Artificial intelligence was supposed to save employees time, but for many workers it has created a second shift. After full workdays, they now spend evenings and weekends learning ChatGPT, Claude, Cursor, AI agents, automations, prompts, and constant updates so they can keep pace with changing workplace expectations.
The article says this has produced a new kind of “learning tax,” the unpaid personal time employees use to stay relevant. An EY survey cited in Business Insider found that 85% of office workers in the United States are learning to work alongside AI agents outside working hours, 83% say most of their AI knowledge is self-taught, and 59% report that insufficient organizational training is a barrier to skill development.
The benefits are real. BCG’s AI at Work 2026 report found that 42% of regular AI users save at least eight hours a week, nearly a full workday. But the piece argues that the saved time is often not returned to workers, and instead becomes pressure to produce more, faster, and with fewer people. A presentation that once took half a day may now be expected in an hour, and a document that needed hours is now supposed to start with AI.
The article also highlights the less visible work of supervising AI. Glean’s Work AI Index 2026 calls this “botsitting,” meaning workers spend hours each week checking, correcting, and improving AI output. The new skill is not just using AI, the article says, but managing it, doubting it, and catching when it is wrong, misleading, or risky. In this environment, workers are judged by how quickly they can learn the next tool, while employers must decide who provides training, guidance, and time.
