Meta has revised its performance expectations for engineers, telling managers that artificial intelligence usage metrics such as token counts and adoption dashboards should not be used to determine employee impact, months after the company intensified its push for workers to integrate AI into their daily workflows.
The updated guidance was communicated in an internal memo from Meta executives Maher Saba and Santosh Janardhan, according to reports. Managers have instead been asked to evaluate engineers on factors including the quality of their output, speed of execution, complexity of problems addressed and the scope of their work.
The change marks a shift in how Meta measures the impact of AI adoption internally, rather than a retreat from its broader AI strategy.
Meta had previously encouraged employees to become more "AI native", with AI-driven impact becoming an increasingly visible part of its workplace expectations. Engineers were encouraged to incorporate chatbots, coding assistants and AI agents into their workflows, while internal dashboards allowed AI adoption to be tracked.
The approach reportedly contributed to a practice referred to by some employees as "tokenmaxxing", where workers increased their consumption of AI tokens amid concerns that higher usage could be interpreted as stronger AI adoption.
Meta's latest guidance seeks to separate the volume of AI usage from the business impact generated by employees. The company told engineers that it would not use AI adoption dashboards or token counts to evaluate impact, according to The Information.
The internal memo reportedly states that AI adoption is not the objective by itself. Instead, Meta wants employees to deliver better outcomes, including higher-quality products and faster iteration.
The company said 93% of code changes are now assisted by AI agents, highlighting the extent to which artificial intelligence has already entered its engineering processes. Meta Chief Technology Officer Andrew Bosworth had also told employees earlier this year that AI tools should not be used merely to increase usage numbers and that token consumption alone was not a measure of impact.
The revised approach comes as Meta continues developing AI tools for its workforce. Employees have recently begun testing Hatch, an agentic AI system designed to perform tasks across applications and interact with the web.
Meta's change therefore does not signal the end of its AI-first workplace push. Instead, it shifts the focus from how frequently employees use AI tools to what they produce with them, as companies continue to work out how artificial intelligence should be incorporated into productivity and employee evaluation systems.
Disclaimer: This article may include information derived from interviews, press releases, public statements, research, company communications and other publicly available or third-party sources. Such material may be summarised, paraphrased or contextualised for journalistic and editorial purposes. All rights in third-party content remain with their respective owners.