Overview

• In the past month, 40% of U.S. desk workers say they received AI-generated output that looked finished but failed to move work forward, with recipients estimating 15.4% of incoming content fits this pattern.
• Employees report spending about one hour and 56 minutes fixing each instance, which researchers estimate costs roughly $186 per worker per month, or about $9 million annually in a 10,000-person company.
• Reputation takes a hit as roughly half of respondents view senders as less creative, capable and reliable, with 42% reporting reduced trust, 37% perceiving lower intelligence and about one-third flagging incidents to teammates or managers.
• The problem cuts across hierarchies and sectors, flowing peer-to-peer as well as up and down the org chart, with professional services and technology cited as hardest hit.
• Coverage urges leadership-led fixes that include clear AI usage standards, human review before external sharing, real AI literacy training and organizational guardrails, while an MIT Media Lab report cited in the reporting finds 95% of organizations see no measurable AI ROI.