Amazon has shut down its internal "Kirorank" AI leaderboard after employees engaged in "tokenmaxxing," a practice that inflated operational costs by using artificial intelligence for unnecessary, menial tasks.

Key Points

• Amazon deprecated the Kirorank beta dashboard, which tracked employee AI usage, after determining it encouraged inefficient spending.
• The company identified "tokenmaxxing" as a primary driver of rising costs, where staff performed trivial tasks to boost their leaderboard rankings.
• Senior Vice President Dave Treadwell previously issued a memo warning employees against using AI tools solely for the sake of usage metrics.
• Similar trends are emerging at other major firms, including Meta, which closed its own "Token Legend" leaderboard, and Uber, which exceeded its 2026 AI budget.
• Companies like Salesforce, DoorDash, and Microsoft are increasingly rationing AI access as they struggle to balance high token costs with measurable business returns.

Why it Matters

These developments signal a shift in corporate strategy as businesses move from experimental AI adoption to prioritizing cost-efficiency and tangible productivity gains. This trend highlights the growing challenge of managing the high financial overhead of generative AI while companies work to distinguish between high-value applications and wasteful usage.