: [what they get]. Our genuine strengths: [2–3 strengths].
Task: write 3 possible replies to the customer.
Constraints: no discounts, no pressure, no promises beyond the strengths listed. Under 80 words each.
Format: each reply, followed by one sentence on which kind of customer it suits.

What to fill in:

• The customer’s exact words, not your summary of them.
• What the price actually covers.
• One or two of your own replies if you want to keep your tone.

What to check in the answer:

• No guarantees or promises you don’t actually make.
• Price, timelines and scope match reality.
• It sounds like you, not a sales brochure.

2. Follow-up email after a meeting

When to use it. You’ve just had a call or meeting and need to put what was agreed in writing — fast.

You're an assistant to the owner of a [industry] business.
Context: notes from a meeting with a client: """[notes or transcript]""".
Task: write a follow-up email summarising the meeting.
Constraints: only what's in the notes, no new commitments. Professional but warm. Under 150 words.
Format: one line of thanks → what we agreed (bullet list) → next steps with dates → who's responsible for what.

What to fill in:

• Raw notes or a call transcript — the model is good at structuring them.
• Dates, amounts and roles (skip unnecessary personal details).

What to check in the answer:

• Dates, amounts and scope match what was actually agreed.
• Nothing has been added that wasn’t discussed.

3. Replying to a negative review

When to use it. A critical review has appeared on Google, Trustpilot or social media and needs a public reply.

You're a customer care manager in [industry].
Context: review on [platform]: "[review text]".
What actually happened: [your side of the story]. What we're prepared to offer: [solution].
Task: write a public reply to the review.
Constraints: no excuses or arguing, don't reveal any customer details, under 90 words.
Format: acknowledge their experience → one-sentence explanation → what we're doing about it → where to contact us to resolve it.

What to fill in:

• Your honest version of events — without it the model will either deny everything or admit to everything.
• A specific fix you’re genuinely willing to offer.

What to check in the answer:

• The reply doesn’t admit fault for something that didn’t happen.
• No names, phone numbers or order details.
• The offer is real.

4. Cold email to a prospect

When to use it. You’re writing to a company that doesn’t know you yet and you want a reply, not the spam folder.

You're a B2B copywriter.
Context: we're [company, what we do]. We're writing to a [job title] at a [type of company].
Their typical problem: [problem]. Our proof: [case study or number].
Task: write a first-touch email.
Constraints: under 120 words, one offer, no filler like "hope this finds you well". The ask is a quick reply, not an hour-long call.
Format: 3 subject line options + the email body.

What to fill in:

• A specific problem that type of company actually has.
• Real proof: a result, a number, an industry.

What to check in the answer:

• The proof is true and the numbers aren’t inflated.
• The model hasn’t invented details about the recipient’s company.

ChatGPT prompts for marketing

Four marketing prompts that cover the most common tasks for a small business without a dedicated marketer.

5. Service page copy

When to use it. A service page on your website is empty or full of generic phrases and needs rewriting.

You're a copywriter who writes websites for service businesses.
Context: service — [name]; for — [customer]; problem it solves — [problem]; how we work — [steps]; price — [price or "from"]; timeline — [timeline]; what makes us different — [2–3 facts].
Task: write the copy for the service page.
Constraints: no "tailored solutions" or "end-to-end approach", only facts from the context, 250–350 words.
Format: H1 → subheading → three H2 sections (who it's for, how we work, price and timeline) → a 3-question FAQ → call to action.

What to fill in:

• The words customers use to describe the problem — from calls, emails and reviews.
• Real prices and timelines, even as a range.

What to check in the answer:

• No promises of results you can’t guarantee.
• Prices and steps match reality.
• The headline uses the phrasing customers actually search for.

6. A month’s content plan

When to use it. You post on social or your blog whenever you can and want a plan tied to real customer questions.

You're a content marketer.
Context: business — [industry, city]; audience — [who]; channels — [Instagram, blog, newsletter]; common customer questions: [list].
Task: create a 4-week content plan.
Constraints: 3 posts a week; every topic answers a real question from the list; no holiday greetings or generic motivational posts.
Format: table — week | channel | topic | format | which customer question it answers.

What to fill in:

• A list of 10–20 questions customers genuinely ask.
• The channels you actually run, and how many posts you can realistically manage.

What to check in the answer:

• Every topic links back to a customer question.
• The plan fits your resources.
• No repeated topics.

