How do you benchmark a team's AI ability?

You benchmark a team's AI ability by measuring the same real-work signals consistently across people and over time: which tasks each person can improve with AI, whether the gains are verified, and how capability trends month to month. Consistent evidence, not a one-off test, is what makes a benchmark meaningful.

A benchmark is only useful if it is comparable. That means measuring capability the same way for everyone, from real work rather than self-report, so differences reflect ability and not who rated themselves generously.

It also has to move. A single snapshot tells you where a team was; a living benchmark tells you where it is going. Tracked over time, it shows which teams are pulling ahead and where support will pay off.