How to write better AI prompts: a practical guide with examples

You do not need to be technical to write good prompts. You need to be specific. The same skill that makes a good brief for a coworker makes a good prompt for an AI: say who should do it, what you want, what they need to know, and what "done" looks like. Everything below is just a structured way to do that.

The four-part prompt framework

Almost every strong prompt has four ingredients. You do not always need all four, but reaching for them in order fixes the majority of weak results.

1. Role and goal. Tell the AI who to be and what outcome you want. "Act as a hiring manager. Help me write a job description for a senior nurse."
2. Context. Give it the facts it cannot guess: the audience, the situation, the constraints. "This is for a 200-bed hospital, night shift, and we struggle to attract experienced candidates."
3. Format. Describe the shape of the answer. "Return a 150-word posting with three bullet points on what makes the role appealing."
4. Refinement. After you read the answer, point at what is wrong and ask for a targeted fix instead of starting over.

Before and after

Here is what the framework does to a real request.

The second prompt did not require more skill — just more of the information that was already in your head. The AI cannot read your situation. Your job is to hand it over.

Five habits that raise quality fast

1. Say what you do not want

Negative constraints are powerful. "No jargon," "do not use the word 'leverage,'" "avoid a hard sell" all steer the output more than another positive instruction would.

2. Give an example of what good looks like

If you have a sentence, a past email, or a format you like, paste it and say "match this style." Showing one example is worth a paragraph of description.

3. Ask for the reasoning when it matters

For decisions and analysis, add "explain your reasoning step by step before giving the answer." For creative or simple tasks, skip it — it just adds noise.

4. Break big asks into steps

Instead of "plan and write my entire launch campaign," ask for the plan first, react to it, then ask for each piece. You will catch wrong turns early instead of regenerating a huge output.

5. Refine, do not restart

When an answer is 80% right, do not rewrite the prompt from scratch. Say exactly what to change: "Keep everything, but make the second paragraph shorter and remove the apology." Iteration beats regeneration.

Common mistakes to avoid

• Being polite instead of clear. "Could you maybe help with something?" wastes a turn. State the task directly.
• Assuming shared context. The AI does not know your company, your last conversation, or your deadline unless you tell it.
• Asking for everything at once. A single mega-prompt usually produces a mediocre answer to each part. Sequence instead.
• Accepting the first draft. The first answer is a starting point. The value is in the second and third pass.
• Not verifying facts. AI can state wrong things confidently. Check names, numbers, dates, and quotes before you rely on them.

A copy-and-paste template

When you are stuck, fill in this skeleton:

Why this skill matters more over time

Prompting is becoming a baseline work skill, like writing a clear email or a good search query. The people who get the most out of AI are rarely the most technical — they are the ones who describe what they want precisely and iterate quickly. The framework above works whether you are drafting a document, analyzing data, or getting help in a live conversation.

That last case is where real-time tools come in. When you are in a meeting or an interview, you do not have time to write a paragraph-long prompt. Tools like LiveCue apply the same idea automatically — turning what is being said into short, useful cues in the moment — so you get the benefit of a good prompt without stopping to type one.

Bottom line

Good prompting is not a trick. It is clarity: role, context, format, refinement. Give the AI the information it cannot guess, show it what "good" looks like, and treat the first answer as a draft. Do that consistently and you will get better results from every AI tool you touch.