Amazon tells its engineers AI will be central to software work, but outcomes still matter most

Focus on what is not changing

According to Amazon, engineering leaders including Distinguished and Senior Principal Engineers recently met to reaffirm what it means to be an SDE at the company. They drew on observations from the most successful SDEs to produce the updated guideline. The company says its most prolific engineers have shown that neither the developer role nor its output is defined by the use of AI, and it describes the present as the era of the Renaissance Developer, which blends AI tools with human judgment and domain expertise.

The memo says Amazon has historically avoided naming specific technologies in SDE guidelines, given the variety of its businesses. Leaders debated whether AI warrants an exception. They concluded that AI is changing software engineering more quickly and more fundamentally than earlier trends such as the cloud, microservices and serverless. Even so, the authors say it is still early, and any definition of good AI-powered practice would likely look antiquated within a year. Their solution was to concentrate on what stays constant: impact on customers, businesses and teams.

Measuring AI use, with caveats

The memo acknowledges that long-term outcomes are the best measure but can be delayed, hard to measure and hard to separate from other trends. In some cases Amazon has used input metrics such as token usage or AI tool adoption, or throughput metrics such as code changes and reviews. The authors call these metrics deeply flawed because they are disconnected from what matters. If treated as targets, they warn, such metrics could raise costs, noting that tokenmaxxing is at odds with frugality, and increase complexity, since measuring lines of code is at odds with invent and simplify.

Still, the memo says these input metrics correlate well with the outputs that matter and are expected to keep doing so for some time. Amazon attributes this to higher productivity and delivery on what it calls frontier teams, the ones that have embraced AI the most, and to the fact that teams and individuals not adopting AI or raising output metrics are falling behind. Using the metrics with care and judgement is described as appropriate. Leaders are advised to look at the 10th percentile of AI adoption to find teams not embracing AI effectively, and the 90th percentile to check whether heavy users are seeing matching delivery gains.

The memo also says that if someone's work does not benefit from current AI tooling, that is "a finding, not a failing." Engineers are encouraged to document why, share it with peers and managers, and look for solutions.

Expectations for leaders and closeness to customers

Technical and organizational leaders are expected to understand AI hands-on rather than relying on what they hear. According to the memo, sufficient understanding means having used the tools on production work, being able to describe a task where they helped and one where they did not, and explaining that judgement to the team. Leaders should also balance exploring new approaches against doubling down on existing ones, and take ownership of connecting inputs and outputs in their organizations.

Amazon says the leaders with the most impact are those who emphasize learning and curiosity, both for themselves and for their teams, and that returns on this investment have been disproportionately large. It also reports that SDEs and software development managers who are closer to their products and businesses get the most from AI and agentic software development. Because some individual contributors and managers have become less connected to customers as Amazon has grown, the company wants to reverse that trend.

What it means

For Amazon's engineers, the message is that AI fluency is expected soon, and that judgement about where AI helps and where it does not is valued. The company says it will keep evolving the SDE role as AI's effect on its business and engineering practices becomes clear. Amazon also presents its approach as a possible model for customers, positioning itself as a leader in agentic AI with thousands of SDEs.