Amazon Warns Engineers Against Using AI Tools Just to Boost Productivity Metrics

Amazon has issued updated guidance to its software engineering teams, warning employees against using artificial intelligence tools simply to increase their usage statistics. The guidance, published in an internal memo and reported on October 28, 2026, encourages engineers to focus on the actual results of their work rather than the number of AI tokens they consume or lines of code they generate. Terk Terkowitz, Amazon’s vice president and distinguished engineer at Worldwide Amazon Stores, described the emerging ideal as the “Renaissance Developer” — an engineer who combines AI capabilities with human judgement, technical expertise and creativity. The company believes that successful software development will increasingly involve AI, but insists that employees must understand when the technology genuinely improves their work.

A major concern highlighted in the memo is a practice known as tokenmaxxing, which involves deliberately increasing AI usage to appear more productive. AI models process information using tokens, and employees can consume large quantities when interacting with coding assistants and automated agents. Amazon reportedly discovered that some employees were using AI systems for unnecessary activities to increase their positions on an unofficial internal leaderboard called Kirorank. The company eventually discontinued the leaderboard in May after concerns emerged about employees manipulating the rankings and generating unnecessary computing expenses. Amazon now wants its engineers to understand that high token consumption is not proof of valuable work, particularly when the resulting output requires additional corrections or creates unnecessary complexity.

Terkowitz also criticised the use of AI adoption statistics, code changes and review numbers as direct measures of employee performance. According to the memo, these indicators can become misleading when managers turn them into strict targets. Employees may concentrate on increasing measurable activity instead of improving software quality, solving customer problems or delivering useful products. Amazon warned that this approach could increase computing costs and undermine its long-standing business principle of simplifying processes. The company therefore wants managers to evaluate whether AI adoption produces meaningful improvements rather than rewarding employees simply for using more tools. It also encouraged engineers to explain when existing AI technology does not provide a clear benefit to their work.

Despite the warning, Amazon is not abandoning its commitment to artificial intelligence. The company expects AI to become a central part of future software engineering and continues encouraging employees to learn how to use AI tools effectively. Its updated guidance recommends that managers examine both employees with relatively low AI usage and those with particularly high usage. Lower adoption may indicate that some workers are missing opportunities to improve their productivity, while unusually high usage should prompt managers to determine whether it is producing better results. Amazon’s position is that the appropriate level of AI adoption depends on the nature of the work, the quality of the output and the value delivered to customers.

The announcement reflects a wider shift in the technology industry as businesses reconsider how they measure the benefits of AI investment. Several companies have encouraged employees to adopt AI coding assistants and automated agents, but rising computing costs and inconsistent results are forcing organisations to pay closer attention to actual productivity. Amazon’s latest guidance suggests that the next stage of AI adoption will not simply be about using more powerful models or generating more code. Instead, companies will increasingly need to determine where AI saves time, improves quality and creates measurable value. For software engineers, the message is clear: AI should strengthen human expertise and decision-making, not replace critical thinking or become a performance target in itself.