On the surface, it appears that companies in nearly every industry are gobbling up AI contracts. But internally, employees are increasingly complaining that AI doesn’t save them any time on productive tasks and increases stress.
Many employers are now starting to mandate the use of AI in the workplace, and some are even going so far as to fire employees who don’t participate in AI to justify spending large amounts of money on contracts with the tech industry. While the forced adoption of AI has a major impact on the economic viability of the technology as a whole, it also provides perverse incentives for office workers to increase their use of AI in non-productive tasks.
As a good example, financial times According to reports, Amazon office staff are increasingly using the company’s in-house AI agent, MeshClaw, to perform personal tasks to meet quotas.
In an effort to get more than 80% of developers to use AI each week, Amazon introduced employee-specific AI usage goals, in addition to a broader “token consumption” leaderboard that tracks how much each employee uses AI (in machine learning terminology, a token refers to the basic unit of data that an AI model uses to understand text).
However, according to the staff interviewed, F.T.employees are abusing the system by increasingly using mandated AI systems to automate personal tasks. This is a tactic known as “token maxing.”
“The pressure to use these tools is huge,” one anonymous employee told the paper. “Some people simply use MeshClaw to maximize their token usage.”
Meanwhile, Amazon said: F.T. “Every day, thousands of Amazonians automate repetitive tasks,” it said, adding that the retailer is “committed to developing and deploying generative AI in a safe, secure and responsible manner for our customers.”
A quick peek at Team Blind, a message board for verified employees of companies like Google and Apple, shows that this practice is widespread, or at least widely acknowledged.
“I burned the token, [project manager]” one Amazon employee wrote in a May 8 post. It makes good use of the GPU. ”
When I asked a Microsoft employee what that meant, an Amazon staffer replied, “We just paste in the Slack conversation history and tell our agents to use 10 subagents to analyze that guy.”
Less maliciously, there are plenty of other threads from employees across the tech industry asking how other companies are maximizing the use of their tokens, and the replies are filled with free advice.
In one thread, a SnapChat employee encouraged people to “get creative.” “If companies are using brain-dead metrics to judge people, you need to learn how to beat them fast.”
Learn more about leveraging AI: AI is giving bosses more tools than ever to be monsters
