Amazon warns developers against ‘Tokenmaxxing’, explains how AI use will be judged

Amazon has updated its guidelines for software engineers, asking employees to start using artificial intelligence in their daily work or risk falling behind. The company believes AI is changing software development faster than previous technology shifts, forcing engineers and managers to rethink how they write code, solve problems and measure productivity.

The updated guidelines were shared in an internal memo published on Amazon’s official website by Terk Terkowitz, Vice President and Distinguished Engineer at Worldwide Amazon Stores. The memo, developed following discussions among senior engineering leaders, makes it clear that AI adoption is becoming a major part of software engineering at Amazon. Yet the company has warned managers against judging employees simply by how many AI tools they use or how much AI-generated code they produce.

:

Amazon wants engineers to use AI in everyday work

Amazon has asked software development engineers (SDEs) and software development managers (SDMs) to actively learn and experiment with AI tools.

“If you haven’t already, you should take the time to learn and apply AI into your daily work as soon as possible,” the memo states.

The company believes AI will become central to software engineering practices, even as the best ways to use the technology continue to change.

Amazon describes its future software engineer as a “Renaissance Developer”, someone who combines AI tools with human judgement, technical knowledge and a clear understanding of customer needs.

:

Amazon warns engineers against ‘tokenmaxxing’

One of the more interesting parts of the memo concerns how Amazon plans to measure AI adoption.

The company acknowledges that tracking AI token consumption, code changes and tool usage can provide useful information about productivity. It also admits these measurements can create problems.

Amazon specifically warns against “tokenmaxxing“, a practice where employees consume large amounts of AI tokens without delivering meaningful results.

For example, an engineer generating thousands of lines of code with AI may appear productive. That output means little if the software creates more problems or fails to help customers.

Amazon wants managers to examine actual results rather than reward employees for generating more code.

:

Bottom 10% of AI users could face closer examination

Amazon has recommended that managers pay special attention to teams in the bottom 10 per cent and top 10 per cent of AI adoption.

For teams using AI less frequently, managers should understand why adoption remains low. For heavy users, the question is whether their work has improved enough to justify the increased AI usage.

The memo clarifies that engineers can explain situations where AI does not help their work. They are encouraged to document such findings and discuss them with managers.

:

What does this mean for software engineers?

Amazon expects technical managers to gain firsthand experience with AI tools and understand their limitations.

The changes indicate that AI skills are becoming a workplace expectation for developers. Amazon is linking those skills to engineering judgement, customer understanding and measurable results.