Where AI has no business being: industrial cases where the real problem is elsewhere
I’d rather say it plainly, because it’s what sets me apart from the dream-sellers: AI is not the answer to everything. Knowing it also means knowing where it has no business being. Here are the four cases where I systematically pump the brakes.
The technology looking for a problem
The most common trap. You saw a dazzling demo, you want to “do AI,” and then you look for where to put it. That’s the reverse of the right approach. You start from an irritant the floor already raises, not from a tool in search of a use. A solution without a problem is a gadget: it impresses in a meeting and dies in three weeks.
The hallucination you didn’t frame
An AI can invent an answer with full confidence. Handing it a decision on a vague question, without verifiable sources, is taking a needless risk. Until it’s framed on a restricted, controllable dataset, with clickable sources, you don’t trust it to decide. You use it to rough out and cross-check; validation stays human.
The ritual no one keeps alive
You build a fine tool, and it runs out of steam after a few weeks because no one keeps it alive. A tool no one animates dies, exactly like a dashboard no one opens anymore. If you haven’t planned who carries the usage over time, don’t launch the tool: you’ll just add one more disappointment to the list.
The managerial problem disguised as a technical one
The most important one. When a team malfunctions, when information doesn’t flow, when a decision drags on for weeks, no AI will fix that for you. Technology can even serve as a smokescreen to avoid dealing with the real subject, which is human and organizational. There, my job is first to tell you so.
The rule that sums it up
My compass is three words: verify, validate, valorize. If an idea doesn’t pass those three filters, it isn’t ready, and sometimes the right recommendation is to do no AI at all. It’s precisely that honesty that makes the remaining projects solid.
For the full picture, read the guide AI in industry. See also: Inspection robots in the plant. Wondering where to start? Gauge your AI maturity in 2 minutes, or let’s talk for 20 minutes.