Amazon Warns Engineers Against AI Productivity Traps in Latest Memo

As concerns grow over AI’s impact on software engineering jobs, Amazon has issued a memo outlining what it expects from its engineers as AI reshapes software development.

While acknowledging AI’s capabilities in software engineering and its long-term prospects, Amazon said, “It is difficult to define, in a way that will stand the test of time, what good AI-powered engineering practices look like.” 

“Any definition we write now will likely look antiquated by this time next year,” the company added. 

And these ‘antiquated definitions’ risk slowing down progress, says Amazon, which wants to focus on what isn’t changing in software engineering roles. 

“Specifically, impact on our customers, our businesses, and our teams. The outcomes that matter aren’t changing, even if the inputs, tools, and techniques are changing faster than ever,” said the company. 

The company said that metrics that often grab attention, such as token usage, the number of AI tools used, code changes, the number of reviews, and more, are “deeply flawed”. 

Amazon said that these metrics are “disconnected from what matters”. 

It adds that if teams use these metrics as targets—not just measurements—they could drive negative outcomes, increasing costs and complexity. 

“We believe that it is best to measure, and be measured by, long-term outcomes. But these outcomes can be time-delayed, difficult to measure, and hard to isolate from other trends,” said the company. 

Amazon recommends that engineering teams use AI adoption metrics carefully, paying particular attention to the 10% of teams with the lowest adoption and the 10% with the highest adoption.

The company wants managers to investigate why some teams struggle to incorporate AI into their workflows and whether teams using AI extensively deliver corresponding productivity improvements. 

Amazon said its most AI-forward teams have seen productivity and delivery improvements. It also warned that engineers and teams that don’t adopt AI or increase their output risk falling behind.

However, the company said engineers shouldn’t be expected to use AI for every task. 

“If your work does not benefit from current AI tooling, that is a finding, not a failing,” Amazon said. 

It added that an engineer who can explain where AI falls short is demonstrating the judgement the company expects.

Amazon also wants engineering leaders to gain hands-on experience with AI tools rather than relying on second-hand accounts of their capabilities. 

Managers should be able to identify tasks where AI has improved their teams’ work, as well as those where it has not, and explain the reasoning behind their decisions.

Another priority is reconnecting engineers with customers and the businesses they serve. Amazon said some individual contributors and software development managers had become less connected to customers and business priorities as the company grew.