How to prompt ChatGPT for your small business (and stop getting generic answers)

May 2, 2026

Vague question. Vague answer. Vaguely disappointed. Close the tab.

If that's been your experience with ChatGPT, the tool isn't the problem. The approach is.

This post covers two things: which tasks in your business are worth handing to ChatGPT, and exactly how to brief it so what comes back sounds like you wrote it. Not like a response to a contact form no one actually reads.

Key takeaways

ChatGPT works like a new staff member who knows nothing about your business until you tell it. Vague brief, vague output.
Bucket one tasks, including payment reminders, booking confirmations, and document follow-ups, can go straight to AI. Your client just wants the outcome.
Bucket two tasks, including proposal drafts, new enquiry responses, and post-meeting summaries, use AI for the first draft, but you review and send.
A good prompt tells ChatGPT who you are, who the client is, what just happened, and what you want them to do next.
Bucket three tasks, including complaints, first real conversations with new clients, and anything where someone is worried, never go to AI.

Why does ChatGPT give such generic answers?

Because you haven't told it anything.

Imagine you've just taken on a work experience kid. First day. They're keen, they've got no idea what your business does, no idea who your clients are, no idea how you talk to people. You walk up, hand them a task, and say: write me a follow-up email to a client.

What are they going to give you? Something generic. Safe. Completely forgettable. Not because they're hopeless. Because you gave them nothing to work with.

ChatGPT is that work experience kid. Capable. Willing. But it knows absolutely nothing about your business until you tell it. The brief is the work. And most people skip it entirely.

"The brief is the work. And most people skip it entirely."

Which tasks in a small business should go to ChatGPT?

Not all of them. The bucket framework from the previous video in this series gives you the filter. The question that sorts every task:

Would the client you'd least want to lose feel less valued knowing a tool handled this?

Yes means it stays with you, full stop. No means keep reading.

Bucket one: full AI, no review needed

These are tasks where your client just wants the outcome. They don't care whether a person or a system handled it. They just need it done.

Examples: payment reminders sent three days after a job, appointment confirmations, document follow-up messages to clients who haven't sent something through, FAQ responses to common enquiries, booking confirmations with prep instructions.

Nobody cares if a payment reminder sounds human. They just need to know when to pay. This is where AI saves the most time for the least risk.

Bucket two: AI drafts, you review

These are tasks where your judgment and your knowledge of the client still matter. But AI can do the heavy lifting on the first draft.

Examples: proposal drafts from your meeting notes, responses to new enquiries where you read the draft before it goes, post-meeting summaries, social content where AI gives you the structure and you add your voice.

The AI does the drafting. You check it, tighten it, personalise it, and send it. The quality is yours. The accountability is yours. ChatGPT just cleared some of the desk.

"You're not learning a new skill. You're learning how to hand things over properly."

Bucket three: keep it human, no exceptions

Complaints. First real conversations with a new client, the kind where they tell you what they actually need, not what they typed in your contact form. Any situation where your client is worried about money, a deadline, or whether they made the right call hiring you.

These conversations aren't about information. They're about trust.

How to write a ChatGPT prompt that actually works

The video shows the before/after live. This section gives you the framework in writing so you can apply it to your own tasks.

A prompt that gets a usable result has four elements. You don't need all four every time, but the more you include, the less editing you'll do at the end.

1. Who you are and how you work

One or two sentences. What kind of business you run, who your clients are, and how you normally talk to people.

Example: "I run a small bookkeeping practice in Sydney. My clients are mostly sole traders and small business owners. I write in a warm, direct tone. Like a conversation, not a corporate letter."

You can save this as a note and paste it at the start of every prompt. Once ChatGPT has this, everything else in the session benefits from it.

2. Who the client is

You don't need their name. You need enough context that the AI understands the situation.

Example: "The client is a tradie who's been working with me for two years. He's pretty relaxed, doesn't like formal language."

3. What just happened

What was the meeting about? What did they ask? What did you agree? The more specific, the better the output.

Example: "We just had a call about his pricing. He's been undercharging for travel time and we're going to fix that. He was a bit nervous about putting his rates up."

4. What you want them to do next

The clearest part to include and the most often missed. What's the one action you want the client to take after reading this?

