Stage 01 — Mission Brief

The Engine:
Prompt Lab

Judge the prompt before you trust the output.

< 30 minutes Solo or team No AI tool needed Module 2 · after the Use-Case Finder

Six teammates wrote a prompt for the same campaign task. Some are briefs, some are wishes, and one would put your brand at risk. In this sprint, you read each prompt against the six parts of a good brief and the guardrails it carries, then pick the two you would actually send.

A prompt is not a magic phrase. It is a brief — and a model can only be as specific, accurate, and on-brand as the brief it receives.

C ContextO ObjectiveS StyleT ToneA AudienceR Response

Start the Sprint

Your mission

1Set the campaign task: offer, audience, channel, KPI, tone, a verified fact, and a do-not-say rule.
2Read six fictional prompts written by your teammates.
3Check each for context, objective, style, tone, audience, response — and guardrails.
4Select the two prompts you would actually send.
5Compare your call with the recommended shortlist.

Stage 02 — Setup

Set the Campaign Task

Describe the piece the team must produce. The simulation will write six fictional prompts for that task, the way different teammates would. All fields are required — and nothing you type leaves this page.

Stage 03 — Prompt Dossiers

Read the Prompts

Your task and six prompt dossiers are below. Read the prompts themselves before you look at the scores. A long, polished prompt is not automatically a good one: check what the model is told, what it is not told, and what it is being asked to invent.

Stage 04 — Prompt Selection

Choose What You Would Send

Pick the two prompts you would actually send to your approved AI tool — the ones most likely to return an accurate, on-brand draft a reviewer can approve. Use the Select to Send buttons on the dossiers above.

Stage 05 — Debrief & Feedback

How Did You Judge?

Field lesson

Practical reminder

Wrap-up

From Prompts to Briefs

Three things to carry back into your marketing work:

A prompt is a brief

Context, objective, style, tone, audience, response: leave one out and the model fills the gap with something generic.

Polish is not quality

The most impressive prompt on the board pasted personal data and asked for invented statistics. Read what the prompt asks for, not how it sounds.

Guardrails go in the prompt

Use only verified facts, flag what cannot be checked, respect the do-not-say rules — written into the prompt, not hoped for after it.

← Previous module: Use-Case Finder Build your own prompt in the toolkit →

The Engine:Prompt Lab