Amazon issued a memo on Tuesday revising guidance for software engineers, encouraging AI use while urging employees to apply the tools judiciously. According to Sina Finance, the memo describes a “Renaissance Developer” era in which developers combine AI with their expertise and judgment, and says employees should tell managers when AI tools do not help their work.

Amazon Vice President and Distinguished Engineer Terk Terkowitz wrote that when AI does not benefit a task, identifying why is a discovery, not a failure. He said counting tokens or lines of code is a seriously flawed way to measure productivity because those metrics may not reflect work quality. Turning them into targets could raise costs and add unnecessary complexity, conflicting with Amazon’s principles of thrift and “Invent and Simplify.”

Terkowitz said AI adoption metrics remain closely related to important outputs. He encouraged managers to compare the 10% of their teams using AI the least with the 10% using it the most, to understand the reasons for low use and determine whether high use corresponds to greater output.

Reports said some employees had directed AI agents to perform pointless tasks to increase token use and climb the unofficial internal “Kirorank” leaderboard. Amazon shut down the leaderboard in May amid concerns about unnecessary computing costs. Separately, a company reportedly spent $500 million on Claude tools in one month after failing to set limits on employee use. Terkowitz said teams should use input metrics thoughtfully when considering career development and organizational performance, and work toward better metrics and targets over time.