Amazon is shifting its focus for software engineers, emphasizing meaningful results rather than mere artificial intelligence (AI) usage figures. The tech giant has issued new internal guidelines cautioning against a practice dubbed "tokenmaxxing," where employees might consume excessive AI tokens to boost usage statistics or give the impression of widespread AI adoption.
The Problem with Tokenmaxxing and Misleading Metrics
Tokens are the fundamental units of text and data processed by AI models. Higher consumption directly translates to increased operational costs for companies. Amazon's updated guidance, initially shared internally in September and published online in October, clarifies that metrics such as token consumption, AI adoption rates, and even code changes can provide an incomplete, and potentially misleading, picture of an engineer's true productivity.
The company previously shut down an internal leaderboard in May after discovering employees were manipulating token consumption to climb rankings. While Amazon has not imposed strict caps on token usage, it seeks to ensure that AI tools genuinely contribute to an engineer's output.
Focusing on Tangible Outcomes: "Feature Velocity"
Instead of quantifying AI interaction, Amazon wants to assess if AI actually helps engineers deliver valuable results. Marc Brooker, a vice president and distinguished engineer at Amazon Web Services, noted that the company is evaluating "feature velocity," which includes successful product launches and significant changes implemented in production environments.
The memo explicitly states, "if your work does not benefit from current AI tooling, that is a finding, not a failing." This underscores that engineers are not expected to integrate AI simply to inflate usage numbers.
Industry-Wide Concerns Over AI Investment Returns
This internal directive from Amazon comes amidst broader industry discussions about the return on investment for AI technologies. An IBM report, cited by Mint, indicated that "tokenmaxxing" saw a decline in popularity after peaking in spring 2026, largely due to escalating costs and a lack of clear, visible improvements in productivity.
Uber faced a similar challenge in May, reportedly exhausting its entire 2026 AI budget by April. Its chief operating officer highlighted difficulties in directly linking increased token consumption to improved customer-facing features or enhanced developer productivity. For Amazon, the core objective remains whether AI facilitates engineers in solving critical problems and producing better work, rather than simply how much of the technology they employ.