7. Customer pain points from reviews

When to use it. You want to know what really bothers customers and use it in your website copy and ads.

You're a customer experience analyst.
Context: below are 20–50 reviews of us and other businesses in our niche (names removed): """[reviews]""".
Task: find the recurring pain points and expectations.
Constraints: only what's in the reviews; for each theme, say how many times it comes up.
Format: table — pain point | example quote | how often | what it means for our website or service.

What to fill in:

• Reviews from Google, Trustpilot, marketplaces and social media — with names and contact details removed.
• Not just your own reviews: niche-wide pain points often show up more clearly in other companies’ reviews.

What to check in the answer:

• The quotes really appear in the reviews.
• The counts aren’t invented — spot-check two or three.

The pain points this prompt surfaces are ready-made topics for pages and articles. We show how to turn them into a site structure in Keyword Research: How to Build a Keyword Map.

8. Ad headline variations

When to use it. You’re launching ads and want several headline options to test.

You're a performance marketer.
Context: an ad on [ad platform] for [service]; audience — [who]; main benefit — [benefit]; character limit — [number].
Task: suggest 10 headline options.
Constraints: within the character limit; no "best" or "#1"; 3 options with a number, 3 phrased as a question.
Format: numbered list with the character count next to each option.

What to fill in:

• The platform’s exact character limit.
• One main benefit, not five.

What to check in the answer:

• The length of every headline — count it yourself; models often miscount characters.
• The wording follows the ad platform’s policies.

ChatGPT prompts for analysis and decisions

Four templates that help you get to grips with numbers and documents faster. Checking matters most here: the model can be confidently wrong in its calculations.

9. Analysing a sales spreadsheet

When to use it. You’ve got a sales export and no time to dig through it.

You're a financial analyst for small businesses.
Context: sales data for [period], columns: [list of columns]. [Paste the table or upload the file.]
Task: find the 3–5 most important patterns.
Constraints: use only the data provided; if something's missing, say exactly what; show your working.
Format: finding → the numbers behind it → what to check or do next.

What to fill in:

• The table with column names and no personal customer data.
• The question you want answered: what’s growing, what’s falling, what’s profitable.

What to check in the answer:

• Recalculate one or two key figures by hand.
• Correlation isn’t being presented as cause.
• All rows were included, not just some.

10. Comparing options against criteria

When to use it. You’re choosing a supplier, software, premises or strategy and have more than two options.

You're a decision-making consultant.
Context: we're choosing between [option A], [option B] and [option C]. Weighted criteria: [e.g. price 40%, timeline 30%, risk 30%]. What we know about each option: [facts].
Task: compare the options.
Constraints: only the facts provided; where facts are missing, write "no data".
Format: a 1–5 scoring table → weighted total → 3 questions to answer before deciding.

What to fill in:

• Weighted criteria — without them the model decides what matters.
• Facts about each option, including the uncomfortable ones.

What to check in the answer:

• The scores reflect the facts.
• You make the decision — the table just organises your thinking.

11. Summarising a long document

When to use it. You need to get the gist of a contract, report or proposal quickly.

You're an executive assistant.
Context: [contract / report / proposal]: """[text]""".
Task: summarise it so I can make a decision.
Constraints: under 200 words; call out amounts, dates, obligations and penalties separately; reference clause numbers.
Format: the gist in one paragraph → key terms as a list → what to clarify.

What to fill in:

• The full text, not an excerpt — the model can’t know what’s in the part you left out.
• What matters most to you: money, deadlines or risk.

What to check in the answer:

• Check every amount and date against the original.
• Legal conclusions only after a lawyer has reviewed them.

12. Preparing for a negotiation

When to use it. You’ve got an important conversation coming up with a client, partner or supplier.

You're a negotiation coach.
Context: negotiating with [who] about [subject]. Our goal: [goal]. Our walk-away point: [minimum acceptable terms]. What we know about the other side: [facts].
Task: prepare a negotiation plan.
Constraints: no manipulative tactics; arguments based only on the facts provided.
Format: 5 likely objections with responses → 3 questions worth asking → a fallback option if we can't reach a deal.

What to fill in:

• Your walk-away point — the model needs to know your limit.
• Facts about the other side: what matters to them, what they’ve already offered.

What to check in the answer:

• The arguments rest on your facts.
• None of the prepared lines gives away your walk-away point.

ChatGPT prompts for your team and processes

Three templates for jobs small business owners often end up doing themselves: hiring, onboarding and documenting processes.

