AI can organise evidence and support a decision, but it cannot take responsibility for what happens to the person affected. Important decisions need genuine human oversight from somebody with the information, time and authority to challenge or reject the AI’s recommendation.

Not every use of AI carries the same weight. The matrix helps you see the difference.

The time to agree when you will stop is before everybody becomes determined to carry on.

Someone approved the money. But who takes responsibility when something goes wrong?

Saving time sounds impressive until you count the checking, correcting and extra work.

Before making a process faster, check whether half of it needs to happen at all.

A job title tells you very little about the different kinds of work happening underneath it.

Sometimes the most sensible thing you can do is leave a perfectly decent process alone.

The question that should come before every AI decision, but is often the first one skipped.