AI is cutting hours of office work, but also creating a new kind of busywork

A new survey of 6,000 digital workers across the United States, United Kingdom, and Australia reveals AI tools are saving employees roughly 11 hours per week by automating tasks. However, workers spend over six hours weekly on 'botsitting'—correcting AI errors, refining inputs, and troubleshooting flawed outputs, according to research by the Work AI Institute, which includes academics from Stanford and UC Berkeley. The study, sponsored by AI company Glean, found that while 75% of workers report increased productivity, only 13% of organizations translate these gains into measurable business growth. The discrepancy highlights a gap between personal efficiency and company-wide impact, with much of the time saved by AI absorbed by additional oversight. Workers spend 37% of their AI interaction time fixing errors, while 36% is spent on productive tasks, the survey found. Over a third of AI sessions fail entirely, forcing workers to restart or rework outputs. The report cites an example of a junior software engineer who deployed AI-generated code with undetected errors, burdening senior colleagues with fixes. The study’s lead author, Paul Leonardi of UC Santa Barbara, noted that employees are effectively acting as managers for AI tools, handling tasks like gathering documentation and ensuring accuracy. This 'invisible labor' layer risks undermining productivity gains, as workers offload judgment to AI but still bear responsibility for its outputs. The findings align with recent Silicon Valley trends, where companies push employees to maximize AI use despite unclear returns. Uber, for instance, exhausted its 2026 AI budget in four months without delivering a usable feature, underscoring the challenges of scaling AI adoption. The report concludes that while AI boosts personal productivity, organizations must address the hidden costs of oversight to fully realize its potential.