Example: "I want him to confirm he's happy with the new rate sheet I'm sending separately. Keep it short. Two paragraphs max."

Before and after: the same task, two different prompts

Here's what the difference looks like in practice. Same task. Same tool. One is usable, one isn't.

Tradie: payment reminder

Before (what most people type):

Write a payment reminder email to a client.

What you get: something stiff, formal, and generic. "Dear valued client, this is a reminder that your invoice is outstanding..."

After (with a proper brief):

I'm a tradie in Brisbane. My clients are homeowners, mostly pretty relaxed and informal.

Write a short, friendly payment reminder for a bathroom renovation I completed last week.

The client is a young couple, first home. Invoice is $4,200, due three days ago. Keep it light. I'd rather they pay than feel uncomfortable.

One paragraph. End with the bank details or a prompt to reply if they have any questions.

What you get: something that actually sounds like a real person sent it.

Bookkeeper: reassuring a worried client

Before:

Write an email to a client about an ATO letter.

What you get: formal, cautious, likely to make the panic worse.

After:

I'm a bookkeeper. A long-term client has just received a confusing letter from the ATO and she's in a mild panic.

Write a short email to reassure her I'm looking into it. Warm, calm, direct. No jargon.

Don't overpromise anything. I don't know what's in the letter yet. Just let her know I'm across it and will be back to her by end of day.

What you get: something that reads like it came from someone who knows this client, not a compliance office.

Consultant: follow-up after a discovery call

Before:

Write a follow-up email after a meeting.

What you get: a generic template that could be from any business in any industry.

After:

I run a small business consulting practice. My tone is warm but direct, like a conversation, not a corporate letter.

The client is a sole trader bookkeeper who's been in business four years and is struggling with capacity.

In the discovery call we talked about reviewing her service packages and her pricing. She's interested but a bit nervous about putting her rates up.

Write a follow-up that recaps the key points and gives her one clear next step: book the working session. Keep it short, two paragraphs.

What you get: something the client actually feels was written for her.

The work experience kid didn't get smarter between the bad prompt and the good one. You just became a better manager.

Where to start this week

Pick one task. Just one. Something you already know belongs in bucket one or bucket two. Pick a follow-up you send every week, a confirmation you've typed so many times you could do it in your sleep.

Open ChatGPT. Before you type the task, give it one sentence about your business and your tone. Then give it the task, with context. Who the client is, what happened, what you want them to do next.

Read what comes back. If it sounds like you briefed someone properly, you did it right. If it still sounds off, add more context and go again.

That's the whole process. You're not learning a new skill. You're learning how to hand things over properly. If you've ever brought someone new into your business, you already know how to do it.

The onboarding is the work. And you already know how to do it.

Not sure what to hand off and what to protect?

The AI Task Audit takes three minutes and gives you a personalised result.

Take the free audit

Frequently asked questions

Can I use ChatGPT to write emails to clients? +

Yes, for bucket one and bucket two tasks. For bucket one, including payment reminders, booking confirmations, and document follow-ups, you can let it go out without reviewing every time. For bucket two, including responses to new enquiries, follow-ups after calls, and anything where the client has shared real context, AI writes the draft, you read it and adjust before sending.

Why does AI content sound so robotic? +

Because it doesn't have enough context. When you give ChatGPT a vague prompt, it defaults to safe, generic language that could fit any business in any industry. Add your business, your client, and the specific situation. The output changes completely. The tool isn't the problem. The brief is.

How long should my prompt be? +

Long enough to cover the four elements: who you are, who the client is, what just happened, and what you want them to do next. For most tasks, that's three to five sentences. You don't need an essay. You need enough context that ChatGPT isn't guessing.

Should I automate my client complaints? +

Never. Complaints belong in bucket three without exception. When a client is upset, the automated response that tells them their enquiry has been received and someone will be in touch within two business days is the thing that loses them. A real person responding quickly keeps them.

How do I know which tasks belong in which bucket? +

Use the filter question: would the client you'd least want to lose feel less valued knowing a tool handled this? Yes means bucket three. No means bucket one or two. If you want a structured audit across your full task mix, the AI Task Audit maps your tasks to a specific profile and tells you exactly where to start. It takes about three minutes and is free.

Find out exactly what to hand off and what to protect.

Take the free AI Task Audit — three minutes, personalised result.