A Meta employee built an internal tool called “Claudeonomics” that tracked and ranked colleagues by how many AI tokens they consumed — turning Claude usage into a company-wide competitive leaderboard. The dashboard went public knowledge when The Information reported on it. Two days later, it was gone. Mark Zuckerberg didn’t make the top 250.
What Claudeonomics Was
The dashboard was built internally on Meta’s own infrastructure and gave employees a real-time view of who was using the most AI tokens across the company. It wasn’t an official product — it was a grassroots tool built by one employee as, in their words, “a fun way for people to look at tokens.” The competitive element was clearly intentional: ranking users publicly against each other is a specific design choice that turns a usage metric into a social motivation.
The name itself — Claudeonomics — confirms that Claude is the primary AI model in use at Meta for internal work, which aligns with Anthropic’s own disclosures about enterprise customers. Meta has been one of the largest enterprise AI consumers, and its employees’ usage patterns at the token level give a rare window into how AI is actually being integrated into knowledge work at scale.
Who Was at the Top
The Information reported that software engineers dominated the top of the leaderboard — consistent with the expectation that coding tasks generate the most token usage of any knowledge work category. An hour of AI-assisted coding generates far more token consumption than an hour of AI-assisted writing or research, because code generation involves longer context windows, more iterative back-and-forth, and more complex reasoning chains.
The fact that Zuckerberg didn’t rank in the top 250 is notable not because CEOs are expected to be the heaviest direct AI users — they typically aren’t — but because Meta has made AI usage a stated performance expectation across the company. In January 2026, Meta overhauled its performance review system to incentivize the highest AI users with bonuses of up to 200%. Chief People Officer Janelle Gale told employees that “AI-driven impact” would be a “core expectation” in 2026. The existence of Claudeonomics suggests some employees took that expectation and turned it into sport.
Why It Was Shut Down
The shutdown message left on the dashboard read: “We’ve really enjoyed building this app on Nest for everyone. It was meant to be a fun way for people to look at tokens, but due to data from this dashboard being shared externally, we’ve made the decision to shutter Claudeonomics for now.”
Meta confirmed the employee took the dashboard down voluntarily: “The employee took down the dashboard at their discretion; Meta did not request this action.” The company noted separately that it maintains an official dashboard for token usage metrics geared toward software engineers — suggesting the data itself isn’t proprietary, but a public-facing ranking of individual employees created complications once it escaped internal circulation.
What It Reveals About AI Adoption at Scale
The Claudeonomics episode is a useful case study in how AI adoption actually unfolds inside large organizations — messily, organically, and often ahead of official policy.
The competitive leaderboard dynamic isn’t unique to Meta. Organizations that have made AI usage a performance metric are discovering that employees respond to incentives in the same ways they always have: by competing, by gaming metrics, and by building their own tools when official ones don’t exist. A leaderboard that ranks token usage doesn’t measure whether that usage is producing value — only whether it’s happening. An engineer generating thousands of tokens of mediocre code ranks higher than one generating hundreds of tokens of excellent code.
The gap between “AI-driven impact” as a stated expectation and “AI-driven impact” as a measurable reality is one of the defining management challenges of 2026. How you measure AI contribution without inadvertently optimizing for token consumption over output quality is a problem most organizations haven’t solved yet.
The Broader Context: Meta’s AI Investment
Meta’s 2026 capital expenditure is projected at $115-135 billion — nearly double 2025 levels — driven almost entirely by AI compute demand. The company recently signed a $21 billion deal with CoreWeave for GPU capacity between 2027 and 2032. At that scale of investment, ensuring employees are actually using AI effectively — not just using it to generate high token counts — becomes a material business question, not a management curiosity.
Claudeonomics existed for a few days and is now gone. But the question it raised — how do you actually measure AI-driven impact in knowledge work — is going to be one of the most consequential organizational challenges of the next few years.
Conclusion
An internal AI usage leaderboard at one of the world’s largest tech companies is a small story that points at a large question: how do organizations measure, incentivize, and actually capture value from AI adoption at scale? Browse our directory to explore Claude and every AI tool driving the enterprise adoption wave that stories like Claudeonomics are symptoms of.
