Two weeks to get the appointment. Ten minutes with the doctor. Then the pharmacy says the drug isn’t in stock. Start again.
And here’s the strange part. Nobody in that chain did anything wrong. The doctor did their job, the pharmacist did theirs. It feels like bad luck, but it isn’t. It’s two systems that were never able to see each other, and a process nobody ever stood back and mapped.
This is what I keep meeting as companies rush towards AI. They want to bolt it on top of processes exactly like that one, and the result is predictable.
You can’t automate a mess. You just get a faster mess.
I recently watched a room map a real patient journey through a GP surgery, end to end. The thorn appeared almost at once: doctors prescribing medicines with no view of whether the pharmacy actually had them. Rework, delay, frustrated patients. No AI on earth fixes that until someone maps it and sees it.
And this is the bit most people miss. The skill isn’t mapping the work you do. It’s asking which of that work creates value, and which only exists because of old tools and old habits. In Hampshire, they scrapped rigid appointment slots for blood tests and X-rays, moved to a choice-based model, and wiped out both waiting times and no-shows. They didn’t speed the old process up.
They deleted it.
Before AI can lift anything, your data has to be visible across your systems in real time. Most companies discover their real thorn isn’t the AI at all. It’s that their own data can’t be seen.
So the question isn’t “are we using AI yet”. It’s “have we earned the right to”.
Map the value first. Then automate.