Amazon updates AI coding guidelines as it warns engineers against tokenmaxxing

Highlights

Amazon's September memo updates internal guidance, emphasizing AI's pivotal role in software engineering while acknowledging current metrics are deeply flawed if used in isolation.
Amazon urges engineers to avoid using AI where it provides little value and stresses that demonstrating when AI isn't beneficial reflects sound judgment.
Although Amazon has not capped employee token usage, it shut down an internal leaderboard in May after widespread tokenmaxxing inflated costs without quality results.

Internal guidance reshapes AI coding practice

As first reported by CNBC, an internal Amazon memo circulated in September and posted on the company website on Tuesday updates how the company defines the role of software development engineers and managers in the AI era.

The memo says Amazon sees AI at the center of future software engineering practices, even as standards are still evolving. The company writes that the industry is only beginning to see AI's effect on software engineering and that best practices are still emerging.

Amazon says measures such as token usage, AI tool adoption rates, and code changes or reviews are "deeply flawed" if used to assess teams' or organizations' AI performance on their own. It says those indicators can drive higher costs and unnecessary complexity, though they should still be considered with care and judgment alongside other information.

The guidance also tells employees not to force AI into tasks where it offers little value. Amazon says work that does not benefit from current AI tools should be treated as a useful finding rather than a failure, adding that engineers show sound judgment when they can explain where AI is not yet helpful.

Cost discipline and workforce implications

Amazon has rapidly expanded AI across its businesses under Jassy, who has said the technology is likely to change the scope and speed of daily work. He has also said automation could reduce the company's corporate workforce over the next few years, while creating other roles.

The memo arrives as companies across the sector weigh the cost of heavy AI usage against uncertain productivity gains. Businesses have been trying to measure returns on investment more reliably while also containing rising expenses linked to tokens, the units of data processed and generated by AI models.

Amazon says it has not imposed caps on employee token usage, according to a company spokesperson. But the company shut down an internal leaderboard in May after finding that employees were tokenmaxxing, a practice in which developers use AI as much as possible with little regard for the quality or usefulness of the output.

In our earlier article on the KPMG survey of U.S. CEOs on AI agents, we reported that many large companies are seeing efficiency or growth gains, but only a small minority view AI agents as truly transformational so far. The survey also highlighted continued investment in AI infrastructure and a strong focus on hiring and retraining, with most leaders expecting headcount growth rather than near-term job cuts as human-AI collaboration expands.

This material may contain third-party opinions, none of the data and information on this webpage constitutes investment advice according to our Disclaimer. While we adhere to strict Editorial Integrity, this post may contain references to products from our partners.