Most AI pilots don’t fail on the technology. They fail because nobody decided beforehand what “success” means.

The pattern

It starts with a demo. Someone sees a tool that impresses, a trial period follows, a few colleagues join in. After three months the enthusiasm is gone and the licence isn’t. Nobody can say whether it worked, because nobody had said what working meant.

That is not a shortage of technology. It is a shortage of a decision.

What a decision asks for

A decision about AI is an ordinary decision. It asks for the same three things as any investment:

  • a question in one sentence: what do we want to be different, for whom, by when
  • an owner: who decides, and who carries it afterwards
  • a criterion: how will we see it has worked, and what do we do if it hasn’t

Without those three, a pilot is an experiment without an end. With them, it is a project with a go/no-go.

Why the tool comes later

Choosing a tool before the decision is taken reverses the order. Then the tool decides which question you ask. Choose the decision first, and it often turns out that the tool you already have is enough, or that no tool is needed at all but an agreement.

That is why every assignment at March83 starts with an intake and a quick scan, not with a demo. The result is a decision memo your management can read for itself. Only after that do we talk about tools.

What you can do tomorrow

Take the AI pilot currently running in your organisation. Write down in one sentence which decision it is supposed to support. If you can’t, you know what has to happen first.