Study: AI 'Workslop' Costs U.S. Workers Productivity

Researchers at Harvard Business Review’s BetterUp Labs and Stanford’s Social Media Lab identified “workslop” — AI-generated emails, reports and slides that look polished but lack the substance to meaningfully advance work. In a survey of 1,150 U.S. full-time employees, 40% said they’d received workslop in the past month, and recipients spent an average of about 1 hour 56 minutes fixing or interpreting each instance, an invisible tax the authors estimated at roughly $186 per worker per month (about $9 million annually for a 10,000-employee firm). Beyond lost time, workslop damaged morale and trust — over half said it made them annoyed, many were confused or offended, roughly half viewed senders as less capable, and managers often had to heavily revise or redo AI-produced work. The researchers said the problem flows laterally and up and down organizations, hitting professional services and tech especially hard, because large language models tend to produce fluent but sometimes inaccurate or unclear outputs that transfer effort downstream. That dynamic helps explain why widespread AI adoption hasn’t translated into gains — other studies, including work cited from MIT, find most firms see no measurable ROI — and experts urged firms to pair generative tools with stronger guardrails and human review so AI augments rather than clogs workflows.