Getting good results from AI.
You don't need special phrases or tricks. Ask the way you'd brief a capable new colleague and the answers get much better. Here's how, with everyday examples.
- Give context: who it's for, what it's for and what good looks like. Vague requests get vague answers.
- Treat it as a conversation. The first answer is a starting point, so ask for changes until it's right.
- You're the editor. AI does the first draft, and your knowledge of the work, the people and the facts makes it good.
Prefer to watch?
The video covers the same ground as the page below, in a few minutes.
Brief it like a new colleague
The most common reason people are disappointed by AI is that they type a few words, as they would into a search engine, and get something bland back.
Imagine asking a new temp to “write an email about the meeting”. They'd have questions. Which meeting? Who is it for? What should the reader do next? AI has the same problem, except it won't ask. It will guess, and guesses tend to be generic.
Whatever you type into an AI tool is called a prompt. A good prompt usually covers four things:
- The task. What you want: a draft, a summary, a list of options or an explanation.
- The context. The background it needs: who it's for, what has happened so far and why it matters.
- The format. How long, what style, bullet points or paragraphs, formal or friendly.
- The goal. What a good result would achieve, such as “so the customer feels heard and knows what happens next”.
Everyday examples
Here's the difference in practice. Each better prompt uses placeholders or made-up details, so nothing confidential goes in.
- An email. Instead of “write an email about a delay”, try: “Write a short, friendly email to a customer whose order is running three days late because of a supplier problem. Apologise once, give the new delivery date as [date] and offer to cancel if they'd prefer. Under 120 words, no jargon.”
- A report. Instead of “summarise this”, try: “Summarise this public report for my manager, who has two minutes to read it. Give the three main findings, anything that affects a small retail business, and one question we should be asking ourselves.”
- A spreadsheet. Instead of “help with Excel”, try: “In Excel, column A has invoice dates and column B has amounts. I want a formula in column C that shows ‘Overdue’ if the date is more than 30 days ago and stays blank otherwise. Explain how it works in plain English.”
- A customer query. Instead of “reply to this complaint”, try: “A customer says they were charged twice for one order. Draft a calm reply that acknowledges the problem, explains we're checking with our payments team, and promises an update within two working days. Warm but professional.”
Keep the conversation going
The first answer is rarely the best one. A lot of the skill is in the follow-up, and you can simply say what you'd like changed, just as you would with a colleague. For example:
- “Shorter, and less formal.”
- “That's too salesy. Make it sound more like a real person.”
- “Give me three different versions so I can choose.”
- “What have you assumed here? List anything I should check.”
- “Before you start, ask me any questions you need answered to do this well.”
Habits that help
- Show an example. Paste in some non-confidential text you like and say “match this style”. It's often easier than describing the style.
- Give it the source. If you want an answer based on a particular document, and you're allowed to share it with the tool, attach it and say “answer only from this”. That's far more reliable than asking from memory.
- Say what to avoid. “No jargon”, “don't promise a refund” or “don't mention prices” heads off common problems before they start.
- Give it a role. “Act as an experienced customer service manager and review this reply” can sharpen the feedback you get.
- Break big jobs into steps. Ask for an outline first, agree it, then ask for each part. You stay in control, and each piece is easier to check.
- Ask it to critique your work. “What's unclear in this email?” or “What would a sceptical reader ask?” is a great use of AI, and it leaves the writing in your hands.
- Save what works. Keep a note of prompts that gave good results. Reusing and tweaking them saves time.
- Start afresh when it goes wrong. If a conversation gets muddled, open a new one with a clearer brief rather than piling on corrections.
You're still the editor
However good your prompt, what comes back is a draft. Read it properly, check the facts and make it sound like you. People can often tell when a message has been pasted straight from AI, and it can feel careless or impersonal, especially to customers.
The good news is that the skills behind good prompts are ones you already use: explaining clearly, knowing your audience and knowing what good work looks like. AI just rewards them more.
Before and after
- Pick a routine task you do often, such as replying to a common customer question or writing a weekly update.
- Ask your approved AI tool for it in one short line, the way you might type into a search box. Save the answer.
- Ask again using the four ingredients: task, context, format and goal. Use placeholders instead of any real names or details.
- Ask for one improvement, such as “make it warmer and shorter”, then compare all three versions.
- Save the prompt that worked best somewhere handy. Next time, you'll only need to change the details.
Share it with your manager.
If your whole team is being asked to get up to speed on AI, a short practical session together saves weeks of guesswork.
About team training →Pass it on.
It's free, there's no sign-up, and it's written for people who aren't in IT.