← AI Academy
Employees · 7 min read

Prompting for work: roles, examples, structure

Five techniques that turn average answers into work you can actually use.

1. Give a role and an audience

“You are an experienced HR specialist. Write for new employees without a technical background.” The role sets expertise and tone; the audience sets the level. Together they remove most generic answers.

2. Show an example

Paste one good example of what you want — a previous email, a table row, a product description — and say “in the same style”. One example often beats a paragraph of instructions.

3. Separate instructions from material

Put the text to work on between clear markers, e.g. “Text: ### … ###”. The model then knows what to process and what to do with it — and is less confused by long documents.

4. Ask for a structure

Tables, numbered steps, fixed headings or even CSV that you paste into Excel. “Return a table with columns: task, owner, deadline, risk.” Structured output is easier to check and reuse.

5. Make it check itself

End with: “Before answering, list your assumptions. After answering, check the result against my requirements and fix what doesn’t fit.” Save your best prompts in a shared document so the whole team benefits.

Key takeaways
  • Role + audience + example + structure.
  • Keep instructions and material separate.
  • Build a team library of prompts that work.
Try it

Fill in a ready-made work prompt and open it in ChatGPT, Claude or Le Chat.

Open the prompt library →

Check yourself

  1. Which addition most improves a prompt for a product description?
  2. Why ask for a table or CSV?
  3. What does “list your assumptions first” help with?

Tip: getting at least 2 answers right in the quiz above completes the lesson automatically.

This lesson is part of:AI at Work

Last reviewed: September 2026