Blog · 7 min read

ChatGPT Prompts for Small Business Owners (That Actually Work)

Most prompt lists give you generic output. Here's the structure that makes ChatGPT usable for a small business — plus 10 prompts you can copy.

Most small business owners have already tried ChatGPT. They opened it, typed something like "write a post about my product," got back three paragraphs of beige marketing language, and quietly closed the tab.

The conclusion people draw is that AI isn't there yet for real business work. That conclusion is wrong, but the experience behind it is real — and it comes down to one thing that nobody explains up front.

The problem isn't the AI. It's the missing 80%.

Here's the same request, twice.

What most people type:

Write a post about my product.

What actually works:

I run a small bakery. Write an Instagram post about our sourdough subscription boxes, 80–120 words.
Details: a weekly box with two fresh loaves, delivered Friday mornings. Long fermentation, no additives. Main benefit: fresh bread every week without thinking about it. Tone: friendly.
Structure: an opening line that stops the scroll (no clichés like "We're excited to announce"), then the core value, then a call to action. Add 5 relevant hashtags.

The second one is longer to type — maybe forty extra words. But the first one produces something you'd never publish, and the second one produces something you'd publish with light edits.

The difference isn't skill. It's that the second version answers questions the AI would otherwise have to guess at: who is writing, to whom, about what, in what shape, at what length.

The formula

Every prompt that reliably works has the same five parts. Miss one and the output drifts toward generic.

1. Who you are. "I run a small bakery." One line. This alone changes vocabulary, examples, and assumptions more than anything else you can add.

2. What you want. Not "some marketing" — the specific artifact. A post, an email, a checklist, a reply.

3. For whom. Your customer, briefly. "Parents who buy bread weekly" produces different copy than "office managers ordering for meetings."

4. In what format. Structure the output: a numbered list, three options, a table, sections in a specific order. Without this you get a wall of prose.

5. How long. "80–120 words" or "under 10 words each." Models follow explicit limits reasonably well and ignore vague ones entirely.

There's a sixth part that isn't required but changes the result more than people expect: tell it what to avoid. "No clichés like 'We're excited to announce'" removes an entire category of output that models default to.

Ten prompts you can use today

These follow the formula. Everything in [square brackets] is yours to replace.

I run a [type of business].
Write an Instagram post about [product/service], 80-120 words.

Main benefit for the customer: [what they actually get]
Tone: [friendly / premium / plain-spoken]

Structure: an opening line that stops the scroll (no clichés like "We're
excited to announce"), then the core value, then a call to action.
Add 5 relevant hashtags.

2. A week of content in one sitting

Create a plan of 7 posts for next week for my business: [describe in 1-2 sentences].

For each post: day, content type (sales / value / behind-the-scenes / customer
story), the topic in one sentence, and a 2-3 sentence draft.

No more than 2 of the 7 should be a direct sell.

That last line matters. Without it you get seven sales posts, which is how business accounts train their audience to scroll past.

3. Answering "it's too expensive"

A customer says [product/service] is too expensive. Context: [what it is and the price].

Write 3 responses that don't drop the price but show the value: one comparing to
the alternative, one breaking it down per day or month, one framing the long-term
outcome. 2-3 sentences each, no pressure tactics.

Three different angles matter because "too expensive" means at least three different things: I can't afford it, I don't see why it costs this, and I'm negotiating. One canned answer can't cover all three.

4. A procedure your team can follow

Create a step-by-step SOP for this task: [describe it the way you'd explain out loud].

Who performs it: [role]
How often: [daily / weekly / as needed]

Format: a numbered list where every step starts with a clear action verb.
Add a "Common mistakes" section with 2-3 things to avoid.

The "common mistakes" section is the part people skip and then regret. It's where the actual institutional knowledge lives — the stuff you'd normally only learn by getting it wrong.

5. Replying to an angry customer

A customer sent this complaint: "[paste it]"

Write a reply: first acknowledge the problem without excuses, then a concrete fix,
then an offer to make it right. Tone: genuine, no canned phrases like "your feedback
is important to us."

Explicitly banning "your feedback is important to us" is doing real work here. Models reach for that phrase constantly, and customers read it as a form letter.

6. Making sense of your own numbers

Here are this month's numbers: [revenue, main expense categories].

Explain in plain language what these mean for the business: is this good, what
deserves attention, what are the 2-3 most important takeaways. No financial jargon.

This isn't accounting advice and shouldn't be treated as such. What it's good for is seeing your own figures from a different angle before you talk to someone who does this professionally.

7. A job posting that sounds human

Write a job posting for [role] at a [type of business].

Responsibilities: [3-5 of them]
Terms: [hours, pay range if you'll state it, location]

Structure: brief intro to the company (2 sentences, no clichés), responsibilities
as a list, requirements kept separate from "nice to have", terms, how to apply.
Tone: friendly.

Keeping requirements separate from "nice to have" measurably widens your applicant pool — people, especially good ones, self-reject against long requirement lists.

8. Following up without sounding desperate

I sent a customer a proposal for [description] [X days] ago and got no reply.

Write a short follow-up (2-3 sentences) that doesn't sound pushy or desperate.
Make it easy for them to say "no" without awkwardness.

"Make it easy to say no" is counterintuitive and it works. It converts a message people avoid answering into one they can close out in ten seconds — and a clear no is worth more than silence.

9. Finding the real cause of a recurring problem

We keep running into this problem: [describe it].

Help me find the real cause rather than the symptom: ask me 5 consecutive
"why does this happen?" questions, one at a time, each based on my previous answer.
At the end, state the root cause and 2 possible fixes.

The "one at a time" instruction is what makes this work. Without it, the model writes out all five questions and answers them itself, which defeats the point.

10. Turning a rambling explanation into instructions

Here's how I currently do [task]: [describe it messily, out of order, however
it comes out].

Turn this into clear step-by-step instructions someone else could follow.
Flag anything that seems to be missing.

This is the fastest way to write documentation you've been putting off. You don't have to organize your thoughts first — that's the part the model is genuinely good at.

What still doesn't work

Being honest about the limits is more useful than another list of tips.

Anything with legal consequences. Dismissals, debt collection, contracts. The model will produce confident text that ignores the law in your country entirely. Use it to prepare a conversation, never to determine a procedure.

Decisions about specific people. Whether to promote someone, whether to fire someone, how to handle a conflict between two employees. It can help you structure your thinking; it cannot know the people.

Anything requiring current facts. Prices, regulations, what your competitor did last month. Models don't know and will sometimes invent.

Your actual voice, on the first pass. The output will be competent and slightly generic. Reading it aloud is the fastest test — if it doesn't sound like you, ask for "simpler, less formal, more like how I'd actually say it." That one follow-up does more than any amount of prompt tweaking.

The one habit worth building

Not "learn prompt engineering." Just this: before typing a request, spend fifteen seconds answering who, what, for whom, what format, how long.

That's the whole thing. Everything above is that habit applied to specific situations.

The people who get value out of AI in their business aren't the ones who learned clever tricks. They're the ones who stopped typing four-word requests and started spending an extra thirty seconds on context.

I put together 150 prompts like these, organized by business area — marketing, sales, operations, finance, hiring — along with spreadsheet templates and a seven-day implementation guide. You can try a sample of them here with the fill-in fields already built in.