Technology giants such as Google, OpenAI, Anthropic, and Meta invest billions of dollars annually in developing AI. For years, executives have painted the same future: when AI takes over more tasks, humans can work fewer days. OpenAI has even publicly called on companies to trial a four-day workweek without pay cuts, claiming AI will greatly increase labor efficiency. However, a BBC investigation found that these companies promoting shorter working hours have never lived up to their promises. Former OpenAI employees revealed that during their tenure, they worked at least 70 hours per week, and during peak periods, even over 90 hours, with weekends still being used to catch up or troubleshoot system failures.
Ironically, the more AI increases productivity, the longer employees' working hours become. Meta recently laid off about 10% of its staff while forcing many into AI project teams, which employees privately call "conscription"—refusing means resigning. After joining an AI team, working late into the night and continuing on weekends has become the norm, and even after work, they remain on standby. A former Google employee said that after the company continuously shifted processing power and storage resources to AI projects, internal engineering system failures became more frequent, often requiring them to stay up late to resolve issues. After leaving, their sleep and health conditions significantly improved.
The time saved by AI is consumed by new work
A study by the University of California, Berkeley, tracking hundreds of employees at a tech company for eight months revealed a core paradox: AI makes work faster and tasks broader, but the amount of time spent working each day actually gets longer. A key reason is that employees must constantly check whether AI-generated content is correct, a "supervision cost" that offsets the efficiency gains from AI. MIT innovation researchers pointed out that even if AI does save time, the saved portion will be swallowed by the work of system adjustments, implementation, and ensuring smooth operations.
A deeper driving force comes from employees' own anxiety. Some want to use the time saved to complete more work, while others fear that not continuously increasing output may prevent them from proving their value to employers. New work will always emerge, keeping the idea that "a 20% reduction in workload could lead to a four-day workweek" stuck in theory. Anthrop's CEO, Amodei, has openly stated that the higher the AI productivity, the more jobs may be eliminated in the future, further increasing employees' pressure to not relax. AI was supposed to do many things for humans, but in reality, at least in the core areas of the tech industry, more people are not sleeping better or living more easily, but instead falling into a cycle where the more efficient they are, the busier they become.
