The first thing the AI dashboard showed was who had no licence

When I started measuring AI adoption, I wanted to see who used it most, where the median sat, how usage was distributed, whether it was trending, and whether it clustered. The first thing the report showed was simpler: people who could not use AI at all, because nobody had given them access.

A usage report cannot show this. Usage data lists the people who used something. Someone with no licence appears nowhere on it, so absence stays invisible until you join usage to the directory and ask who is missing from the first list.

Once the join existed, it separated three groups that look identical on a low-usage chart. There are people with no access, people holding a seat they do not use, and people whose seat belongs to someone who has since left. Each needs a different response: give them access, have a conversation, take the seat back. Treat them as one group called low adoption and all three get the same email, and two of the three get the wrong one. Seats and people are not the same count, because some people hold more than one seat. Leave and start dates add two more distinctions. Someone who was away is not someone who has disengaged, and a new starter with low usage may simply not have had time yet. Counting days out, never the reason, keeps that conversation fair. My default is that the gap is the organisation's, not the person's, until shown otherwise.

The federal mandate asks a Chief AI Officer to promote adoption through governance and to advise on workforce. My reading is that neither is possible for people you cannot see, so measurement comes first, not as a reporting extra.

The objection is that a missing licence is an IT provisioning problem, not an AI strategy problem. It is both, and the strategy half is only visible once the provisioning half has been measured. The person without a seat has made no choice about AI at all, and an adoption figure that counts them as a non-user has quietly decided it for them. Where there is no governed way in, some of them turn to unsanctioned tools, and the figure counts them as non-users while they use AI outside any governance.

Do you know who in your organisation has no access at all, as opposed to who uses it least?