Amazon shifts focus from AI metrics

Amazon has instructed its engineers to prioritize the outcomes of their work over using AI to complete tasks, warning against reliance on "tokenmaxxing" metrics. These metrics, which evaluate success through AI-generated tokens or code, were described as "deeply flawed" in a memo from Terk Terkowitz, vice president and emeritus engineer at Amazon.

New definition of developer roles

The memo emphasized that the role of a developer and their productivity should not be measured by AI use alone. It introduced the concept of the "Renaissance developer," a professional who integrates AI tools with human judgment and expertise to innovate and create new solutions.

AI costs exceed expectations

In July 2026, Amazon engineering teams observed cases where switching from handwritten code to AI models resulted in project costs far exceeding budgets. For instance, $1.8 million was spent on Anthropic's Claude Sonnet for a project that was never delivered, as reported in internal presentations.

Microsoft tracks AI spending closely

Microsoft began closely monitoring AI-related expenses in August 2026 after an employee from the "Customer and Partner Solutions" division reported spending $28,000 on AI tokens in just 28 days. An analysis by Business Insider found that the median AI-related expense over 28 days was about $300, with some departments spending up to $975.

Focus on customer impact and outcomes

Amazon’s memo urged employees to concentrate on customer impact, business outcomes, and team success rather than metrics tied to AI token usage. The note, titled Amazon Technical Leadership's Perspective on the Role of the Developer in the Age of AI, emphasized the importance of building a "deep connection" with customers to better understand their needs.

Challenges in managing AI costs

In July 2026, business leaders faced difficulties predicting and controlling AI operational costs due to the shift to usage-based pricing. KPMG reported that nearly half of organizations adjusted their technology deployment timelines as a result, while companies like Amazon and Microsoft worked to help customers leverage AI tools effectively.