Amazon cracks down on AI tokenmaxxing after rising cost, tells employees don't use AI just for…
Amazon has reportedly taken down an internal AI leaderboard after employees allegedly started inflating AI usage. The shift reflects wider concern over rising AI costs and pressure to tie usage to useful work.
When Amazon started ranking employees based on their AI usage, some workers reportedly began “tokenmaxxing” — using AI for unnecessary tasks in an apparent attempt to climb the leaderboard. But as the company’s computing costs started rising, Amazon appears to have learnt its lesson and reportedly decided to shut the system down.
According to a report by the Financial Times, citing people familiar with the matter, Amazon informed employees that its “KiroRank” system, which scored users of the company’s Kiro developer platform based on their AI activity, had been taken offline.
Update: "One of the internal dashboards, called KiroRank, was recently created by a group of employees who wanted to drive awareness for how AI can accelerate work, and was never intended to promote the use of AI for usage's sake. The beta dashboard was not a formal or approved tool, and has since been deprecated. We’re focused on AI adoption and sharing best practices to celebrate innovation and operational efficiency gains across the company, and we’re proud of the way our teams are embracing this technology," Amazon said in a statement to India Today Tech.
Dave Treadwell, a senior vice-president at Amazon, told staff earlier this week that the leaderboard had been created with “good intentions.” However, the system reportedly ended up increasing costs for the company because employees began “tokenmaxxing” — a term used for inflating consumption of AI tokens, which are units of data processed by AI models.
In simple terms, some employees were allegedly using AI tools for unnecessary tasks just to improve their ranking scores.
“Please don’t use AI just for the sake of using AI,” Treadwell reportedly told employees.
Amazon shifts focus from usage to useful work
The report says Amazon has now started using a different metric called “normalised deployments” to evaluate adoption of AI tools inside the company. Instead of measuring raw AI token consumption, the new metric reportedly focuses on whether engineers are regularly using AI to create useful code and complete meaningful work.
The move comes at a time when the technology industry has become increasingly vocal about the real costs of deploying advanced AI systems.
AI costs are becoming a growing problem
AI companies such as Anthropic have recently shifted toward consumption-based pricing models instead of flat monthly subscriptions, a move that has reportedly increased costs for some customers. Some industry insiders have even begun arguing that, in certain situations, hiring a skilled human employee may be cheaper than heavily relying on AI coding systems like Claude.
Recent reports have also suggested that Microsoft has been reducing Claude Code licenses as part of efforts to cut operating expenses ahead of the new financial year in July.
Earlier this year, while speaking to The Information, Praveen Neppalli Naga, CTO of Uber, said the company’s assumptions around AI spending had already been exceeded, largely because of the rapid adoption of advanced coding tools like Claude Code developed by Anthropic.
Amazon is spending heavily on AI infrastructure
Amazon itself relies extensively on Anthropic’s AI models. At the same time, the company has been carrying out sweeping layoffs as it tries to reduce costs while continuing massive investments in AI infrastructure.
The report says Amazon is expected to spend nearly $200 billion in capital expenditure this year, with most of that money going toward AI systems and data centre infrastructure.
When Amazon started ranking employees based on their AI usage, some workers reportedly began “tokenmaxxing” — using AI for unnecessary tasks in an apparent attempt to climb the leaderboard. But as the company’s computing costs started rising, Amazon appears to have learnt its lesson and reportedly decided to shut the system down.
According to a report by the Financial Times, citing people familiar with the matter, Amazon informed employees that its “KiroRank” system, which scored users of the company’s Kiro developer platform based on their AI activity, had been taken offline.
Update: "One of the internal dashboards, called KiroRank, was recently created by a group of employees who wanted to drive awareness for how AI can accelerate work, and was never intended to promote the use of AI for usage's sake. The beta dashboard was not a formal or approved tool, and has since been deprecated. We’re focused on AI adoption and sharing best practices to celebrate innovation and operational efficiency gains across the company, and we’re proud of the way our teams are embracing this technology," Amazon said in a statement to India Today Tech.
Dave Treadwell, a senior vice-president at Amazon, told staff earlier this week that the leaderboard had been created with “good intentions.” However, the system reportedly ended up increasing costs for the company because employees began “tokenmaxxing” — a term used for inflating consumption of AI tokens, which are units of data processed by AI models.
In simple terms, some employees were allegedly using AI tools for unnecessary tasks just to improve their ranking scores.
“Please don’t use AI just for the sake of using AI,” Treadwell reportedly told employees.
Amazon shifts focus from usage to useful work
The report says Amazon has now started using a different metric called “normalised deployments” to evaluate adoption of AI tools inside the company. Instead of measuring raw AI token consumption, the new metric reportedly focuses on whether engineers are regularly using AI to create useful code and complete meaningful work.
The move comes at a time when the technology industry has become increasingly vocal about the real costs of deploying advanced AI systems.
AI costs are becoming a growing problem
AI companies such as Anthropic have recently shifted toward consumption-based pricing models instead of flat monthly subscriptions, a move that has reportedly increased costs for some customers. Some industry insiders have even begun arguing that, in certain situations, hiring a skilled human employee may be cheaper than heavily relying on AI coding systems like Claude.
Recent reports have also suggested that Microsoft has been reducing Claude Code licenses as part of efforts to cut operating expenses ahead of the new financial year in July.
Earlier this year, while speaking to The Information, Praveen Neppalli Naga, CTO of Uber, said the company’s assumptions around AI spending had already been exceeded, largely because of the rapid adoption of advanced coding tools like Claude Code developed by Anthropic.
Amazon is spending heavily on AI infrastructure
Amazon itself relies extensively on Anthropic’s AI models. At the same time, the company has been carrying out sweeping layoffs as it tries to reduce costs while continuing massive investments in AI infrastructure.
The report says Amazon is expected to spend nearly $200 billion in capital expenditure this year, with most of that money going toward AI systems and data centre infrastructure.
