Survey Reveals Workers Spend Significant Time 'Botsitting' AI, Hindering Business Gains
A recent study found that digital workers allocate as much time managing AI tools as they do producing useful work. While 75% of individuals reported increased productivity from AI adoption, only 13% of organizations observed significant business growth. This discrepancy is attributed to the substantial effort workers invest in correcting AI output, gathering necessary information, and restarting frequently failing AI sessions.
Digital workers are reportedly spending a comparable amount of time managing artificial intelligence (AI) tools as they do on productive tasks, according to a new survey. This phenomenon, termed 'botsitting,' is described as a new chore arising from AI adoption.
The study, published by the Work AI Institute, surveyed 6,000 digital workers across the United States, the United Kingdom, and Australia between December and January. Contributors to the institute include academics from Stanford University and UC Berkeley, with sponsorship from AI company Glean.
While a significant 75% of individuals reported a boost in personal productivity due to AI, only 13% of organizations indicated achieving substantial business gains. This suggests a disconnect between individual-level efficiency improvements and overall company growth.
Paul Leonardi, Duca Family professor of technology management at UC Santa Barbara and a co-author of the study, highlighted that workers are spending over six hours per week 'babysitting' their work chatbots. The survey found that for every hour of useful AI output, roughly another hour is spent making it usable. Of the total weekly interaction with AI, 37% is dedicated to 'botsitting,' while 36% is spent on actual work production.
Reasons for extensive 'botsitting' include the time required to correct AI-generated work, gather essential files, documentation, and tacit knowledge for quality output. Workers also reported that more than a third of AI sessions fail outright, necessitating restarts or significant rework. Furthermore, 41% of workers admitted to delivering AI-generated work they couldn't explain if questioned.
Leonardi suggests that these AI tools are currently expecting individual contributors to act as managers, overseeing AI agents without accounting for the additional management workload. (Source: Slashdot)
