Meta has removed AI tool usage from engineer performance reviews, according to internal executives Maher Saba and Santosh Janardhan. The change comes after concerns over 'tokenmaxxing' led to excessive AI token consumption and rising costs. The company now prioritizes the quality, speed, and complexity of engineers' work instead of their AI usage.
The shift follows a trend where employees used AI tools in bulk to appear high-performing on internal leaderboards. This practice, known as 'tokenmaxxing,' has driven internal AI costs toward billions in 2026. Meta plans to implement budgets and a central dashboard starting in 2027 to manage AI resource allocation more effectively.
Meta is also testing its new AI agent tool, Hatch, which is designed to automate computer tasks. However, the tool has faced resistance from employees who are hesitant to link it to their personal accounts due to privacy concerns. The company is still evaluating how to address these internal challenges.
"AI dashboards and token counters won't factor into performance reviews anymore," said Maher Saba, according to an internal memo. The memo emphasized that the focus will now be on the output and impact of engineers' work rather than their AI tool usage.
The decision follows a broader internal push to balance AI efficiency with employee well-being. Meta has acknowledged the need to refine its approach to AI integration, ensuring it supports productivity without creating unintended incentives.
Meta did not specify how it will measure the success of its new review criteria, and the company remains cautious about the long-term effects of the policy change. The move reflects a shift toward more sustainable and ethical AI practices within the organization.
Source: thedecoder
