Tables vs the AI.
A workshop for a room of DPOs. Five or six tables play one case in stages, then hunt the AI's answer for mistakes. Nobody is scored individually, nobody writes a document, and the leaderboard does the rest.
- The room
- twenty to thirty DPOs at five or six tables, one device per table, a projector at the front
- The case
- one realistic scenario, revealed stage by stage; the whole room moves together
- Each stage
- two or three quick calls, one sentence, and a wager on whether the AI will get it right
- The catch round
- after each reveal, points for every genuine mistake or gap in the AI's answer, with the facilitator as referee
- Points
- a call 10 · a correct prediction 5 · a catch 20
- Nobody
- is scored on their own, or asked to draft anything
Projector · running totals (illustrative)
- Table Four60
- The Lawful Basis45
- Recital 4735
- Article 30 Club30
- Late Filers25
How a session runs
Tables name themselves. The leaderboard uses the names, which is half the fun.
A stage is revealed. The facilitator opens it from the projector and every table sees the same information at the same moment. The table argues it out; a spokesperson taps in its calls, its one sentence, and its wagers on the machine.
The AI's answer is shown. Tables hunt for what it missed or got wrong. The facilitator awards points against a pre-marked sheet, rules live on anything new, and reads out the best sentences. The facilitator's ruling is final, and that is a rule of the game.
Next stage. Four or five beats make a good game.
The final slide. How well the room predicted the machine: where it over-rated it, where it under-rated it. That is the workshop's real question, answered live in front of everyone.
What it needs
A room with tables and a projector, one device per table with printed answer cards as the backup, and a facilitator. The game runs on a separate copy of the bench with table accounts, no personal email addresses, wiped afterwards. Your real cases can be used, anonymised, or one of ours.