Michelle Volberg sent an email to her staff about six months ago with a simple ask: “Please stop talking to me like a robot.”
Volberg, the founder and CEO of tech recruiting company Twill, tells me that some workers had increasingly been sending Slack messages and emails they had copied and pasted from Claude. She suspected a bot was to blame, because the pings were more reminiscent of pitches from clueless external consultants than her employees. Instead of hearing from her thoughtful staff, she says, she suddenly received messages that sounded “annoyingly optimistic and dumb.” The punctuation and capitalization all looked perfect — more formal than a typical Slack message might — but upon closer review, the messages lacked context about clients and included hallucinations.
It all began when Volberg rolled out an enterprise Claude subscription to her company of 10. Volberg’s policy now is simple: use AI to brainstorm and work through ideas, but use your own voice for internal communications. “If you wouldn’t say it that way in person, email, or Slack — don’t send it.”
She noticed a shift immediately. “Creative thinking came back, interpersonal relationships came back, so much of the human humanness of working together, whether it’s in person or remote, came back,” she says.
A year ago, researchers from the Stanford Social Media Lab and professional training and coaching company BetterUp Labs coined the term “workslop,” referring to low-quality, extra verbose, AI-generated memos, documents, emails, and slide decks that white-collar workers had begun foisting onto one another. Now, even more emboldened workers are shoveling it up to the C-suite.
CEOs are somewhat complicit — the slop they’re now wading through may have accidentally been unleashed by mandates to use AI. New policies attempt to revive some workplace humanity and parse productive uses of AI from the deluge of wasteful content that distracts from real work. They’re also learning that rules alone may not fix the underlying causes of the mess. Some are even firing workers for slopping it up.
“Workslop is predicted by the conditions more than it is in individual behavior,” says Kate Niederhoffer, chief scientist at BetterUp Labs, adding that bosses need to set up workplaces where people feel safe to “have an effective dialogue about how these tools can really transform the way we work.”
While you’re living by the move-fast-and-break-things ethos to figure out AI, your boss is tired of moonlighting as the slop janitor.
Some slop oozes from lazy workers trying to pass the friction of tasks onto others. But much of it has proliferated as people feel pressure to use AI tools without guidance from the bosses pushing its adoption. Tokenmaxxing, the trend that fell nearly as quickly as it rose, led workers to spin up as much content as they could with AI. Companies have begun measuring AI use in performance reviews, although there’s little standardization across how its use should be evaluated. AI tools haven’t achieved their builders’ promises of supercharged productivity. Some who use them have taken on longer hours and more tasks, which can lead to lower-quality output. People also defer to AI’s judgment, trusting its authoritative tone and speedy delivery, and may not realize that they’re shipping slop.
In Niederhoffer’s initial workslop research published last fall, more than half of those surveyed admitted to sending it, noting that some of their work was “unhelpful or low quality.” The staggering number, she says, reveals that the problem runs deep, and workers’ actions may signal three things: compliance with AI adoption, innovation, or that they’re depleted.
Many of the anti-slop policies companies introduced over the past year try to straddle the line between supporting AI use and drawing some boundaries around experimentation. In August, software startup Clay’s cofounder Varun Anand posted the company’s new AI writing policy on LinkedIn, and it quickly went viral by LinkedIn standards. It’s not a ban on AI in writing, but it requires workers to stand behind every idea in every sentence and qualifies that creating a document should take longer than reading it does. The policy also specifies that writing is a useful part of the thinking process and brevity is often best. “Readers will appreciate hearing your thoughts, even at the expense of supposed ‘polish,'” it reads.
There’s a risk of losing creative thought and judgment when workers use AI to cognitively offload tasks. Spellbook CEO Scott Stevenson told his team in early September that he would rather see five scrappy bullet points when reading an idea proposal than five perfectly formatted pages. He also specified that communication with customers should always be “hand-written,” a term now thrown around to refer to writing that is typed out by hand rather than generated.
Workslop can waste time and erode trust among teams. Eric Fischgrund, who runs public relations company FischTank PR, says he bans the use of AI to write content sent to reporters or clients and relies on an AI detector to check. Last year, he fired one worker who continuously used AI for off-limits purposes. The first time resulted in a confrontation; the second, a warning; and yet the person did it for a third time, Fischgrund says. “It’s obvious, it’s annoying.”
ZocDoc also put out anti-AI slop guidance in August, encouraging people to “never ship your slop,” make bot-to-human interaction transparent, and to maintain original ideation. “If you give this responsibility to AI, it will do it poorly and your value will decline,” the statement reads. Jessica Aptman, chief communications officer at ZocDoc, tells me that the new guidance is “less punitive” for those who churn out slop and more prescriptive about maintaining their own value as they use AI. “Over time, people who are not going to take these expectations seriously will see performance decline — just in how much people trust their work, how much influence they have, how much value they’re giving if they’re outsourcing all of their smarts and creativity to the robots,” Aptman says.
Bosses who want to do away with slop may need to do more than ask for workers to speak as themselves. They need to shape office cultures where people feel their input matters, and trust that AI won’t absorb or devalue their work. It may be more effective to discourage spilling slop than it is to mop it up.
