Amazon Warns Engineers Against Misusing Artificial Intelligence

Chasing Tokens or Real Productivity?

Tech giant Amazon is setting new guidelines for its engineers regarding the use of artificial intelligence tools. In a memo circulated by Vice President and Distinguished Engineer Terk Terkowitz, the company explicitly criticizes the practice of "tokenmaxxing." This term refers to the intensive use of AI where the primary goal is not a high-quality outcome, but rather the consumption of as many "tokens"—the units of data processed by AI models—as possible.

According to Amazon leadership, the volume of token consumption, the pace of AI adoption, or the number of code changes are not accurate measures of productivity. The company states that using such metrics as targets only leads to increased costs and unnecessary complexity, which contradicts the company's core principles of frugality and simplicity.

What Changes for Amazon Engineers?

While the company continues to monitor AI usage, managers are being urged to approach this data with "caution and judgment." Terkowitz stated that if an engineer's workflow does not require artificial intelligence, it does not mean they are performing poorly; it is simply a matter of fact. Amazon maintains that teams using AI actively are generally more efficient, but it emphasises that artificial intelligence is merely a tool, not an end in itself.

The company has already discontinued internal leaderboards that ranked teams based on the frequency of AI usage, as these created unnecessary competition and tension among employees. The focus has now shifted to "feature velocity"—the success with which real changes and new products reach production.

A Global Trend: AI Euphoria Fades

Amazon is not the only company adjusting its AI budgets and usage policies. Microsoft, Uber, and Accenture have already begun taking similar steps to avoid the wasteful expenditure of resources. As AWS Vice President Marc Brooker notes, companies do not want to spend tokens on solving problems that do not actually exist.

For the technology sector in Georgia, this news serves as an important signal. Many local companies and startups are currently actively integrating AI into their workflows. Amazon's approach demonstrates that the effectiveness of artificial intelligence is not determined solely by the money or time spent on it, but by the real value that the technology delivers to the end user.