Work AI Institute Study Reveals Hidden Labor of Botsitting

The Work AI Institute published a study revealing that while generative AI saves digital workers roughly 11 hours per week, it introduces a new form of labor called botsitting. This process involves checking outputs, fixing errors, and rerunning prompts, consuming over six hours of the average worker's week. The research, which surveyed 6,000 employees across the United States, United Kingdom, and Australia, indicates that 75% to 78% of individuals feel more productive, yet only 13% to 18% of organizations report significant business gains.

Data shows a systemic failure rate where more than one-third of AI sessions require restarts or significant rework. Consequently, employees are increasingly acting as managers for AI tools, with 41% of workers admitting to delivering AI-generated work they cannot explain. The report suggests that roughly 37% of AI interaction time is spent on management compared to 36% on production.

The study cites significant corporate inefficiencies, including Uber, which exhausted its 2026 AI budget within four months without shipping a usable feature. Researchers argue that companies often treat AI adoption as a vanity metric based on usage and seat counts rather than tangible business outcomes.