How do you measure AI capability?

You measure AI capability by looking at real work, not quizzes or course completion. The honest signals are whether a person can improve an actual task with AI, whether the gain is verified rather than self-reported, and whether that ability is growing over time.

Attendance and test scores are easy to collect and weak as evidence. They say someone was present, not that their work changed. Capability shows up in output, so that is where it should be measured.

The strongest measure is a before and after on a real task, confirmed by someone other than the author, and tracked over weeks. That gives a manager a picture of capability they can actually trust and act on.