Licences do not equal adoption

The board pack says 12,000 licences deployed. The steering committee calls it adoption. It is not adoption. It is procurement.

A licence is a start line. Adoption is what happens after: whether people change how they work, whether the tool touches the tasks that matter, whether anyone would notice if you switched it off on Tuesday.

Around 6% of organisations qualify as true AI high performers. The other 94 are not short of licences. They are short of the unglamorous middle layer: workflows redesigned around the tool, managers who model the new behaviour, time carved out to learn, and a straight answer to the question every employee is quietly asking, which is what happens to me if this works.

If you want a number worth reporting, stop counting seats and start counting behaviour. Weekly active use on real tasks. Cycle time on the processes you said would speed up. Rework rates. The number of tasks formally moved to the model, and the number deliberately kept human.

Licences tell you what you bought. Adoption tells you what you changed. Boards should only pay attention to the second one.