Amazon Tells Engineers to Judge AI by Outcomes, Not Token Use

Amazon has updated its software-engineering guidance to put long-term results ahead of raw AI usage. The company warns that using token consumption or code changes as targets can raise costs and complexity, while still saying those measures can offer useful signals when handled carefully. The revised role guideline, published by Amazon on October 27, says engineers should assess AI by its impact on customers, the business and their teams. Amazon calls the practice of chasing token usage “tokenmaxxing” and says engineers should use AI where it helps their work, not simply to increase activity (Amazon’s updated engineering guidance).