Image courtesy: LinkedIn
A survey by BetterUp Labs and Stanford’s Social Media Lab found that over half of U.S. desk workers admitted to sending workslop, while 38% reported receiving it. On average, employees spent 3.4 hours per month revising AI-generated work, costing organizations millions in lost productivity.
Speaking on The Knowledge Project podcast, Lütke said staff were producing unexamined emails and code with AI, creating what he called “slop grenades” that burden colleagues with cleanup, according to a report by Fortune. “You’re just letting AI do the work for you,” he warned, noting that AI should synthesize points in an email rather than generate long, sloppy missives. Duolingo CEO Luis von Ahn made headlines last year for declaring the company would go “AI-first” — a policy that tied employee evaluations to AI usage, replaced human contractors with AI tools, and limited headcount growth to only what automation couldn’t handle.
Shopify cofounder and CEO Tobias Lütke has shifted his stance on workplace AI use, cautioning employees not to let the technology do all their work. Researchers have coined the term workslop to describe polished-looking but low-value AI output that drags down productivity. Lütke isn’t alone in reversing course.
A survey of 962 full-time American desk workers conducted this year by BetterUp Labs and Stanford’s Social Media Lab found that more than half of respondents (52.7%) admitted to sending workslop to a colleague, a habit that showed up more often in workplaces that actively encouraged AI adoption. More than a third (38%) said they had been on the receiving end of it, spending an average of 3.4 hours a month cleaning it up. This year’s findings mark a worsening trend compared to last year’s version of the same survey, when 40% of respondents reported encountering workslop and said cleanup took roughly two hours on average. That earlier figure translated to about $186 in lost productivity per affected employee each month, or as much as $9 million annually in lost productivity for a 10,000-person organization — a number that’s now almost certainly higher given this year’s longer average cleanup time.
The survey found that workslop erodes trust between coworkers — recipients rated colleagues who sent it as less competent and less friendly, and more than a third of those who’d received it said they’d actively try to avoid working with that person again in the future. By May, though, Von Ahn was telling Fast Company he’d been swept up by how convincing AI-generated writing looked in demos, without accounting for how it holds up at scale compared to human creativity. He pointed to Duolingo’s need to produce large volumes of story content for language lessons, estimating a meaningful share of AI-scaled output ends up being low-quality filler that has to be caught before it reaches users. Researchers now have a name for exactly what Lütke and Von Ahn are describing: workslop — AI-generated output that looks polished and complete on the surface but lacks the substance to actually be useful, forcing someone else to redo the work. Common examples cited included emails with broken links or code that was needlessly overcomplicated. The damage isn’t just measured in wasted hours. Get the latest technology news and updates. Download the TOI App.