Getting started · 6 min read

Writing AI prompts that get answers you can use.

Most people who say AI is overrated have only ever given it vague, one-line prompts and got vague, generic answers back. That's not the tool failing, it's a fair conversation gone half-finished. The good news is that getting useful answers is a skill anyone can pick up in an afternoon, and it comes down to a simple, repeatable way of asking.

The short version: a good prompt gives the AI four things: a role to play, the context it needs, the task you want done, and the format you want it back in. Add an example if you can, then iterate on the reply rather than expecting one perfect shot.

Why prompts fall flat

Almost every disappointing answer traces back to the same two faults: the prompt was too vague, or it gave no context. Ask for "a marketing email" and the AI has nothing to work with. It doesn't know what you sell, who you're writing to, or what you sound like, so it falls back on bland filler that could belong to anyone.

It helps to picture a brand-new casual on their first shift. Capable, but they know nothing about your business yet. Hand them a one-line instruction and you'll get a guess. Brief them properly and they do good work. AI is the same. It's only as good as the brief you give it.

The four things every prompt needs

Here's the structure that does the heavy lifting. You won't always need all four, but they're the dials to reach for when an answer isn't landing.

  1. A role. Tell it who to be. "You're a friendly receptionist for a plumbing business" sets a tone and a perspective straight away.
  2. The context. Give it the background it can't know: what you sell, who the answer is for, the situation, any constraints. This is the bit people skip, and it's the bit that matters most.
  3. The task. Say exactly what you want done. Not "help with marketing" but "write a short Facebook post announcing we now do hot water repairs".
  4. The format. Tell it the shape of the answer: how long, what tone, bullet points or paragraphs, plain language or formal. Otherwise you take pot luck.

One more thing lifts results noticeably: an example. Paste in a past email or post you were happy with and say "match this style". The AI copies your style far better from a sample than from a description.

Vague versus specific, side by side

The difference is stark once you see it. A weak prompt: "Write a follow-up email to a customer." You'll get something generic and a bit robotic.

A strong one: "You're the owner of a small landscaping business on the Coffs Harbour coast. Write a short, warm follow-up email to a customer I quoted last week for a garden makeover, who hasn't replied. Friendly, no pressure, offer to answer any questions. Keep it under 120 words and sign off from James." That gives a role, context, the task and the format, and the reply comes back close to ready to send.

Same tool, completely different result. The only thing that changed was the quality of the brief.

Iterate, don't chase perfection

This is where good prompting really lives, and it's the habit most people miss. Don't expect the first answer to be perfect, and don't keep rewriting the original prompt trying to nail it in one shot. Just have a conversation with the reply.

"Make it shorter." "A bit warmer." "Add a line about our spring discount." "Take out the bit about timelines." Each nudge gets you closer, fast. Two or three rounds of small corrections beats ten minutes spent crafting the perfect opening prompt. Think of it as editing a junior's draft, not commissioning a finished piece.

A few prompts to try this week

To get the feel, here are starting points you can adapt to your own business. Add your real context to each:

  • Replies: "You run a [trade]. A customer emailed asking [their question]. Write a clear, friendly reply that [your answer]. Keep it short."
  • Marketing: "Write three short social posts for a [business type] in [town], announcing [offer]. Casual, local, no hashtags spam."
  • Tidying up: "Here's a rough quote I wrote. Make the wording clearer and friendlier but keep all the prices and details exactly the same: [paste]."

The same role-context-task-format approach works whether you're using ChatGPT, Claude or Gemini, so the habit is worth building once. If you're still deciding which tool suits you, our comparison of ChatGPT vs Claude vs Gemini lays them out plainly, and our guide to practical ways to use ChatGPT in a small business shows the everyday jobs worth pointing these prompts at.

Bottom line: better answers come from better briefs, not magic words. Give the AI a role, the context, the task and the format, show it an example where you can, and refine the reply instead of chasing one perfect prompt. Do that and any of the big tools becomes genuinely useful.

Want AI to actually pull its weight?

The first conversation is free. You'll get a plain-English steer on the tools and prompts worth using in your business, and the jobs where it's genuinely worth your time.

Start a conversation

Prompt questions, answered.

Why are my AI prompts not working?
Usually the prompt is too vague and missing context. If you ask for "a marketing email" the tool has no idea who you are, who it's for or what you sell, so it guesses, and the result feels generic. The fix is to spell out the role, the context and exactly what you want back.
What makes a good AI prompt?
A good prompt gives the tool four things: a role to play, the context it needs, the specific task, and the format you want the answer in. An example of the style you're after helps too. Be specific where it matters and the answer comes back far closer to usable.
Do I need different prompts for ChatGPT, Claude and Gemini?
No. The same simple structure works across ChatGPT, Claude and Gemini, because they all respond better to clear context and a specific ask. You might tweak wording between them over time, but the role-context-task-format approach travels well to any of them.
How long should an AI prompt be?
As long as it needs to be to give the right context, and no longer. A one-line prompt for a quick rewrite is fine. A first-draft customer email or marketing piece usually needs a short paragraph of background. More useful detail almost always beats a shorter, vaguer prompt.
Should I expect the first answer to be perfect?
No, and expecting that is the most common mistake. Treat the first reply as a starting point, then refine it: "make it shorter", "warmer", "add a line about X". Iterating gets you there faster than chasing one perfect prompt. JDCS offers a free first conversation if you'd like a hand getting going.