Amazon has revised software engineer job guidelines to focus on actual customer and business results rather than the amount of AI tools used. While encouraging active use of AI, it has moved to curb the practice of treating simple usage volume as a measure of productivity.

Business Insider reported on Oct. 27 that Amazon recently released new guidelines for engineers setting out roles and evaluation priorities in the AI era.

Terk Terkowitz (테르크 테르코위츠), an Amazon vice president and principal engineer, stressed in an internal memo that AI is changing how software is developed faster and more fundamentally than past technology innovations such as cloud computing.

Amazon defined the shift as the era of the “Renaissance Developer.” It means the company needs developers who combine AI processing power with human judgment and expertise to create new products and services.

It drew a line, saying how much AI a developer uses does not determine capability. The company has tracked AI tool usage rates, token consumption and the amount of code written, and used them to rank some teams.

It warned that if such metrics become targets, “Tokenmaxxing” could emerge, where people focus on increasing usage rather than delivering results. Unnecessary AI use can raise costs, and an obsession with code volume can make software architecture more complex, it said.

That does not mean usage metrics will be scrapped. Amazon said a correlation appears between AI use and productivity, and explained the metrics should be used cautiously by linking them to actual results.

It recommended that managers identify adoption obstacles for teams in the bottom 10 percent for AI use, and check whether high usage by teams in the top 10 percent translates into actual work results.

It also stressed that managers should apply AI directly to their work to understand its benefits and limits, and share that with team members. Developers assigned tasks where AI is not currently helpful were told to document why.

Amazon's position is that AI use should be a core part of development work, but a developer's value should be judged by long-term results that contribute to customers and the business, not by tool usage volume.