The field guide
What to ask when somebody brings you an AI proposal, a business case, or a system that is already running.
These are the principles the simulations turn on, collected in one place. Each is a rule you can apply without knowing anything about how a system works, which is the whole point: none of the strong answers in any scenario requires technical knowledge, and several of the weakest ones sound the most technically impressive.
Six rules per competency are set out below. Underneath each is the full set — currently a hundred and twenty — for looking things up rather than reading through.
Read this after playing, if you can. The scenarios work partly because you fall for something, and knowing the shape of the traps in advance takes a little of that away. It will not spoil anything outright — no rule here tells you which option to pick — but the order is worth having if you have the choice.
There is also a printable handbook — the twenty-four core rules, a worked decision and the framework, in something you can save as a PDF and hand to somebody. This page is the complete and current version; the handbook is the part that stands still.
Taking it back to your own organisation
The situations in the simulations are set somewhere else on purpose. The names, the numbers and the industry belong to another organisation; what carries across is the decision. Three questions worth putting to your own:
- What have we already handed over without deciding to? Not what the policy permits, but what the work actually does now, and who would notice if it changed again.
- Which number in front of us has nobody asked the denominator of? Pick the figure your next decision rests on and find out what population it averages over, and what it leaves out.
- Who is answerable on day four hundred? Name them. If the answer takes more than a minute to establish, that is the finding.