Meta Platforms is reversing its approach to engineer evaluations, scrapping AI-related usage metrics from performance reviews after previously pushing staff towards an "AI-native" work culture. The company has firmly told employees that internal dashboards tracking AI adoption rates and system token consumption will no longer factor into assessments of their individual contributions.
An internal memo dated Tuesday, penned by executives Maher Saba and Santosh Janardhan and reviewed by The Information, outlines the updated expectations for engineers company-wide. The revised document strips away references to specific AI tools from the engineers' career guidelines and eliminates hard requirements tied to any particular AI usage. Instead, the new policy re-centers reviews on the caliber of work and the value of actual output delivered.
According to the memorandum, the new evaluation direction states that managers should focus on the quality of output, the speed of delivery, the complexity of problems tackled, and the overall volume of tasks handled. The document reveals that currently, a staggering 93% of code changes at Meta are assisted by AI agents, yet it stresses that using AI tools is not an end in itself. The real objective remains achieving superior business results with higher-quality deliverables and faster iteration cycles.
A Meta spokesperson has declined to comment on the record. This policy shift arrives after months of the company actively nudging employees to transform into "AI-native" workers. Earlier this year, Meta had begun experimenting with incorporating AI usage into workflows and evaluations. Late last year, the firm attempted to gamify AI adoption, tracking usage data and motivating staff to boost their AI interaction numbers. This led some employees to falsely believe that higher AI consumption would automatically translate into better performance scores, spawning behaviors like competing for the unofficial "Token Legend" title and deliberately burning through massive amounts of tokens to signal enthusiastic AI engagement.
Meta has already started clamping down on such metric-gaming practices. In a previous internal memo from April, Chief Technology Officer Andrew Bosworth told employees that no one should use AI merely for the sake of using AI, adding that token consumption alone is a completely inadequate measure of an individual's work contribution.