We are use case led. Almost every organisation says this about AI now, and it sounds like discipline. Increasingly I think it’s where discipline goes to hide.
Here’s the pattern I’ve watched, in our organisation and others. The call goes out for AI use cases. Lists bloom. Workshops harvest ideas by the dozen, the register swells, and the number itself becomes a proud metric. It feels like momentum. It’s actually an arms race of ambition, and most entries share one fatal property: nobody who’ll own the outcome put them forward.
A use case proposed on someone else’s behalf is a wish, not a commitment. And the gap between the two only becomes visible after the pilot, because pilots flatter everything. A pilot is controlled: the data gets curated, the process gets white-gloved, the project team hovers over it like anxious parents. Of course it works. Then comes operations, where data is messy, exceptions are constant, the project team has moved on, and the people expected to run the thing were never the ones who asked for it. That’s where use cases go to die, quietly, in the gap between demonstration and ownership.
So we changed the qualifying question. Not “is this a good use case?” but “who owns this outcome, and are they committed?” Committed means specific things: they’ll make the decisions, ready their process and data and people, put their name on the governance, and carry it into operations after the project team leaves. Size doesn’t matter, an agent doing something small with a committed owner beats a grand ambition with none.
Our register got shorter. It also got real.
The seduction of the long list is that identification feels like progress and costs nothing. Ownership costs something, which is exactly why it predicts success.
Ask whoever runs your AI portfolio two questions. How many use cases have we identified? And how many have a named owner committed to running the result? The distance between those numbers is your smoke-and-mirrors quotient.