Look familiar? “Write a briefing note on X.” Most of us will recognize this as a type of prompt we use regularly.

The result? Text that is often generic, lacking the right tone or voice. You usually end up spending more time editing the AI’s response than if you had written everything from scratch in the first place.

CREATER can help you change that.

Adapted from CREATE – developed by Dave Birss, CREATER has an extra “R” for a reason: iteration. Your first draft is never your final draft. Sometimes the two don’t even look alike. For the best results, you need to review – revise – repeat until you get what you want.

Here’s how CREATER works:

C — Character

Define the role you want the AI to play. This will help shape the tone, judgement and level of detail in the AI’s response. A policy analyst, a communications advisor, and a director all write differently. Tell the AI which role to take: don’t let it guess.

R — Request

Be specific about what you need. What are you asking for? Who needs this information? Why does this information matter? The clearer the request, the less time you spend fixing the answer.

E — Examples

Show the AI what “good” looks like. A single sentence, an approved document, or a link to a style guide can anchor tone and quality surprisingly well. Remember: no AI product can read your mind (at least not yet).

A — Adjustments

Add guardrails and limits for the AI to use. What should it avoid? What is out of scope? Where should it focus its efforts? This helps prevent drift, jargon and the occasional hallucination.

T — Type of Output

Define the format. Headings, structure, word count—spell it out. If you already have a template or an approved example available to the public, link to it or upload a copy into the AI.

E — Extras

Push for quality. Ask questions like: What assumptions were made? What’s missing? What’s weak? What isn’t needed or has been repeated? Have you been brutally honest with me?

R — Review, revise, repeat

Don’t stop at the first draft. Tell the AI what to change: tighten it, reorder it, cut it down. You can also add feedback from earlier outputs. Iteration is where the real value shows up.

Before / After Example

Before:
“Write a report on service delivery challenges.”

After:
You are a public sector analyst writing for decision-makers.

Write a 300–400 word executive-focused report on key service delivery challenges.

Use plain language (for example, “Delays may affect service standards”). Keep the tone neutral and operational.

Structure your response as:

  • Overview
  • Key Challenges
  • Impacts
  • Suggested Actions

Only include information that is:

  • explicitly provided in the prompt, or
  • generally accepted, non-controversial public sector knowledge

If you are uncertain or making assumptions, clearly signal this (for example: “This analysis assumes…” or “Information on this point is limited”). If key inputs are missing, ask clarifying questions instead of proceeding. If insufficient information is available, provide a structured placeholder response. Be brutally honest with me concerning gaps, misconceptions and possible errors.

Do not invent statistics, programs, departments, jurisdictions, or specific data points.

Avoid speculation presented as fact.

Identify gaps in available information and note where additional input would be needed.

Prioritize clarity, accuracy, and actionable insights over completeness.

Revise your output for concision and plain language and design techniques.

When can you use CREATER?

You can use CREATER for:

  • Complex tasks such as analysis, drafting, or summarising multiple sources
  • Work going to senior management, where “good enough” is not good enough
  • Situations where you want consistent, reusable outputs across a team

A well-built CREATER prompt can be saved and reused. That makes it less of a one-off product and more of a team habit.

We public servants like forms, and we are used to them. So here is a quick CREATER template:

C: You are…
R: Your task is to…
E: Use this style or example…
A: Keep in mind…
T: Format as…
E: Check for…
R: Revise to improve…

Key Takeaway

CREATER helps make expectations explicit and improves AI outputs through structured iteration. It does not replace professional judgement—it reduces rework and makes first drafts significantly more usable.