Twelve-Ticket Jury
Put any AI number before twelve real tickets, and quote it honestly.
The job
In under two minutes, in front of the sceptic, checks that an AI-produced percentage matches the tickets behind it, shows how precisely it can be quoted, and stops unchecked figures leaving in the brief without an "unverified" stamp.
The moment
The AI says 38% of housing complaints are about eligibility. The chief of staff puts that figure to a jury of twelve real complaint tickets.
The twelve cards flip one by one: ten really are about eligibility, and two are actually about payment delays.
The AI put all 12 in this group, but only 10 belong. The label is accurate somewhere between 55 and 95% of the time, so the safe line is "about a third", not 38%.
The chief of staff rewrites the minister's line on the spot, and the brief marks the figure as jury-checked instead of unverified.
What it does
- Rule a card In / Out / Can't tell
- Stamp the figure and accept the honest wording
- Send Out cards to the AI category owner
- Open the slice or the Out cards
- Share a verdict
What you see
Courtroom jury box: twelve face-down ticket cards flipped one by one, ending in a stamped verdict on the original figure
What it moves
The share of AI-produced figures in the weekly brief that carry a jury check (target 100%), and how accurate each AI category is, in points, with its range.
Built for
- Executives & leadership
- Analysts & data
- Regulators & auditors
- CX leaders & VoC
