CO-STAR Workshop Exercise — MindfulAI™
MindfulAI™ Masterclass — CO-STAR Framework

The Information Brief
Workshop

Use the Teleserv call center scenario to work through each CO-STAR dimension. Then trade with a partner to find exactly where your brief fell short.

Scenario
Meet Teleserv.

Teleserv is a 320-agent inbound call center handling customer service for three telecom clients. Agents work live calls — billing disputes, cancellation threats, plan changes — under pressure to keep handle time under six minutes while hitting CSAT targets.

Leadership wants to deploy an AI assistant that surfaces real-time response scripts directly in the agent's browser during live calls. The AI needs to know what to say, how to say it, and what it is never allowed to promise — without a supervisor in the room.

Your task

Use CO-STAR to write the prompt that tells the AI how to respond when a customer threatens to cancel over a billing dispute.

In the next exercise, you'll use SPACE™ to define the system that governs everything the AI does before a single customer call begins.

Session timer 07:00
Step 1 of 7
C
CO-STAR Step 1
Context
3 min
From the Information Brief: "What is this for, and who is it for?" — Fill in confirmed facts and project context before you prompt. The AI cannot know what you haven't written down.

What is this project? What will the AI be working on? Include the format, the purpose, and any relevant background.

What is already decided? List only what is truly fixed — mark status: confirmed, assumed, or pending.

List actual sources — don't summarize them here. Name the document and the key fact it provides.

O
CO-STAR Step 2
Objective
2 min
From the Information Brief: "How will you know this was done well?" — Your success criteria IS your objective. If you can't state it clearly, the AI can't meet it.

One clear sentence. What should the AI produce? Avoid vague verbs like "help with" or "work on."

How will you know the output is good? Be specific — vague criteria produce vague outputs.

The most commonly skipped section. Flag what is NOT decided yet — this prevents wrong guesses.

S
CO-STAR Step 3
Style
2 min
From the Information Brief: "Voice & Style Constraints — Brand language, tone, things to avoid." Examples teach faster than adjectives.

Describe the writing style in concrete terms — not just "professional." Think about structure, sentence length, vocabulary level.

✓ An example I like
✗ An example I don't like
T
CO-STAR Step 4
Tone
1 min
From the Information Brief: Tone is not style. Style is structure; tone is emotional register. The same facts delivered with urgency vs. reassurance produce very different responses.

What should the reader feel? Choose 2–3 precise words — not just "professional."

Tone traps the AI often falls into for this kind of output.

A
CO-STAR Step 5
Audience
2 min
From the Information Brief: Audience determines vocabulary level, assumed knowledge, and what needs to be explained vs. taken for granted. The brief failed if your partner had to guess who would read this.

Specific role, not a broad group. What do they already know? What do they care about most?

Who else might read this, and does that change anything?

Legal, compliance, or brand requirements that must appear no matter what. (From the Information Brief.)

R
CO-STAR Step 6
Response
2 min
From the Information Brief: "Success Criteria — How will you know this was done well?" The response format is how success becomes visible. If you don't specify it, the AI guesses.

Exact structure you want back. Headings? Bullets? Table? Length?

Hard limits on length, file type, section count, reading level, etc.

Your assembled CO-STAR prompt
Fill in the dimensions above, then click "Assemble prompt" to see your CO-STAR prompt come together.
Partner exercise
Trade briefs. Find the gaps.
Exchange your assembled prompt with a partner. They attempt the task using only what's written — no follow-up questions allowed. Wherever they have to guess is exactly where the brief failed. Check every gap below.
  • They didn't know who the reader was
  • They guessed at the length or format
  • They assumed a tone you didn't intend
  • They used jargon the audience wouldn't know
  • They filled in a pending decision as if it were settled
  • They missed a non-negotiable requirement
  • They couldn't tell what "done well" meant
  • They produced the wrong output format
Debrief — what did the gaps teach you?
MindfulAI™ Masterclass Series · Right AI™ LLC · CO-STAR Framework Workshop
The quality of what AI produces is bounded by the quality of what it's given.