The AI performance metric fiasco and the reemergence of old management pitfalls

In 2026, companies like Meta and Amazon faced backlash over internal AI performance metrics that incentivized employees to prioritize token usage over actual productivity. Meta’s internal leaderboard, dubbed Claudeonomics, ranked top AI token consumers, awarding titles like 'Token Legend' and 'Cache Wizard.' Employees burned through over 60 trillion tokens in a single month, while a Disney employee interacted with an AI assistant 460,000 times in nine days. JPMorgan, KPMG, Amazon, and Accenture reportedly tracked employee AI activity, integrating it into performance reviews. The trend led to the emergence of 'tokenmaxxing,' where employees manipulated metrics by using AI for tasks that didn’t require it. Meta and Amazon reportedly shut down their leaderboards after public backlash, with critics arguing they encouraged cheating. The article references V.F. Ridgway’s 1956 paper on the dysfunctional consequences of performance measurements and Charles Goodhart’s law, which warns that when a measure becomes a target, it ceases to be a good measure. The piece highlights a recurring pattern in management practices, as companies repeatedly fail to learn from past mistakes.