Prompt Engineering: A Practical Guide With 10 Templates

The difference between a vague AI answer and a genuinely useful one is usually the prompt. You don’t need tricks or magic words; you need to give the model what a smart new colleague would need to do the job well. Here’s how, with ten templates you can copy. (Creating pictures instead? See how to write better image prompts.)

The six elements of a strong prompt

1. Role: who the model should act as. “You are an experienced B2B copywriter.”
2. Context: background, audience and purpose. “This email goes to IT managers at mid-sized companies.”
3. Task: exactly what you want. “Write three subject lines and a 120-word email.”
4. Constraints: tone, length, what to avoid. “Plain English, no hype, no exclamation marks.”
5. Format: how the output should look. “Return a table with columns for subject line and angle.”
6. Examples: a sample of what good looks like, if you have one.

You won’t need all six every time, but the more important the task, the more of them you should include.

Techniques that reliably help

Be specific about the audience

“Explain this for a busy CFO” produces a very different answer from “explain this for a first-year student”.

Ask for a plan before the answer

For complex tasks, ask the model to outline its approach first, then produce the result. You can correct the plan before it writes 2,000 words in the wrong direction. Whatever the prompt, check facts in the output; see why AI hallucinates.

Give examples

Showing one or two examples of the style, format or level you want (called few-shot prompting) is often more effective than describing it.

Break big tasks into steps

Research, outline, draft and edit in separate prompts. Each step is easier to check.

Ask it to check its own work

“Review your answer for factual errors, missing steps and unclear wording, then give me a corrected version.”

Iterate, don’t restart

Treat the first answer as a draft. “Shorter.” “More formal.” “Add a concrete example for point 2.” Small follow-ups beat rewriting the whole prompt.

10 templates you can copy

1. Summarise a document

“Summarise the text below for [audience] in [number] bullet points. Start with the single most important takeaway. Flag anything that needs a decision.”

For long files, see how to summarise PDFs with AI safely.

2. Draft an email

“You are a [role]. Write a [tone] email to [recipient] about [topic]. Goal: [outcome]. Keep it under [number] words. End with a clear next step.”

3. Explain a concept

“Explain [concept] to [audience]. Use one everyday analogy, one concrete example and avoid jargon. Finish with a two-sentence recap.”

4. Brainstorm ideas

“Give me 15 ideas for [goal]. Mix safe, bold and unusual options. For each, add a one-line reason it could work.”

5. Improve writing

“Edit the text below for clarity and concision. Keep my voice. Show the edited version, then list the three most important changes.”

6. Make a plan

“Create a step-by-step plan to [goal] in [timeframe]. Include milestones, risks and what to do first this week.”

7. Compare options

“Compare [option A] and [option B] for [use case]. Use a table covering cost, effort, risks and best fit. Then recommend one and explain why.”

8. Prepare for a meeting

“I’m meeting [person/role] about [topic]. List the questions they’re likely to ask, strong answers for each and one question I should ask them.”

9. Practise an interview

“Act as an interviewer for a [role] position. Ask one question at a time, wait for my answer, then give specific feedback.” For engineers, pair this with Backend Architect’s system design interview framework.

10. Turn notes into a document

“Turn these rough notes into a clear [document type] with headings, short paragraphs and an action list. Don’t invent facts; mark anything unclear with [?].”

Common mistakes

• Too vague: “Write about marketing” invites generic answers.
• Too much at once: ten tasks in one prompt leads to shallow results.
• Trusting without checking: models can state errors confidently. Verify facts, figures and citations.
• Hiding the goal: tell the model why you need something; it helps it choose what matters.

Frequently asked questions

Is prompt engineering still useful as models improve?

Yes. Better models need fewer tricks, but clear context, goals and formats still make a big difference.

Do these templates work in any AI chatbot?

Yes. The principles apply to all major AI assistants, though results vary between models.

How long should a prompt be?

As long as it needs to be. A short question is fine for simple tasks; important tasks deserve a paragraph of context.