Amazon scraps internal AI leaderboard after employees game system through tokenmaxxing

3 Sources

Share

Amazon has taken down KiroRank, an unofficial internal AI leaderboard, after employees began creating unnecessary AI agents to boost their rankings—a practice called tokenmaxxing. The initiative backfired as workers burned through costly tokens, driving up expenses while Amazon pushed for over 80% of developers to use AI weekly.

Amazon AI Leaderboard Becomes Costly Experiment

Amazon has discontinued its internal AI leaderboard after an unofficial initiative to promote AI adoption spiraled into an expensive lesson about gamification. The beta dashboard, known as KiroRank, tracked employee usage of Amazon's Kiro agentic AI development platform, ranking workers according to their AI activity

1

. What began as a grassroots effort to drive AI awareness among Kiro employees quickly transformed into a competitive race that had unintended financial consequences for the tech giant.

Source: Analytics Insight

Source: Analytics Insight

The leaderboard was created by a group of employees who wanted to demonstrate how AI can accelerate work, but senior managers discovered that workers were engaging in tokenmaxxing—creating pointless agents and unnecessary AI tasks solely to climb the rankings

1

2

. This practice resulted in excessive token consumption, as AI algorithms process data by breaking it down into small chunks called tokens, which require GPU processing power and generate substantial costs under pay-per-token models

2

.

Pressure to Use AI Drives Unintended Behavior

The tokenmaxxing phenomenon didn't occur in a vacuum. Amazon had introduced aggressive targets requiring more than 80% of developers to use AI each week

3

. This mandate created an environment where employees felt compelled to demonstrate AI usage, with some workers possibly engaging in inflating usage scores out of fear of redundancy or to prove their value as employees

2

.

The situation highlights a broader challenge facing enterprises pushing for efficient AI adoption. When companies implement metrics without considering how employees might game the system, the results can undermine the original goals. Amazon's approach essentially communicated to staff: use AI for your job or lose your job to AI, creating pressure that manifested in counterproductive behaviors

2

.

Source: InfoWorld

Source: InfoWorld

Rising Costs Force Rethinking of AI Strategies

The increased AI costs from the leaderboard experiment come as major AI companies like OpenAI and Anthropic shift from flat subscription models to pay-per-token models due to ballooning expenses

2

. This pricing structure makes every token generated a direct cost, turning what might have been harmless competition into painfully pricey bills for Amazon.

In response to the situation, Amazon confirmed the removal of the tracker, stating that "the beta dashboard was not a formal or approved tool, and has since been deprecated"

3

. The company emphasized its focus on operational efficiency and sharing best practices to celebrate innovation rather than promoting AI usage for its own sake

3

.

This incident serves as a cautionary tale for organizations implementing AI awareness initiatives. As pay-per-token models become increasingly commonplace and AI continues generating limited viable returns relative to investment, companies may need to reconsider how they measure and incentivize AI adoption. The challenge lies in fostering genuine productivity gains rather than creating metrics that encourage wasteful behavior disguised as engagement.

Today's Top Stories

TheOutpost.ai

Don’t drown in AI news. We cut through the noise - filtering, ranking and summarizing the most important AI news, breakthroughs and research daily. Spend less time searching for the latest in AI and get straight to action.

Instagram logo
LinkedIn logo
Youtube logo
© 2026 TheOutpost.AI All rights reserved