At Meta, an employee built an internal leaderboard ranking colleagues by AI token consumption. Badges included “Token Legend”. Engineers reportedly left AI agents running idle overnight to climb it. When the dashboard leaked, it was gone within days.
Google, meanwhile, says it now processes 3.2 quadrillion tokens a month, seven times more than a year ago. The number is true. It tells you almost nothing.
This is Goodhart’s law: when a measure becomes a target, it ceases to be a good measure. The moment token usage became the score, people optimised for the score. We’ve been here before. We once measured programmers by lines of code, and we know how that ended.
If you want to know whether AI is really changing how your teams work, ask them different questions:
- What decision got faster this quarter?
- What work stopped existing?
- What would you fight to keep if the tools vanished tomorrow?
- What valuable work are you doing now with the time saved?
Those answers are harder to put on a leaderboard, and that’s the point. They measure value, not activity.
Here’s a final twist. If these companies had used AI to help define their AI adoption metrics, they might well have avoided the problem altogether.
Try it with your team. The answer won’t be a token effort.
Pia and Dan