The idea factory: surfacing the right AI use cases with your teams
The day I arrive on a site, I don’t ask to see the software. I ask to spend the day in production, at the foot of the machines, with the people who keep them running. That is where the good AI use cases live, and nowhere else. Not in a meeting room, not in a consultancy slide deck: on the floor, in the words of those who deal with the irritants every single day.
We start with what works
Before hunting for problems, I ask the teams what works well. It can be surprising, but it’s deliberate. First because it’s true: a plant that runs is, above all, know-how that holds together. Second because it changes the climate. When you open by pointing at what’s broken, people close up, they feel judged. When you start by acknowledging what they master, they open up, and the real irritants surface naturally, without anyone having to pry them out.
Let the irritants come from those who live them
An irritant isn’t an improvement idea I project from the outside. It’s what annoys the operator at the start of a shift, the re-entry they redo for the third time, the information they wait half a day for, the file no one can find. A consultant who shows up can’t invent those things. They come out across the day, between two machines, once trust is in place. My job isn’t to have the ideas for them: it’s to ask the right questions and actually listen.
Score, sort, let it settle
In the evening, I synthesize. Every idea that surfaced, I score on two simple axes: impact (what changes if we solve it) and feasibility (can we do it fast, or is it a major build). That crossing naturally separates two families. Quick wins: high impact, doable in a few weeks, often nothing more than data already captured that we finally make usable. And structural projects: the ones that take time, a budget, real change management.
Then I let it settle. I don’t hand over a list the next morning. A few days of distance keep you from mistaking the irritant of the moment for the real underlying issue, and let you rank things with a cool head. An idea that survives a week of settling is almost always a good idea.
Where AI has no business being
Let’s be honest: for some irritants, the best answer isn’t AI. Sometimes it’s an organizational fix, a rule to clarify, a ten-euro sensor, or simply two departments made to talk to each other. If I bolt AI onto that, I add complexity for nothing and I disappoint. Part of my job is precisely to say when a use case doesn’t deserve AI. It costs a recommendation in the short term, it builds trust over the long term.
The real differentiator of this method fits in one sentence: the best ideas don’t come from a consultancy, they come from your teams. My work is to surface them, score them, and turn them into projects you can actually carry. AI with you, not in your place.
For the full picture, read the guide AI in industry. See also: The data that sleeps. Wondering where to start? Gauge your AI maturity in 2 minutes, or let’s talk for 20 minutes.