Parochial surveys are starting to address worker output that appears complete but often require extra inputs from colleagues, impacting productivity, morale.
As organizations race to adopt AI, a survey involving Stanford University suggests that the promised productivity revolution remains elusive, and in some ways, AI could be slowing things down.
Detailed in Harvard Business Review, the survey report has coined the term “workslop” to describe AI-generated materials that appear polished but are hollow, and necessitate extra work for colleagues to improve upon — exacting heavy costs on organizations and morale.
Workslop includes well-designed slides, lengthy reports, summaries, or code that, despite its neat packaging, is vague, incomplete, or contextually off-base. Instead of saving time, it shifts complex decision-making to recipients, who are forced to interpret, correct, or even redo the work, creating a ripple of inefficiency across the workplace.
Out of 1,150 full-time US knowledge workers surveyed, 40% had cited encountering workslop in the past month, and AI-generated content now accounts for about 15% of all their work-related material. Unlike traditional “sloppy work”, which at least takes some human effort, workslop can be mass-produced almost effortlessly with a single AI prompt.
The AI productivity paradox
Real-world stories highlight the cost: One finance worker described agonizing over whether to rewrite an AI-generated document, request a revision, or simply accept inferior results. Another in retail detailed spending hours of research and extra meetings redoing reports that looked fine at first glance but were ultimately useless.
According to a separate MIT Media Lab report, 95% of respondent cited seeing no real return on investment from their AI spending, even though AI tools and processes are now common in their workplaces — presumably a gap created by superficial mandates and a lack of clear guidelines for using AI productively.
Still, this characterization comes from a snapshot in time involving a specific subset of workers, and does not necessarily represent the broader global workforce or all industries. In fact, the Stanford researchers acknowledge that poor-quality work — regardless of whether AI is involved — has long existed in various forms, and that the phenomenon is not entirely new.
