Amazon Issues New AI Guidelines to Curb “Tokenmaxxing” Among Engineers

Amazon circulated an internal memo in September and posted it publicly this week, outlining revised expectations for software development engineers and managers as artificial-intelligence tools become integral to the company’s workflow. The guidance warns staff against “tokenmaxxing”—the practice of driving AI models to generate the maximum number of tokens without regard for the quality or relevance of output. Amazon’s leadership emphasizes that responsible AI use, rather than sheer volume, will define future engineering best practices.

CNBC

The memo arrives amid a broader corporate push to embed generative-AI across Amazon’s extensive operations. Under CEO Andy Jassy, the firm has accelerated AI adoption, with Jassy previously stating that AI will reshape the speed and scope of daily work and could eventually reduce the overall workforce through automation while creating new roles. At the same time, many technology companies are confronting soaring AI-related expenses and the difficulty of measuring productivity gains, prompting a wave of internal policies aimed at controlling token consumption and exploring lower-cost, open-source alternatives.

CNBC

The company asserts that the new guidelines are intended to balance innovation with cost-effectiveness and to prevent wasteful usage of AI resources. Jassy’s earlier remarks, cited in the memo, reinforce the view that AI will both automate tasks and generate new employment opportunities, underscoring the firm’s long-term strategic commitment to the technology.

CNBC

Following the announcement, industry observers note that Amazon joins a growing list of firms imposing token limits or shifting to cheaper, open-source models for certain development tasks. These measures aim to curb rising AI bills while still leveraging the productivity benefits of generative tools. Analysts suggest that such internal controls could become a standard component of corporate AI governance as enterprises seek to quantify return on investment and avoid unchecked resource consumption.

CNBC

Amazon has not disclosed a specific timeline for enforcing the new token-usage limits, but the memo indicates that the policies will be integrated into engineering workflows in the coming months. The company plans to monitor adoption and adjust guidelines as AI capabilities continue to evolve.

CNBC