Amazon Issues New AI Guidelines for Engineers, Warns Against Tokenmaxxing and Focuses on Productivity

Amazon has introduced new AI guidelines for software engineers, encouraging responsible AI adoption, customer-focused innovation, hands-on learning, and measurable productivity while warning employees against meaningless AI usage metrics.

Amazon’s New AI Rulebook: Key Guidelines for Software Engineers

Amazon has issued updated internal guidelines explaining how its Software Development Engineers (SDEs) should use artificial intelligence (AI) in their daily work. The company is encouraging engineers to adopt AI tools where they deliver meaningful results, rather than using the technology simply to demonstrate activity or increase usage statistics.

The guidance highlights Amazon’s emphasis on customer satisfaction, engineering quality, cost efficiency and measurable productivity. It also outlines how engineers and managers should evaluate AI’s actual contribution to software development.

Amazon Embraces the Era of the ‘Renaissance Developer’

Amazon believes AI is transforming software engineering more rapidly and fundamentally than previous technological shifts. While developments such as cloud computing, microservices and serverless architecture changed how software was built, the company says AI is reshaping the engineering profession itself.

The memo introduces the concept of the “Renaissance Developer”—an engineer who combines AI capabilities with human judgment, technical expertise and a deep understanding of the business.

According to the guidelines, engineers should actively learn how to incorporate AI into their workflows. However, simply using AI does not define an employee’s value or the quality of their work.

Amazon maintains that its fundamental engineering principles remain unchanged. Developers are still expected to build reliable, high-quality products that address genuine customer needs.

Amazon Warns Engineers Against ‘Tokenmaxxing’

One of the central concerns highlighted in the memo is “tokenmaxxing,” a practice in which employees assign unnecessary tasks to AI tools or agents to increase token consumption and appear more productive.

Amazon has cautioned that this approach creates misleading impressions of performance without necessarily improving software quality or delivery speed.

The company reportedly shut down an internal AI leaderboard in May 2026 after discovering that employees were manipulating the system to improve their rankings.

The updated guidelines also criticise the use of AI token consumption and the volume of AI-assisted code changes as primary performance indicators.

Amazon considers these measurements unreliable because they can encourage wasteful behaviour, increase computing costs and conflict with its long-standing principle of frugality.

Real Productivity Matters More Than AI Usage

Instead of evaluating engineers based on how frequently they use AI, Amazon wants managers to focus on measurable business outcomes and long-term productivity.

Marc Brooker, a vice president at Amazon Web Services (AWS), told CNBC that the company is concentrating on “feature velocity”—a measure of how effectively teams deliver successful updates and new features to customers.

This approach shifts attention from AI activity to the practical value created by engineering teams.

The memo also recommends examining teams at different levels of AI adoption. Leaders should understand why some teams are struggling to integrate AI effectively, while also checking whether the heaviest users are genuinely delivering results faster.

The objective is to distinguish productive AI adoption from activity that increases usage without improving customer outcomes.

Managers Must Gain Practical AI Experience

Amazon’s new guidelines make it clear that technical and organisational leaders should develop hands-on experience with AI tools instead of relying solely on reports or second-hand opinions.

Managers are expected to use these technologies in real production workflows and understand where they improve performance, where they fall short and what limitations remain.

The company encourages leaders to follow its principle of “learn and be curious” by setting aside time for themselves and their teams to experiment with AI.

Practical experience, Amazon believes, will help managers make better decisions about adoption, training and the effectiveness of different tools.

Engineers Should Not Force AI Into Every Task

Although Amazon encourages employees to develop AI skills, it does not expect every task to benefit from the technology.

The memo recognises that some engineering activities may not improve when current AI tools are introduced. Employees are encouraged to identify such situations, document their findings and communicate them to their managers.

Amazon views this feedback as an opportunity to understand the limitations of available tools rather than evidence of an employee’s failure.

This approach reinforces the importance of critical thinking and sound technical judgment. Engineers must determine when AI offers a genuine advantage and when traditional methods remain more effective.

Customer Focus Remains Amazon’s Top Priority

Amazon also wants engineers and managers to strengthen their understanding of the products they develop and the customers who use them.

As organisations expand, technical teams can become increasingly disconnected from customer needs and business priorities. The company wants to address this gap by encouraging employees at every level to remain closely involved in product direction.

According to the guidelines, engineers who understand customer problems can make better decisions about where AI should be applied—and where it should be avoided.

The message is clear: AI is a tool for solving problems, not a goal in itself. Amazon wants its engineers to combine technological innovation with accountability, sound judgment and a strong focus on delivering value to customers.

Conclusion

Amazon’s updated AI guidelines signal a shift from measuring technology usage to evaluating meaningful results. By discouraging tokenmaxxing, promoting hands-on learning and prioritising customer outcomes, the company aims to ensure AI improves engineering productivity without compromising quality or cost efficiency.