Tokenmaxxing: The Productivity Metric That May Be Wasting Billions

Tokenmaxxing, a trend in which companies incentivize employees to maximize their AI token usage as a measure of productivity, has been gaining traction. Google has dubbed it a 'breakout' term for 2026, with search volume growing more than 5,000% this year.

A token is the basic unit of input and output for Large Language Models (LLMs), and companies pay by the token, whether in computing power or money. IBM Principal Product Manager Markus Eisele points out that tokenmaxxing was initially based on the conviction that heavy AI consumption would lead to better outcomes.

However, its drawbacks became apparent quickly. With tokenmaxxing, employment decisions are made based on who's consuming the most tokens, which is not a good direct measurement of performance, according to IBM Staff Writer Dave Bergmann.