Token‑count metrics mislead engineers

Measuring AI work by tokens consumed or raw adoption rates incentivizes waste (higher compute bills, unnecessary tool‑use, and added complexity) without proving value. Amazon's internal guidance tells managers to prioritize long‑term outcomes and to treat heavy AI usage as a signal to investigate, not an automatic success metric. — How firms measure AI matters for costs, procurement, governance, and whether AI adoption produces real productivity gains or merely 'slop' and lock‑in.