Key facts
- Tech leaders have historically stated that AI development would lead to reduced working hours.
- A former OpenAI employee reported working at least 70 hours per week, with sprints at AI companies like OpenAI and Anthropic potentially reaching over 90 hours.
- Meta employees described being "drafted" onto urgent AI projects, often working nights and weekends.
- Research from UC Berkeley indicated that workers using AI tools adopted a faster pace and took on more tasks, extending their work hours.
- Neil Thompson, an innovation scholar at MIT, suggested that time savings from AI are often consumed by the implementation and verification of the technology.
Despite pronouncements from tech leaders that artificial intelligence will lead to reduced working hours, employees at major AI development companies report facing intense work cultures and significantly longer hours than the traditional 40-hour week.
Executives at companies like Google and OpenAI have previously suggested that AI would usher in an era of shorter workweeks, with some even predicting a four-day week by 2025. However, a former OpenAI technical employee described a demanding environment characterized by frequent "crisis meetings," weekend work, and "super cut-throat" performance reviews, stating they regularly worked at least 70 hours a week.
Workers at other prominent tech firms, including Meta and Anthropic, are also reporting extended hours. Employees at Meta have described being "drafted" onto urgent AI projects, often working into the night and on weekends, with little option to refuse. Anthropic has highlighted its chatbot Claude's ability to operate for seven hours autonomously, framing it as a productivity gain.
Even employees not directly developing AI tools are experiencing increased workloads. A former Google employee cited a need to work late hours due to internal engineering functions failing as crucial resources were diverted to AI projects. Research from UC Berkeley, which followed hundreds of tech workers using AI, found that employees worked at a faster pace, took on more tasks, and extended their working hours.
Experts like Neil Thompson from MIT suggest that any potential time savings from AI are often absorbed by the need to implement, manage, and verify AI outputs. Furthermore, workers tend to fill any saved time with more tasks, either by choice or due to perceived pressure to prove their value, leading to a situation where AI's promise of less work is not translating into reality for many.