Every week another tool arrives that “changes everything”. The questions you have to ask don’t change.
1. Which decision should this support?
Not “what can the tool do” but “what will we do differently”. A tool that changes nothing about a decision or an action is a subscription. Write the decision down in one sentence. If you can’t, it is too early for a tool.
2. Who takes it, and by when?
An AI application without an owner dies in the trial period. The owner is not the enthusiastic colleague who gave the demo, but the person who will still be responsible for it in a year. And every decision has a date: without a date there is no go/no-go, only postponement.
3. What happens on a no-go?
This is the question that rarely gets asked. If the pilot doesn’t deliver what you hoped: does the licence stop, is the data deleted, do the staff know they may go back to the old way of working? A no-go that costs nothing is a no-go you dare to take. And that is exactly what a good pilot has to make possible.
And only then the tool
Whoever can answer the three questions has half of a decision memo. The other half, value, risk and feasibility, is the work of a quick scan. After that, choosing a tool is usually simple, because the question is sharp.