The question I get most: “We want to do AI — where do we start?” Wrong question. The right one: where are you losing time today?
Follow the time, not the hype
AI is only worth it where it returns time or quality. So we start by observing. Concretely:
- Which task comes back every week and makes everyone groan?
- Which piece of info takes ten minutes to find when it should take one?
- Where do people copy data from one tool to another, by hand?
Those three questions beat any seminar on generative AI.
Three workstreams, not thirty
An SME can’t run everything at once. We pick three, on two criteria: business pain (does it really hurt?) and team demand (do they want it?). What ticks both goes first.
Measure from the first pilot
Before / after. Time per case, error rate, satisfaction. If you don’t measure, you’ll never know if it works — and you won’t convince the next team.
That’s it. No magic. A serious AI audit is this: look at the reality on the ground, prioritise, measure. The rest is noise.