The rise of tokenmaxxing is creating unexpected challenges for companies trying to encourage AI adoption. Amazon has shut down an internal AI leaderboard after employees reportedly began generating unnecessary AI activity to climb rankings, driving up costs in the process.
The employee-created tool, known as Kirorank, was a short-lived beta project that tracked how often workers used AI tools across the company. What started as a way to encourage AI adoption quickly turned into something else entirely.
Apparently, when you turn AI usage into a leaderboard, people start treating tokens like high scores.
TL;DR
- Amazon shut down its internal AI leaderboard, Kirorank
- Employees were reportedly engaging in AI tokenmaxxing
- Token usage increased costs without delivering meaningful benefits
- Meta and other companies have faced similar AI spending challenges
- Businesses are becoming more cautious about AI usage while adoption continues to grow
What Is Tokenmaxxing?
Tokenmaxxing refers to the practice of intentionally generating large amounts of AI activity to increase token counts and climb usage rankings.
Instead of using AI for meaningful tasks, employees reportedly began assigning trivial or unnecessary work to AI systems simply to boost their numbers.
According to reports, growing token consumption significantly increased operational costs while delivering little additional value.
Amazon eventually removed the leaderboard and urged employees to stop using AI simply for the sake of generating usage metrics.
AI Costs Are Becoming Harder to Ignore
The issue highlights a broader challenge facing companies rushing to integrate AI into every corner of their operations.
While AI adoption continues to grow, so do the costs associated with running large language models, especially as businesses deploy coding assistants, AI agents, and always-on automation systems.
More AI activity means more tokens, and more tokens mean larger cloud bills.
Turns out “just add AI” becomes a much more expensive strategy when someone has to pay for every prompt.
Amazon Isn’t Alone
Amazon is not the first company to encounter problems tied to AI usage incentives.
Meta reportedly shut down its own employee-run AI leaderboard earlier this year after similar tokenmaxxing behavior emerged.
Meanwhile, companies across the industry are beginning to reassess AI spending as usage costs rise faster than expected.
Reports indicate that organizations including Salesforce, DoorDash, and Microsoft have started tightening controls around AI-related spending and tool usage.
The AI race is still happening. Companies are simply becoming more selective about where they put fuel in the engine.
AI Usage Is Still Growing
Despite the pullback, AI consumption continues to rise at a rapid pace.
Google recently reported that Gemini usage increased from 480 trillion monthly tokens in 2025 to 3.2 quadrillion monthly tokens in 2026.
Much of that growth comes from AI agents, coding assistants, and automated systems that consume far more resources than traditional chatbot interactions.
The challenge for businesses now is figuring out which AI workloads create value and which simply create invoices.