Prompting is one small part. A team gets more from four skills: describing the task clearly, checking where information came from, testing the result on awkward cases, and making the work usable by someone else.
Does everyone need to learn to build agents?
No. Most people need to explain what a tool should do and how they would check it. Start with a task the team already understands.
1. Brief the work.
Practise writing one task, the inputs it may use and the output you expect.
Add what must not happen, what might be missing and when to ask a person.
Practical check: Could someone else tell whether the task was done? A clear brief is more useful than a long prompt.
2. Question the information.
Practise tracing a statement back to an approved document or record.
Check its date, its scope, its definitions and whether it actually supports the statement.
Explain what can be concluded, and what is still unknown.
Practical check: Practise on public or fictional information first.
3. Test the help.
Practise with a normal case, a case with missing information and an ambiguous request.
Decide what to accept, correct, reject or pass to someone else, and keep a short reason.
Practical check: Looking polished is not a reason to accept it.
4. Make it usable by someone else.
Practise explaining what starts it, the steps, the checks and the fallback.
Try a handover. Ask a colleague to run a fictional example and say what they’d do if it failed. Note what was unclear and improve the instructions.
A team exercise.
Pick one task the team repeats. One person writes the brief, another tests it on three cases, a third runs it from the instructions alone. Compare notes on what each person had to guess.
Read more: AI Skills for HR Professionals: Where to Start · How to Show AI Skills in an HR Interview or Review
These exercises are not a certification. This is general guidance, not advice for your situation.