THE WORLD IS NOT STANDING STILL.RSS
BIG CHANGE.
BIG CHANGE, one tap away
BIG CHANGE, one tap away

Make room for the big picture.

Add BIG CHANGE to your home screen or desktop. It opens in its own window, with the latest stories when you’re online.

iPhone & iPad
Open bigchange.ai in Safari. Tap Share, then Add to Home Screen. Turn on Open as Web App if shown, then tap Add.
Android
Open bigchange.ai in Chrome. In the browser menu, choose Install app or Add to Home screen.
Desktop
In Chrome or Edge, use the install icon in the address bar or the browser’s app menu. In Safari on Mac, choose File → Add to Dock.

Markdown edition

# Amazon's AI engineering guidance looks beyond usage counts

> Amazon's guidance weighs AI activity alongside longer-term outcomes.

By BIGCHANGE.AI Editorial

Published: 2026-10-29T11:13:02.795Z
Updated: 2026-10-29T11:13:02.795Z
Canonical: https://bigchange.ai/blog/amazon-ai-engineering-guidance-usage-outcomes

![Editorial illustration of a hand at a keyboard beside a display of code.](https://bigchange.ai/api/media/file/hero.png)
AI-generated editorial illustration by BIG CHANGE. Conceptual software work, not a record of an Amazon workplace or product interface.

Amazon's [engineering guidance published on October 27](https://www.aboutamazon.com/news/company-news/amazon-software-engineer-role-guideline-ai) encourages AI use while emphasizing long-term results for customers, the business and teams. It warns that making activity counts targets can increase cost and complexity.

That leaves a practical question for an engineering manager: how can a count help identify something worth investigating when it cannot establish the result on its own? The distinction affects what a leader asks after seeing a dashboard, and what evidence an engineer can offer about a task.

## The big change

- More activity on a dashboard does not necessarily mean a team has delivered more value. A useful measurement system needs to show what happened after code was produced, including the quality of the result and the effort needed to get there.
- A measure of activity can direct attention to a pattern that needs investigation. It takes additional evidence to determine what that pattern means for the work customers receive.
- An account of a task can give managers a way to examine the work itself. Other research shows why teams also need to agree on what they mean by productivity and quality before drawing conclusions from their measures.

## What the numbers can and cannot tell a manager

Amazon says long-term results can be delayed, hard to measure and difficult to isolate from other trends. Tokens count consumption; code changes count activity. A smaller change that solves a problem can be more useful than a larger one.

The memo recommends investigating ineffective adoption at the 10th percentile and checking whether heavy use at the 90th matches greater delivery. The questions concern obstacles and task fit at one end and results at the other.

Amazon attributes its reported correlation partly to higher productivity and delivery among heavy adopters, without providing supporting data or methodology. Correlation alone does not establish causation.

## How to investigate an unhelpful task

The memo asks engineers to document and share unhelpful uses, and expects leaders to use the tools in production work and explain tasks where they helped and did not. BIG CHANGE reviewed the published guidance and supporting research; it did not test Amazon's internal practices.

A manager can ask what the tool did, what the engineer kept and what work remained.

Microsoft researchers' separate [EngThrive measurement framework](https://www.microsoft.com/en-us/research/publication/engthrive-make-it-fast-and-easy-to-do-great-work/) illustrates a way to pair a result with information about the work behind it. It combines outcome-oriented measures with diagnostic measures, system telemetry and developer surveys, across speed, ease and quality, with developer well-being as a guardrail. The framework describes Microsoft's approach; it does not establish that Amazon uses it or that AI caused a particular result.

The distinction matters when a team is trying to interpret a change in its figures. A measure of a delivery result tells the manager whether that result moved; the accompanying details and developer accounts can help identify what to examine next. Both are needed to discuss a specific task, and neither alone proves why a result changed.

## Agree on what good work means

Even the intended outcome needs a shared definition. In a [survey study of developers and managers](https://www.microsoft.com/en-us/research/publication/how-developers-and-managers-define-and-trade-productivity-for-quality/), Microsoft researchers found that the two groups differed in how they described productivity. Respondents also differed on whether quality could be traded for productivity or was part of it. This study is about reported views, not an experiment on AI tools or proof that Amazon teams disagree.

That finding gives the measurement conversation a concrete starting point: agree what the team expects a successful change to accomplish and which quality properties matter. Otherwise, a faster change and a better change may be described with the same word while people are assessing different things.

For the related question of what happens after AI-generated code enters the delivery process, [BIG CHANGE's earlier analysis](https://bigchange.ai/blog/ai-coding-ci-bottleneck-dependable-delivery) examines review, validation and the limits of productivity studies. Together, these questions connect how a team works with what reaches its users.

## Sources

- [Amazon engineering memo](https://www.aboutamazon.com/news/company-news/amazon-software-engineer-role-guideline-ai) — Correlation without supporting data.
- [EngThrive: Make It Fast and Easy to Do Great Work](https://www.microsoft.com/en-us/research/publication/engthrive-make-it-fast-and-easy-to-do-great-work/) — Microsoft researchers describe outcome and diagnostic measures, system telemetry and surveys.
- [How Developers and Managers Define and Trade Productivity for Quality](https://www.microsoft.com/en-us/research/publication/how-developers-and-managers-define-and-trade-productivity-for-quality/) — Survey research on how developers and managers define productivity and consider quality tradeoffs; not an AI experiment.