Token leaderboards distort AI adoption

Companies that measure AI use by raw consumption (tokens, calls, leaderboards) create incentives for employees to inflate usage — automating unnecessary tasks or routing work through internal models — so metrics read well even when productivity or safety do not improve. That gaming hides real impacts, encourages lock‑in to internal tools, and undermines trustworthy measurement of AI adoption. — This reveals a recurring governance failure: metric design and public leaderboards can reshape behavior in ways that mislead managers, regulators, and the public about AI’s true use and value.