What happened
Meta will no longer factor AI dashboards or token counters into engineer performance reviews, executives Maher Saba and Santosh Janardhan said in an internal memo. Reviews now focus on quality, speed, and complexity.
Why it matters
Meta's prior policy of rewarding AI use backfired: employees burned tokens in bulk to look good on leaderboards, pushing internal AI costs toward billions in 2026. Meta will roll out budgets and a central dashboard starting in 2027.
What to watch
Meta's new AI agent tool Hatch, which handles computer tasks, is meeting internal resistance as employees hesitate to connect it over privacy concerns. Whether this changes adoption remains the key test.
WHO IT HITSEngineering managers at Meta must reset how they evaluate reports, focusing on output quality rather than AI tool metrics. Finance and operations teams will watch AI usage costs, which are significantly trending upward heading into 2026.
Ask the AI about this article →
Summaries like this, in your inbox every morning.
The move by Meta to decouple AI usage from performance signals a correction after an initial, overly aggressive push to encourage AI adoption. The previous approach created a perverse incentive structure, directly leading to "tokenmaxxing." The financial consequence of this gaming was severe, with internal AI use costs projected to hit billions in 2026, forcing a more controlled approach starting in 2027.
This adjustment happens as Meta introduces Hatch, a new AI agent tooling meant to automate computer tasks. However, the internal rollout is facing friction due to employee privacy concerns about linking it to personal accounts. The company now balances its financial reality with the cultural adoption of AI.
The outcome ultimately hinges on whether Meta can foster sustained AI adoption without a rigid metric-based review structure. It must demonstrate the value of tools like Hatch to privacy-conscious employees while implementing the new cost-control measures. The success may therefore depend on internal communication to rebuild trust after the unintended consequences of the initial performance metric.
For example, today's edition would include:
AI-summarized, only the topics you pick — one digest a day via Email, Slack, or Discord.
Free · 30 seconds with Google · unsubscribe anytimeWhat is AIToday? →
Ask AI anything about this article. Q&As are published on this page for other readers too.
Perplexity CEO Aravind Srinivas told CNBC on September 4 that banks like Morgan Stanley and JPMorgan want "air…
Pampa Energía plans a facility of up to 500 megawatts in Neuquén province, seeking investors with first contra…
SoftBank introduced an AI system for video interview screening in May 2020, cutting screening time by about 70…
OpenAI's Astra model is now available, beating the ARC-AGI-3 benchmark and scoring the highest on Zapier's Aut…
Mistral AI raised €3 billion (about $3.58 billion) at a post-money valuation of more than €21 billion (about $…
A new OpenAI blog post announces that its GPT‑6 Astra model can now act as a "semi-autonomous agent" to handle…
