Amazon memo warns engineers against 'tokenmaxxing' and wasteful AI use

Internal guidelines say measuring success by token consumption or code generation rates is 'deeply flawed'

By LineZotpaper AI staff

Published27 October 2026

Read Time2 min

Sources2 outlets

Illustration: Amazon memo warns engineers against 'tokenmaxxing' and wasteful AI use

Amazon has circulated an internal memo telling software engineers to be more selective in using artificial intelligence tools, discouraging 'tokenmaxxing' and warning that metrics such as token usage and code generation rates are 'deeply flawed' measures of productivity.

The memo, sent in September and posted to the company's website on Tuesday, revises the role of software development engineers and managers as Amazon pushes AI into its operations. Under CEO Andy Jassy, Amazon has rapidly integrated AI across its business, with Jassy predicting the technology will change the scope and speed of work and potentially shrink the corporate workforce through automation while creating other jobs.

As employees rush to adopt AI tools, companies have grappled with surging costs and difficulty measuring returns. Some have set limits on token usage, the units of data processed and generated by AI models, or switched to cheaper open-source models. Amazon told CNBC it has not capped employees' token usage, but in May it shut down an internal leaderboard after discovering employees were 'tokenmaxxing' to climb the ranks.

The memo says metrics like token usage, adoption rates, and code changes remain 'deeply flawed' ways of measuring AI usage, as they can lead to higher costs and complexity. However, Amazon stopped short of telling staffers to ignore them entirely, saying they should be considered 'with care and judgment' alongside other data.

'We believe that it is best to measure, and be measured by, long-term outcomes,' the company wrote.

Marc Brooker, a vice president and distinguished engineer at Amazon Web Services, told CNBC the company looks at metrics such as 'feature velocity', how many launches and successful changes are made in production, to gauge AI's impact on productivity. 'Broadly, we think about this as a return on investment,' Brooker said. 'What we don't want to see is wasted effort.'

The memo also encouraged engineers not to use AI when it doesn't suit the task: 'If your work does not benefit from current AI tooling, that is a finding, not a failing.'

RSS

The Daily Edition

The front page in your inbox each weekday morning: the lead, the picks, and why each one earned its place.

Analysis

Why This Matters

Cost control: Tokenmaxxing drives up AI expenses for companies without guaranteed productivity gains, and Amazon's stance may influence how other tech firms manage AI costs.
Productivity measurement: The memo highlights the broader challenge of accurately measuring AI's return on investment, an issue affecting businesses across industries.
Workforce implications: Amazon's guidelines shape how engineers work with AI, potentially affecting job roles and the pace of automation in its corporate workforce.

Background

Amazon under CEO Andy Jassy has aggressively integrated AI across its operations, including retail, cloud computing, and logistics. Jassy has said AI will change the scope and speed of work and may reduce the corporate workforce through automation while creating new roles. The company's internal guidelines come amid an industry-wide scramble to adopt AI tools, with many firms struggling to assess whether the technology provides meaningful productivity improvements or merely inflates costs.

Key Perspectives

Amazon (via memo): Long-term outcomes are the best measure of success; metrics like token usage and code generation are 'deeply flawed' and can encourage wasteful behaviour. Engineers who identify where AI does not help show good judgment. Marc Brooker, AWS vice president: Amazon looks at 'feature velocity', the rate of successful launches and changes, to measure AI's return on investment. The goal is to avoid wasted effort on problems that don't matter while allowing room for experimentation.

What to Watch

Whether Amazon introduces formal token caps or further refines its productivity metrics in response to cost pressures.
How other tech companies, particularly those investing heavily in AI, respond to the tokenmaxxing phenomenon and whether they adopt similar guidelines.
The impact on Amazon's internal AI adoption rates and engineering culture as employees adjust to the new guidance.

Sources

Checked

12 claims in this story checked against 1 source: 8 supported.

Show each claim

Partly supportedThe memo warns engineers against tokenmaxxing and wasteful AI use.Tech: “Amazon also discouraged staffers from using AI when it doesn’t make sense for a given task. “If your work does not benefit from current AI tooling, that is a finding, not a failing,” Amazon said in the memo.”
Partly supportedThe summary says token consumption and code generation rates are deeply flawed measures of success.Tech: “In the memo posted Tuesday, Amazon said metrics like token usage, the rate of adoption of AI tools, and code changes or reviews remain “deeply flawed” ways of measuring teams or organizations’ AI usage”
SupportedAmazon circulated a memo urging software engineers to be more selective about using AI and discouraging tokenmaxxing.Tech: “Amazon laid out its AI usage guidelines for software engineers in a memo to staffers, and discouraged employees from turning to “tokenmaxxing” in their push to embrace artificial intelligence tools.”
Partly supportedThe opening says token usage and code generation rates are deeply flawed measures of productivity.Tech: “metrics like token usage, the rate of adoption of AI tools, and code changes or reviews remain “deeply flawed” ways of measuring teams or organizations’ AI usage”
Partly supportedThe memo says token usage and code generation should be considered with care and judgment alongside other data.Tech: “Token usage and code generation should be considered “with care and judgment” alongside other data, Amazon said.”
SupportedAmazon says it is best to measure, and be measured by, long-term outcomes.Tech: ““We believe that it is best to measure, and be measured by, long-term outcomes,” the company wrote.”
SupportedBrooker said Amazon uses feature velocity, including launches and successful production changes, to gauge AI’s effect on productivity.Tech: “Marc Brooker, a vice president and distinguished engineer at Amazon Web Services, told CNBC in an interview that the company has looked to metrics such as “feature velocity” to measure how AI is improving engineer productivity. The metric looks at things like, “how many launches are we doing, how many successful changes are we making in production,” he said.”
SupportedBrooker said, “Broadly, we think about this as a return on investment.”Tech: ““Broadly, we think about this as a return on investment,” Brooker said.”
SupportedBrooker said, “What we don’t want to see is wasted effort.”Tech: “He added, “What we don’t want to see is wasted effort where we are going and spending those tokens solving problems that don’t matter and sometimes it’s hard to know, so we do need to have some appetite for experimentation.””
SupportedThe memo says, “If your work does not benefit from current AI tooling, that is a finding, not a failing.”Tech: ““If your work does not benefit from current AI tooling, that is a finding, not a failing,” Amazon said in the memo.”
SupportedThe memo was circulated internally in September and posted to Amazon’s website on Tuesday.Tech: “The memo, which was circulated internally in September and posted to the company’s website on Tuesday, revises how the company defines the role of software development engineers and managers in the age of AI.”
SupportedAmazon said it has not capped employees’ token usage and shut down an internal leaderboard in May after discovering employees were tokenmaxxing to climb the ranks.Tech: “An Amazon spokesperson said the company hasn’t instituted caps on employees’ token usage. Amazon in May shut down an internal leaderboard after discovering employees were tokenmaxxing to climb the ranks.”