355
Audio & Video Production344
Automation & Workflow224
Software Development250
Marketing & Growth192
AI Infrastructure & MLOps173
Writing & Content Creation203
Data & Analytics140
Design & Creative169
Customer Support130
Photography & Imaging156
Sales & Outreach125
Voice & Speech135
Operations & Admin87
Education & Learning131
Amazon removed an internal AI usage leaderboard after some staff ran unnecessary AI tasks to raise scores, which increased computing costs.
In short: Amazon has taken offline an internal AI usage leaderboard after some employees tried to boost their rankings in ways that raised computing costs.
Amazon shut down a service called “Kirorank,” an internal dashboard that scored employees based on how much they used AI tools on Amazon’s Kiro developer platform, according to people familiar with the matter.
The leaderboard appears to have encouraged some workers to create extra AI activity that did not help their work. In some cases, employees assigned AI agents, which are automated helpers that can take actions for you (like a junior assistant following instructions), to do unnecessary tasks just to climb the rankings.
Dave Treadwell, an Amazon senior vice-president, told staff not to “use AI just for the sake of using AI,” according to the same people. He also pointed to “tokenmaxxing,” meaning employees were trying to inflate their use of “tokens,” which are small chunks of text and data that an AI system processes (like items on a meter that counts usage).
Amazon said the dashboard was in beta, was not a formal or approved tool, and has now been deprecated.
This is a simple example of what can go wrong when companies measure tool use instead of results. A leaderboard can turn work into a points game, like rewarding someone for turning on the lights all day instead of for finishing the job.
The issue also connects to rising AI costs. Amazon is spending heavily on AI and data centers, and some AI providers have moved to pricing that charges based on how much you use, not a flat monthly fee. Amazon is reportedly shifting toward measures that focus on useful output, such as “normalised deployments,” instead of raw token use.
Source: Financial Times