Start with a real situation.
Bring 3–8 people, one facilitator, a timer, and somewhere to write. Choose one recent pull request, sprint, or AI adoption decision. You do not need to have read the books or played a tabletop game.
Use the matching deck and start with the two examples below. Read the front and back together; the prompts in this guide help turn recognition into action. Without a deck, open the linked bestiary entries and try the conversation first.
Name patterns, not colleagues. Say “we lost the reasoning for this module,” then point to the evidence. Give people room to disagree or pass. If the name does not fit, use plain language. A card is a conversation starter, not a verdict.
Volume I / Developers, reviewers & tech leads
A 15-minute AI coding retro.
Bring one recent AI-assisted change. The question: “Where did speed make the scope, reasoning, or ownership less clear?”
- 0–3 minutes: Put the work on the table. State the intended result. Show a pull request, acceptance criterion, or decision note. Agree on what happened before naming a pattern.
- 3–7 minutes: Pick the card that helps. Read the Scope Creep Kraken and Phantom Intern together. Each person names one observation from the work. Choose the better fit; no match is also a useful answer.
- 7–11 minutes: Test the diagnosis. Ask one prompt below. What evidence supports it? What would make the team change its mind? Keep assumptions marked unknown.
- 11–15 minutes: Choose one countermove. A countermove is a small, specific response to the pattern. Write the action, a human owner, the evidence you expect, and a follow-up date.
Scope Creep Kraken
Recognize it: Helpful extras became commitments before the team chose to maintain them.
Ask: “Which part of this change serves the original goal? Which addition needs its own decision?”
Try: Restate the acceptance criteria. Move one extra into separately approved work with an owner, a reason, and a maintenance estimate.
Read next: Volume I, Chapter 8, “Aberrations: The Scope Creep Kraken and the Temporal Tangle Worm”; Chapter 12, “Diagnosis and Intervention.”
Phantom Intern
Recognize it: The code works, but the team cannot explain the decisions or confidently change it.
Ask: “If the author were away tomorrow, where would we find why this works this way?”
Try: Have the author record the important decision, alternatives, and assumptions alongside the code. Ask another maintainer to trace one real input and explain the result.
Read next: Volume I, Chapter 10, “Spirits: The Phantom Intern and the Probability Pixie”; Chapter 12, “Diagnosis and Intervention.”
Explore the Volume I book for the deeper diagnosis and interventions. Get its matching book + deck kit to bring the conversation to your next review.
43 illustrated cards, 2.5 × 3.5 in. Printed deck ships in days; this guide works with it out of the box.
Volume II / Managers, team leads & AI adoption teams
A 20-minute AI adoption discussion.
Bring one proposed purchase, pilot expansion, or AI workflow. Include its owner and the evidence you have. The question: “What are we choosing, and what would justify that choice?”
- 0–4 minutes: Name the decision. Write the problem, who experiences it, and one useful outcome. Describe the proposed change and the decision date.
- 4–9 minutes: Pick the card that helps. Read the Vendor Bard and Cost Singularity together. Discuss the claims or cost assumptions in this decision. Describe the pressure on the decision without assigning a character to a person.
- 9–15 minutes: Ask for evidence. Use one prompt below. List what you know, what is unknown, and the smallest test or forecast that would close the consequential gap.
- 15–20 minutes: Choose the next move. Record whether to continue, bound the pilot, or pause expansion while the owner gathers evidence. Set a follow-up date and the condition that would justify expansion.
Vendor Bard
Recognize it: The promise of joining the future is doing more work than the evidence for your actual problem.
Ask: “Which of our specific problems did the demo solve? What does this product not do well?”
Try: Write three problems before evaluating the tool. Have an owner test one ordinary case and one difficult case from your workflow, and record the limitations.
Read next: Volume II, Chapter 5, “Vendor NPCs and Anti-Patterns.”
Cost Singularity
Recognize it: A cheap pilot is becoming a popular workflow before anyone has priced success.
Ask: “What does this cost if everyone loves it—including retries, supporting services, and human review?”
Try: Forecast the hoped-for adoption. Name a cost owner, set a usage boundary and a review trigger, and check what happens at that boundary before expanding.
Read next: Volume II, Chapter 8, “The Cost Singularity.”
Explore the Volume II book for the organizational cast and countermoves. Get its matching book + deck kit for your next adoption review.
The organization on cards: executives, vendors, FinOps, and the monsters between teams.
Leave with a record.
Copy this into a team note, ticket, or decision record. Choose one move the team can verify; a list of concerns is only the beginning.
Work or decision: ______________________________
Pattern + evidence: _____________________________
One countermove: ______________________________
Human owner: _________________________________
Check back on: _________________________________
We will know it helped when: _____________________
At the follow-up: Did the action happen? What changed in the work? Keep, adjust, or drop the move based on the evidence. For an adoption decision, record whether the evidence now justifies expansion.
Example: “The export PR picked up an unplanned integration. Maya will separate it from the current change before Friday’s review. We will check that the remaining diff meets the original acceptance criteria.”
The card starts it. The book takes it further.
Use the card to recognize a situation, the question to make it concrete, and the matching chapter to understand the causes and choose a response. Read the relevant entry after the session; bring one adjustment back to the next meeting.
These sessions build on the AI Code Review Field Kit and AI Team Field Kit. Use those checklists when you want to examine a whole pull request or rollout.