Prompt engineering basics: a good prompt is a good brief. Give the AI context (who it is for and why), a clear task, any constraints (length, tone, what to avoid), examples of what good looks like, and the output format you want. Then iterate: ask it to critique its own draft, and refine with follow-up instructions. Specific prompts produce useful answers; vague prompts produce generic ones.

Key takeaways

• Structure: context, task, constraints, examples, format.
• Showing an example beats describing one.
• Ask for reasoning or steps on complex problems, then check them.
• Iteration is normal. Treat the first answer as a draft.

A simple prompt structure

Part	Question it answers	Example
Context	What’s the situation?	”I manage a 6-person support team at a software company.”
Role or audience	Who is this for?	”Write for new hires with no technical background.”
Task	What exactly do you want?	”Draft a one-page guide to escalating tickets.”
Constraints	What limits apply?	”Under 400 words, plain English, no jargon.”
Examples	What does good look like?	”Match the tone of this paragraph: …”
Format	How should the output look?	”Numbered steps, then a 3-row table of escalation levels.”

Before and after

Weak: “Write an email about the project delay.”

Strong: “I’m a project manager. Write a short email to a client explaining that our website launch moves from 10 to 24 March because of a payment-provider integration issue. Tone: honest, calm, solution-focused. Include what we’re doing to fix it and one action we need from them (approve the revised timeline by Friday). Under 150 words.”

Techniques that work

1. Give examples (few-shot prompting). Paste one or two samples of the output you want.
2. Ask for step-by-step reasoning on calculations or logic, then check each step yourself.
3. Break big tasks into stages: outline first, then sections, then edit.
4. Ask for a critique: “List the three weakest points of this draft and how to fix them.”
5. Specify what to avoid: “Don’t use bullet points” or “Don’t invent statistics. Mark gaps as [DATA NEEDED].”
6. Provide the source material instead of relying on the model’s memory: “Using only the text below, summarise…”

Common mistakes

• No context: the model writes for nobody in particular.
• Multiple tasks crammed together: split them up.
• Accepting the first output: iterate.
• Trusting facts without checking: models can invent sources and numbers.
• Sharing sensitive data in unapproved tools. See the safe-use rules in how to use generative AI at work.

Five reusable prompt templates

1. Summarise: “Summarise the text below for [audience] in [N] bullet points, then list any decisions or open questions.”
2. Explain: “Explain [concept] to someone who knows [X] but not [Y]. Use one analogy and one worked example.”
3. Rewrite: “Rewrite this for [audience] in a [tone] tone, keeping all facts unchanged. Under [N] words.”
4. Analyse: “Here’s a table of [data]. What are the three most important patterns? What would you check before trusting them?”
5. Plan: “Turn this goal into a [N]-week plan with weekly milestones and risks: [goal].”

Where prompt skills lead

Prompt skills are increasingly part of analyst, marketing, operations and support roles. See AI jobs without coding and digital marketing skills. If you want to understand what happens inside the model, read the machine learning roadmap.

Practise with a mentor

Tutopiya is planning a 6-week Generative AI for Working Professionals cohort where you build prompt workflows for your own job and get feedback on them. It isn’t running yet. Register interest in the generative AI cohort. There’s nothing to pay.

Sources

• Anthropic: prompt engineering overview
• OpenAI: prompt engineering guide
• Google: prompting guide for Gemini

Frequently asked questions

What is prompt engineering? +

Prompt engineering is the practice of writing instructions for AI models so they produce useful, accurate and well-formatted output. For most professionals it means writing clear briefs: context, task, constraints and the format wanted.

Is prompt engineering a real job? +

Dedicated prompt engineer roles exist but are relatively few. Far more common is prompt skill as part of other jobs, in marketing, analysis, operations, support and software, where using AI well is a productivity advantage.

Do I need to learn coding for prompt engineering? +

No. Everyday prompt writing needs clear thinking and writing. Coding helps if you want to build AI into applications or automate workflows through APIs.

What is the most common prompting mistake? +

Being too vague. A short prompt with no context forces the model to guess your audience, purpose and format, and it usually guesses generically.

Founding cohorts · nothing charged

Want to learn this live, with a mentor?

Tutopiya is planning small, mentor-led online cohorts in English test prep, data & AI, business & finance and digital careers. Tell us which one you want. The programmes with the most interest open first, and the people on the waitlist hear about them first.

See the planned programmes