Why most prompts disappoint
A vague request gets a vague answer. Managers get better results by giving an AI assistant the same briefing they would give a new colleague: who they are, what to do, the background, the format and the limits.
The five parts
| Part | What to include | Example |
|---|---|---|
| 1. Role | Who the assistant should act as | You are an experienced operations manager |
| 2. Task | The single job to do | Summarise these notes into decisions and actions |
| 3. Context | Audience, background, constraints | For a leadership review; the team is behind on a vendor delivery |
| 4. Format | How you want the output | A table with owner, action and date |
| 5. Limits and checks | What to avoid and how to verify | Do not invent dates; mark anything uncertain |
Five templates to adapt
| Use | Skeleton |
|---|---|
| Meeting notes to actions | Act as a project coordinator. Turn the notes below into decisions, actions, owners and due dates. Flag anything unclear. [paste notes] |
| Email reply | Act as a clear, courteous manager. Draft a reply to the message below that declines the request but offers an alternative. Keep it under 120 words. [paste message] |
| Options analysis | Act as a business analyst. Compare options A, B and C for [goal] on cost, risk and time. Present a table and a recommendation with reasoning. |
| Policy first draft | Act as an HR policy writer. Draft a one-page policy on [topic] for a company of [size]. Use plain language and list open questions for legal review. |
| Data question | Act as a data analyst. Given this table, explain the three most important patterns and one limitation of the data. [paste table] |
Habits that make it safer
- Do not paste confidential or personal data unless your company's policy allows it
- Treat outputs as drafts: check facts, numbers and names before use
- Ask the assistant to list assumptions and uncertainties
- Keep a human accountable for anything sent or decided
- Save prompts that work so your team can reuse them
Go further: how to choose an AI course, AI for HR professionals and AI governance for managers.
Frequently Asked Questions
What makes a good prompt?
A clear role, a single task, enough context, a specified format and limits or checks. Briefing an AI like a new colleague works well.
Can I share company documents with an AI tool?
Only if your organisation's policy and the tool's terms allow it. When in doubt, remove confidential and personal details.
Why does the AI make things up?
These systems can produce plausible but wrong statements. Verify facts, ask for assumptions and keep human review in place.
Do I need technical skills to prompt well?
No. Clear writing and a habit of checking outputs matter far more than technical knowledge.