Serious.Games
AI Committee Under Pressure
ManagementFormationFree

AI Committee Under Pressure

Build a robust AI governance framework for L&D despite incidents, tensions, and trade-offs.

Duration · 90–120 min
Participants · 8–24
Level · Intermediate

In short. AI Committee Under Pressure is a serious game (management) for 8–24 participants, lasting 90–120 min. Created by serious-games.com. Use it for: prise de décision, gestion du changement, alignement.

Card by serious-games.com, updated 22 November 2026. Original game by serious-games.com.

Participants embody the AI governance committee of a fictional Training department that wants to deploy generative and analytical AI uses. In teams, they must transform vague risks into clear rules, balancing speed, compliance, educational quality, and learner trust. The game combines prioritisation, role negotiation, incident cards, and executive committee voting. At the end of the session, it produces an AI governance charter tested by real scenarios.

Walkthrough

  1. 1

    Game Framework and Mission

    10 min

    The facilitator presents the fictional organisation: an L&D service with 80 internal trainers wants to use AI to create content, personalise pathways, and analyse learning data. They say: "Your mission is to deliver a usable AI policy, not a perfect one, before the executive committee." Participants listen, ask questions, and discover the 4 scoring criteria: ethics, effectiveness, security, adoption. This establishes a realistic tension between ambition and risk management.

    Tip : Display the 4 scoring criteria from the start and remind that no criterion can be sacrificed without consequence.

  2. 2

    Formation of Committees and Roles

    10 min

    The facilitator forms teams of 4 to 6 and distributes a role card to each participant: L&D Manager, Data Lead, Ethics Officer, Technology Specialist, Intellectual Property Lawyer, Learner Representative. They say: "You defend your angle, but you only win if the overall governance holds." Each team receives 12 Resource tokens, an AI Charter sheet, and 6 Decision cards. The goal is to create structured debates rather than a hasty consensus.

    Tip : If there are fewer than 6 people per team, assign two roles to some players, but ask them to speak by announcing the role they are representing.

  3. 3

    AI Risk Market

    20 min

    The facilitator lays out 12 Risk cards face up: data breach, recommendation bias, pedagogical hallucination, unauthorised content reuse, lack of human validation, supplier opacity, tool dependency, poor user training, weak traceability, digital exclusion, control overload, prompt drift. They say: "Choose 5 risks to address now; the others remain in the shadows." Teams rate each risk from 1 to 5 in impact and probability, then select their priorities. This step forces prioritisation and reveals blind spots.

    Tip : Prohibit choosing only the easy risks: each team must retain at least one data risk, one ethical risk, and one quality risk.

  4. 4

    Construction of the AI Charter

    25 min

    The facilitator initiates a writing sprint with a constraint: each team must write a maximum of 8 rules on their AI Charter sheet. They say: "A good rule specifies who decides, when, with what evidence, and what limits." For each rule, the team spends 1 to 3 Resource tokens; 1 means light, 2 standard, 3 robust. Participants draft rules covering data, ethics, intellectual property, quality control, and user support. The limited number of rules prevents a catalogue and encourages formulating applicable safeguards.

    Tip : Provide a neutral example of wording: "All AI-generated content intended for learners is reviewed by a subject matter expert before publication."

  5. 5

    Incident Cards and Trade-offs

    20 min

    The facilitator gives each team 3 Incident cards drawn randomly from the untreated or partially treated risks. They say: "Your charter has just encountered reality: apply it or modify it, but each modification costs 1 token." Teams have 5 minutes per incident to respond: immediate decision, responsible person, evidence to retain, planned communication. If their charter already covered the incident, they earn 2 points; otherwise, they can correct it but lose 1 point for each missing token. This mechanism tests robustness without turning the exercise into simple writing.

    Tip : Play the role of a hurried executive: regularly ask, "Who signs? Who verifies? What do you tell the learners?"

  6. 6

    Simulated Executive Committee

    20 min

    The facilitator organises presentations: each team has 3 minutes to defend their charter and 2 minutes of questions from another team. They say: "You do not present everything; you prove that your governance is clear, realistic, and defensible." Auditors each assign 1 vote in two different categories among ethics, effectiveness, security, adoption. The facilitator adds the points earned during the incidents and announces the total score, without designating the charter as the sole truth. This step values clarity of argumentation and cross-learning.

    Tip : Strictly limit the time with a visible timer; the constraint reveals the maturity of the trade-offs.

  7. 7

    Common Consolidation

    15 min

    The facilitator displays the best rules heard and asks the group to choose 6 for a common charter. They say: "We are moving from the game to usage: which rules would you keep tomorrow in your context?" Participants vote with 3 stickers or 3 digital votes, then reformulate the winning rules to make them actionable. The facilitator concludes by linking each rule to learnings about responsibilities, evidence, limits, and adoption. This step transforms competition into exploitable collective capital.

    Tip : Ask for an operational reformulation before validating a rule: if it does not specify who does what, it goes back for discussion.

Variants

  • Short version 60 minutes: reduce to 8 Risk cards, request a maximum of 5 charter rules, and play only 2 Incident cards per team.
  • Advanced version: add a budget constraint. Each team receives only 8 Resource tokens and must explicitly accept 2 residual risks with justification.
  • Distance version: use a collaborative board with zones by team, Risk cards on post-its, votes by digital dots, and presentations in breakout rooms followed by plenary feedback.
  • Onboarding version: replace expert roles with simple business roles and provide a list of short definitions: personal data, bias, human validation, intellectual property, traceability.

Debrief guide

  • Which risk did you underestimate at the start, and what made you see it differently?
  • What tensions did you feel between rapid innovation, security, educational quality, and learner trust?
  • Which rules were truly applicable, and which resembled good intentions?
  • How did the roles influence your decisions and disagreements?
  • What elements of evidence or traceability are essential to govern AI without blocking usage?
  • What would need to be adapted to use this charter in your real organisation?
  • What concrete first action could you launch in the next 30 days?

Facilitator reviews

No review yet. Have you run this game? Your feedback will help the next facilitators.

Same objectives or same category, to vary or follow up.