Practice a customer conversation with AI
Run a bounded one-question-at-a-time simulation and critique one behavior at a time.
Role-play transcript and coaching note
Finish with an artifact another professional can review, challenge, and act on.
When to use this
Run a bounded one-question-at-a-time simulation and critique one behavior at a time. Use it when the decision matters enough that source authority, uncertainty, and human review must stay visible.
Match the depth to the consequence
Quick
Produce the minimum useful role-play transcript and coaching note from the controlling sources and flag gaps.
Working
Compare sources, classify evidence, produce the role-play transcript and coaching note, and run the task-specific checks.
Thorough
Add a countercase, conflict review, stakeholder challenge, and explicit validation plan before action.
Bring only what the task needs
- The decision this work must support
- Current, approved source material
- Known constraints and open questions
Boundary: Approved customer information in an employer-approved environment; otherwise use public or fictional material.
Do not add credentials, unnecessary personal information, private security details, unapproved pricing, contracts, or customer material.
Work sequence
- State the decision the role-play transcript and coaching note must support.
- Select the controlling sources and record their date, owner, and authority.
- Separate customer evidence from seller interpretation before asking AI to synthesize.
- Build the role-play transcript and coaching note and label Known, Inferred, Assumed, and Unknown.
- Challenge the result with contrary evidence and complete the verification checks.
Practice a customer conversation with AI
GOAL Rehearse a difficult customer conversation without treating simulation as customer intelligence. CONTEXT The account, audience, decision, and current sales stage are [CONTEXT]. CONTROLLING SOURCES Use only [APPROVED SOURCES]. Identify the source behind every material fact. TASK Act as the selected fictional stakeholder, ask one question at a time, stay within the supplied evidence, and provide a separate critique after the user ends the role-play. CONSTRAINTS Do not invent customer intent, authority, dates, budgets, pricing, commitments, product capabilities, legal conclusions, or missing facts. REQUIRED OUTPUT Role-play transcript and coaching note UNCERTAINTY BEHAVIOR Separate Known, Inferred, Assumed, and Unknown. Show conflicts instead of resolving them without evidence. Turn important unknowns into questions. VERIFICATION REQUEST List the claims, calculations, names, dates, owners, commitments, and source conflicts a human must verify before acting. Construct the strongest evidence-based case against the proposed conclusion.
Northstar example
A fictional Northstar security leader challenges dependency and migration risk. The exercise tests listening and clarification, not prediction.
Task-specific checks
Why this works
The sequence starts with the decision and source authority, so fluent output cannot quietly replace evidence. The countercase turns uncertainty into focused customer questions.
A weak AI request
“Create a polished role-play transcript and coaching note and make it persuasive.”
This request has no controlling sources, evidence boundary, uncertainty behavior, or verification step. It rewards confident prose instead of dependable work.
Practice safely
Practice the vendor-neutral workshop opening with the fictional Northstar security leader.
Open Northstar exercises