IT之家 October 28 news: On the 27th local time, Amazon updated its working guidelines for software engineers in the AI era, emphasizing that evaluations should focus on how much business value engineers create, rather than how much AI they use.
Amazon Vice President and Distinguished Engineer Turk Tekowitz pointed out that compared with previous technological changes such as cloud computing, AI’s impact on software development has come faster and more profoundly. However, the key to measuring engineer performance remains what results they bring to customers and the business, not the frequency or amount of AI use. “We all must adapt to change and take time to learn, otherwise we risk being left behind.”
AI programming tools and agents are changing the software industry, and programmers’ work content is changing accordingly. Amazon calls this stage the “Renaissance Developer Era,” believing that combining human judgment with machine intelligence can develop new products and services faster and more efficiently.
The software industry is undergoing change, and the professional positioning of software engineers is also changing. In recent years, Amazon has increasingly emphasized engineers’ and technical managers’ use of AI and has tried to incorporate it into evaluations. The company has counted employees’ number of AI uses, number of tokens consumed, and lines of code written, and has also created leaderboards based on some of the data to compare different teams.
However, the memo warns that using these data to measure work performance is different from requiring employees to maximize the data. Over-pursuing metrics may backfire, for example, abusing AI to increase token consumption without writing better software or improving customer experience. Tekowitz said: “If these metrics are treated as goals rather than references for measuring work, they may cause negative results.”
According to Business Insider, Amazon had previously been trying different methods, hoping both to measure engineering teams’ AI usage and to encourage employees to use AI more.
Earlier this year, Amazon began tracking engineers’ use of internal AI tools, and some teams also launched leaderboards based on data such as token consumption. This move triggered dissatisfaction among some employees, who worried that adding unnecessary operations to meet AI usage requirements would make it easier to gain recognition than actually improving work efficiency.
The new guidelines are Amazon’s most explicit public acknowledgment to date that the company is aware of this risk. Amazon believes that detailed rules on how engineers should use AI may soon become outdated due to technological progress, so the new guidelines focus on the long-term outcomes that customers and the business can obtain.
IT之家 has learned that Amazon has not therefore reduced its emphasis on AI applications. The company emphasizes that AI should become “the core of software engineering practice” and requires engineers to use this technology in their daily work.
The memo also requires managers to personally use AI tools in actual projects, understand what problems AI can solve and where it still falls short, and explain their judgment basis to their teams.
Tekowitz pointed out that Amazon found that data such as token consumption, AI tool usage, and code review do have some connection with the work outcomes the company values, but one cannot look only at the numbers. Consuming more tokens means costs may increase, and blindly pursuing lines of code can easily make software unnecessarily complex.
For teams with large differences in AI application, managers need to focus on checking two groups: the bottom 10% of teams, to understand why they failed to use AI effectively; and the top 10% of teams, to confirm whether heavy AI use really accelerated work delivery.
Amazon found that the “frontier teams” that use AI most actively achieved better results in work efficiency and delivery performance. Conversely, teams and employees that did not adopt AI, or failed to improve output metrics, gradually fell behind others.
However, the new guidelines do not mean that all engineers must prove that AI can improve their work efficiency. For positions that cannot currently benefit from existing AI tools, employees can record the specific reasons, communicate with colleagues and managers, and work together to find solutions. The memo mentioned: “Engineers who can explain why AI cannot help for the time being precisely demonstrate the judgment we hope to see.”