13. Writing a job ad

When to use it. You’re hiring and want an ad that attracts the right candidates.

You're an HR specialist for small businesses.
Context: role — [title]; tasks — [5–7 real tasks]; hours — [schedule]; location — [office / remote / hybrid]; salary — [range]; what we offer — [benefits].
Task: write the job ad.
Constraints: no "fast-paced environment" or "rockstar" clichés without explanation; be honest about the harder parts of the job; under 300 words.
Format: title → about us (2 sentences) → what you'll do → must-haves and nice-to-haves → what we offer → how to apply.

What to fill in:

• The real tasks of the first month, not generic duties.
• A salary range — ads that include one tend to attract more relevant applicants.

What to check in the answer:

• No discriminatory requirements (age, gender, family status) — they may also break employment law.
• The terms match what you actually offer.

14. An onboarding plan

When to use it. A new hire starts next week and there’s no plan for their first month.

You're an operations manager.
Context: a new [job title] starts on [date]. Key tools: [list]. What they should be able to do after a month: [outcome].
Task: create a plan for their first month.
Constraints: a realistic workload; one measurable outcome per week.
Format: table — week | tasks | who helps | outcome for the week.

What to fill in:

• The specific outcome you expect after a month.
• Who on the team can support the new hire.

What to check in the answer:

• The first week’s workload is realistic.
• Tool access is set up before day one.

15. Documenting a process

When to use it. A process lives in your head and needs handing over to the team.

You're a business process consultant.
Context: here's how we currently [process, e.g. handle a new enquiry], in my own words: """[description]""".
Task: turn this into a standard operating procedure.
Constraints: only steps from my description; if you spot a gap or contradiction, ask me rather than inventing a step.
Format: purpose → roles → numbered steps (who, what, by when) → what to do if something goes wrong → questions for me.

What to fill in:

• The description as you’d explain it to a new hire — even if it’s messy.
• Roles rather than names: “account manager”, “bookkeeper”.

What to check in the answer:

• You’ve answered the model’s questions and added the answers.
• Roles and deadlines are realistic.

What shouldn’t you hand over to ChatGPT without checking?

ChatGPT sounds convincing even when it’s wrong. OpenAI warns right in the interface that ChatGPT can make mistakes and recommends checking important information. So in three areas, its answer is only ever a first draft.

• Facts and figures. Statistics, competitor prices, names of regulations and dates can be made up. If a number is going into public copy or a decision, find the original source.
• Legal, tax and medical questions. The model can help you understand a document and prepare questions for a professional, but it doesn’t replace a lawyer, accountant or doctor.
• Customers’ personal data. On personal accounts, your conversations may be used to train OpenAI’s models by default; you can switch this off with the “Improve the model for everyone” setting in Data Controls. According to OpenAI, content from its business workspaces isn’t used for training by default. Even so, keep names, phone numbers and order details out of prompts unless you really need them — data protection rules such as GDPR still apply.

How do you share prompts so the whole team can use them?

Even the best prompt is useless if it lives in one person’s chat history. Turn your prompt templates into a shared library so ready-made prompts save time for the whole team, not just one person.

• One document or spreadsheet with columns for the task, when to use it, the prompt, an example of a good answer and who owns the template.
• Projects with instructions. ChatGPT and Claude both let you set up a project with standing instructions: your company description, tone of voice, prices and phrases to avoid. Then nobody has to paste in the context every time.
• Keep examples of the best answers next to each prompt — they’re the clearest guide for both the model and new team members.
• Review the library every quarter: what’s out of date, which prompts have stopped delivering, which tasks are new.

Frequently asked questions

Seven questions about using ChatGPT for business that we hear most often.

What to do next

Pick two or three tasks that come up every week, copy the matching templates and fill in your details. Within a week you’ll have the first version of your prompt library — and an honest sense of where ChatGPT really saves time and where it doesn’t.

Want AI working inside your processes rather than in a separate chat window? Get in touch for a free review of your website and workflows.

ChatGPT doesn’t know your business. A good prompt is the five minutes it takes to introduce them.

— Iryna Lehusha, founder of Be Visible

Sources

Prompt structure recommendations come from OpenAI’s and Anthropic’s official guides; data settings from OpenAI’s help centre.

1. OpenAI — Prompt engineering
2. Anthropic — Prompting best practices
3. OpenAI Help — Data Controls FAQ