Logins are not adoption.

A login is the easiest thing to count and the least useful. Real adoption is someone reaching for AI on a Tuesday because it is genuinely the faster way to do the job, which is a smaller and far more honest number.

A managing director told me proudly that 90% of his team had "started using AI". I asked what that number actually measured. It turned out to be the share of people who had logged into the tool at least once since it was rolled out. By that standard, I have adopted every gym I have ever joined.

Logins are the easiest thing to count and the least useful. They tell you procurement worked, not that anything changed.

What adoption actually looks like

Real adoption is someone reaching for AI on a Tuesday, unprompted, because it is genuinely the faster way to do the thing in front of them. It shows up as a change in how the work gets done, not as a line on a licence report. If you want a number, the honest one is harder to get: what share of your team used it this week on something that actually mattered, and for what.

That is a smaller figure than the login count, often much smaller. It is also the only one worth tracking, because it moves when capability improves and stalls when it does not.

Why the vanity metric is dangerous

The trouble with the login number is not just that it flatters. It is that it tells you to stop. If 90% have "adopted", the rollout looks finished, the budget gets closed, and the actual work of building capability never starts. Meanwhile the people who logged in once, found it awkward, and went back to their old way are quietly counted as wins.

Measure the behaviour you want

Pick two or three real tasks per team and look at whether AI is genuinely part of how they get done now. Ask people what they have stopped doing the slow way. Watch for the workflows that have changed shape. It is more effort than reading a usage dashboard, but it measures the thing you actually care about, and it shows you where the capability gap still is.