Inside the Tech Industry’s Newest Metric: Workers Forced to Prove Productivity by Burning AI Credits

Forcing workers to use AI, and rating their workplace performance on AI use, can easily go wrong.

BY KIT EATON @KITEATON

Illustration: Getty Images

In the race to embed AI into pretty much every corner of business, are some companies going too far, too fast, and putting too much pressure on their workers? A report about Amazon’s embrace of AI suggests yes. Meanwhile, CEOs of other tech companies are turning to a new and unusual measure of employee performance: How many AI use credits they burn through per day. The theory here is the more the better, because executives are looking for the much-hyped productivity boosts that AI promises. All of this news may prompt you to think about your own company’s AI deployment—possibly framed as a “don’t do it like this” lesson.

Amazon, which is a technology company just as much as a global-spanning e-retailer, is pressing hard for its workers to use AI. Over half a dozen current and former Amazing staffers spoke to the U.K.’s Guardian newspaper about the initiative and what they revealed was pretty shocking. They alleged that the tech giant is deploying AI in a “haphazard way” even as it simultaneously tracks worker AI use—a mismatch that, you might think, is both risky and unfair to workers. Worse, the Amazon workers said that Amazon is pushing for staff to use AI tools in every aspect of their jobs, whether or not it may prove useful or helpful. And in some cases this pressure is actually harming productivity.

A New-York based Amazon software developer called Dina told the newspaper that her job used to be writing code, but is now all about repairing things that AI has broken. Dina explained that Amazon’s “Kiro” tool often hallucinates and spits out flawed code that forces her to meddle frequently with the bad code. Dina explained that she, and other coworkers, argue the AI is not accelerating their output, “but from management, we are certainly getting messaging that we have to go faster, this will make us go faster, and that speed is the number one priority.” 

The management-frontline worker expectation gap for AI use has been well-documented, but this example shows how harmful it can be. 

How Anthropic's Claude AI Became a Co-Founder

Separately a New York Times report covers how other tech firms are tracking their workers’ AI use. Staggeringly, one engineer at market-leading AI firm OpenAI burned through 210 billion AI “tokens,” which is how AI interactions are instigated and tracked. This all happened in one week, was the most of any employee, and generated enough text content to “fill Wikipedia 33 times.” At rival AI firm Anthropic, one user of the company’s AI code-writing system burned through AI tokens costing over $150,000 in just one month. These are staggering AI use rates, and represent a vast acceleration of AI use compared to a 2024 report—eyeopening at the time—which said the CEO of drug company Moderna was pressing his workers to interact with ChatGPT at least 20 times a day, in search of purported productivity gains. 

The Times notes that managers in some tech companies have started to include a workers’ AI use statistics into the performance review process. Managers are said to be “rewarding workers who make heavy use of A.I. tools and chastening those who don’t.” On the flip side of this management pressure, workers in tech firms are setting up their own AI use metrics, and are competing with each other on how many AI tokens they’re using up as part of their day job. Meanwhile, the companies’ leaders are rewarding high-flying coders with more generous “token budgets.” This kind of workplace perk is a far cry from typical traditional workplace benefits like bonuses or gym memberships. Some workers are using these tokens, the Times says, to “automate as much of their own work as possible.”

This last quote may trigger readers’ worries about one supposed long-term impact of AI, which is that it’ll replace human staff with automated AI systems. Amazon worker Dina expressed these concerns to the Guardian, noting some workers are worried that by using AI at an accelerated rate they’re simply training AI systems to eventually replace them—a “demoralizing” worry.

© Inc.com