The target of the guidance is "tokenmaxxing", the practice of maximising AI token consumption in the belief that more AI usage means more productivity. The memo acknowledges Amazon has seen some correlation between token use and delivery on such teams, but argues the metrics are "deeply flawed" because they are disconnected from what matters. Used as a target rather than a measure, it warns, they could drive negative outcomes, increasing costs and producing overly complex code at odds with Amazon's aim to "invent and simplify".

An Amazon spokesperson confirmed the memo's intent was to discourage tokenmaxxing, pointing to comments from AWS CEO Matt Garman in June encouraging developers to measure outcomes rather than token consumption. The memo also calls out some teams for becoming "less connected to their customers and business", saying more engaged teams take the most advantage of AI and agentic development. It adds that if work does not benefit from current AI tooling, "that is a finding, not a failing".

The approach closely mirrors what Microsoft told its own developers in August, when EVP Jay Parikh said "tokenmaxxing is not what we are optimizing for" and urged focus on outcomes for customers and the business. With two companies of that size questioning token consumption as a productivity measure, The Register suggests more enterprises may instead turn to cheaper open-weight models suited to the task, an approach Garman also advocated in June.