— How-to

How to write an AI prompt that actually works

Beau Robards
By Beau Robards, Certified Claude Expert
Updated 15 July 2026 · 6 min read
The short answer

To write a good AI prompt, give it four things: context, a role, a clear task, and a format — then add a short example of what “good” looks like. Be specific: a vague ask gets a vague answer. That’s the whole method, and it works the same across ChatGPT, Claude, Copilot and Gemini.

Most disappointing AI results come down to one thing: the prompt was too vague. The good news is that writing a good one is a simple, repeatable skill — no coding, no jargon. Here’s the method, then a worked example turning a weak prompt into a strong one.

The four parts of a prompt that does the job

A strong prompt usually gives the AI four things. You won’t always need all four, but the more you include, the closer the first draft lands:

  1. Context — the background it needs. Who are you, what’s the situation?
  2. Role — who it should act as. “Act as our office manager / a friendly copywriter.”
  3. Task — exactly what to produce. Be specific about the output.
  4. Format — how you want the answer. Length, structure, tone.

If you want the underlying idea, we cover it in what prompt engineering is. But you can start with the four parts right now.

Step by step: turn a vague ask into a good one

Say you want AI to reply to a customer chasing a delivery update. Here’s the weak version most people type:

Weak prompt
Write a reply to this customer.

Now add the four parts, one at a time:

Strong prompt
Act as our customer service lead at a small homewares store. A customer ordered a lamp; it's delayed a week from our supplier. Write a reply that apologises, gives the new timeframe, and offers free express shipping. Warm and genuine, under 100 words, no jargon.

Same tool, wildly different result. The second version returns something you could almost send as-is.

Add an example (the biggest single upgrade)

If you have a past reply you were happy with, paste it in: “Match the tone of this example.” Showing the AI what “good” looks like is often more powerful than describing it, and it’s the fastest way to get output that sounds like you rather than a generic bot.

Iterate: tell it what to fix

Your first result won’t always be perfect — that’s normal. Don’t start over; just steer it: “Make it warmer,” “shorter,” “drop the last line,” “more formal.” A couple of quick nudges usually gets you there, and it teaches you what this tool responds to.

Save it: turn a good prompt into a reusable one

Once a prompt works, don’t retype it next time. Save it as a template with blanks for the details that change:

Reusable template
Act as our customer service lead at [business]. Situation: [what happened]. Write a reply that [goals]. [Tone], under [N] words, no jargon.

Now the recurring job takes ten seconds: paste, fill the blanks, done. A small library of these for your common tasks is where the real hours come back — and it’s exactly the kind of thing your whole team can share.

— Build your own prompt library

Reusable prompts, for your real tasks

In our AI Training & Enablement program, non-technical Australian staff build and save reusable prompts for the recurring work in their own week — so the time saving compounds.

Book a call

Frequently asked questions

How do I write a good AI prompt?
Give the AI four things: context (the background it needs), a role (who it should act as), a clear task (exactly what to produce), and a format (how the answer should look). Then add a short example of what ‘good’ looks like. The more specific you are, the better and less generic the result — a vague ask gets a vague answer.
What should an AI prompt include?
A strong prompt usually includes context, a role, a specific task, a desired format, and ideally an example. You don’t need all five every time, but the more of them you provide, the closer the first draft will be to what you actually want. Vague, one-line prompts are the most common reason people get disappointing results.
Why does my AI give bad or generic answers?
Almost always because the prompt was too vague. The AI can only work with what you give it, so ‘write me an email’ returns something generic. Add who it’s for, what it’s about, the tone, and the outcome you want, and the answer improves dramatically. The quality of the output tracks the quality of the instruction.
What is a prompt template?
A prompt template is a reusable, fill-in-the-blanks version of a prompt you use for a recurring task. You write it well once, then each time you just drop in that day’s details rather than re-explaining the job. Building a small library of templates for your common tasks is where a lot of the real time saving comes from.
Do prompts work the same across ChatGPT, Claude and Gemini?
Largely, yes. The core skill — being clear about context, role, task and format — transfers across ChatGPT, Claude, Copilot and Gemini. Each tool has small quirks, but a well-structured prompt works well in all of them, which is why the skill is worth learning independently of any one product.

Keep reading

All resources