What is workslop?
Engineering culture and software practice
Workslop is AI-generated work product that looks finished, polished and plausible but lacks the substance to advance the task, so that the effort of making it useful lands on whoever receives it. The term was coined in September 2025 by researchers from BetterUp Labs and Stanford's Social Media Lab in a Harvard Business Review article. It names the office version of AI slop: the memo that reads beautifully and says nothing.
What this means
Workslop is what you get when someone uses an AI tool to produce something that looks like good work, sends it on, and leaves the real thinking to you. The document is well formatted and confident. It just does not do the job. Opening it, you feel a moment of confusion, then the sinking realisation that you will have to redo it.
The key idea is that the effort has not disappeared; it has moved. The sender saved ten minutes. The receiver loses time working out what the document was supposed to say and fixing it. Multiply that across a team and the organisation is busier but not better off.
The word borrows from "slop", the general term for low-value AI-generated content. Workslop is the workplace strain of it: not spam on the internet, but the report, the slide deck or the email that a colleague sends you.
Why it matters
Workslop explains a puzzle many leaders are living through. Investment in AI tools is up, individual staff feel more productive, and yet team throughput does not improve. If polished-but-empty work is being passed around, the apparent productivity of the sender is cancelled by the hidden cost to everyone downstream.
It also damages trust, which is harder to repair than a single document. When people learn that a colleague's output routinely needs redoing, they start checking everything that colleague sends, which is slower for all concerned. Reputations take the hit.
For a leader, the danger is inadvertently encouraging it. A blanket instruction to use AI, with no guidance on quality, tells staff to produce AI content whether or not it helps. The point is not less AI; it is AI used with judgement, where the person sending the work still owns its quality.
As of September 2026 the term is about a year old and usage is still settling, but the underlying pattern, effort shifted from producer to receiver, is not new and is easy to recognise once named.
How it works
Where the term came from
Workslop was coined in a Harvard Business Review article published on 22 September 2025 (updated 25 September 2025) by Kate Niederhoffer, Gabriella Rosen Kellerman, Angela Lee, Alex Liebscher, Kristina Rapuano and Jeffrey T. Hancock, working across BetterUp Labs and the Stanford Social Media Lab. They defined it as AI-generated work content that masquerades as good work but lacks the substance to meaningfully advance a task.
The article reported a survey of 1,150 US-based full-time employees across industries. Forty percent reported having received workslop in the last month, and those who had spent an average of 1 hour and 56 minutes dealing with each instance; the researchers estimated a cost of about 186 US dollars per employee per month, which for a 10,000-person firm works out at over 9 million dollars a year. These figures were widely repeated in the business press.
The term spread fast, which is the test of whether it will last rather than fade. Mainstream outlets picked it up within days, and practitioner and management writing through 2026 adopted it as a standard label. Some commentators noted that the coining article sat close to commercial interests, a fair caution to keep in mind, but the word itself filled a genuine gap and has been used well beyond its origin.
The mechanism
Workslop works because modern AI is very good at surface polish and only sometimes good at substance. The output has the shape of competent work: headings, confident prose, a tidy structure. What it lacks is the specific judgement, context and accuracy the task needed. Because it looks finished, the receiver does not immediately distrust it, and only discovers the emptiness after investing time. The reader's time is the hidden cost that the sender's time-saving quietly created.
The trust effect on teams
The study's most striking finding is social, not financial. Roughly half of those surveyed viewed colleagues who sent workslop as less creative, capable and reliable; 42 percent saw them as less trustworthy, and 37 percent saw them as less intelligent. This matters because collaboration runs on the assumption that what a colleague hands you is genuinely done. Workslop breaks that assumption, and once broken it is slow to rebuild.
How this differs from good AI-assisted work
The difference is review, judgement and accountability. In healthy AI-assisted work, a person uses the tool to draft, then reads the result critically, corrects it, adds the context only they have, and takes responsibility for what they send. Workslop skips those steps: the draft is forwarded as if it were finished. The tool is the same; the discipline is missing.
Examples
At a mid-sized retailer, a manager asks for a competitor analysis. A team member generates one in minutes and sends it on. It looks thorough, but the figures are generic and some are simply wrong, so the manager spends an afternoon rebuilding it from real data. The ten minutes saved cost half a day, and the manager now reads that person's work with suspicion.
In a council department, a policy officer uses AI to draft responses to residents. The letters read smoothly but miss the specific circumstances of each case, so a senior colleague has to rewrite them before they go out. The department's output looks high, but the useful output has not risen at all, and residents nearly received answers that did not address their questions.
At a software team of eight, by contrast, an engineer uses AI to draft release notes, then checks each line against what actually shipped, cuts the vague parts and adds the detail users need. The notes go out reliable and clear. Same tool, opposite result, because a person owned the quality before pressing send.
Common misunderstandings
The first misunderstanding is that workslop is the same as an AI hallucination. It is not. A hallucination is false content; workslop can be entirely accurate and still useless, because the problem is not that it is wrong but that it does not advance the task and pushes the real work onto the reader. Something can be factually correct and still be workslop.
The second is that workslop means AI is bad at work. Rather, it names a way of using AI badly: sending unreviewed output as if it were finished. The same tool, used with judgement, produces the opposite.
The third is that only junior staff produce workslop. Leaders produce it too, and their workslop travels further because people are reluctant to push back. A vague AI-drafted strategy note from the top can waste far more time than a junior's report.
The fourth is that the fix is simply to add a review step and tell everyone to check AI output before sending. That helps but is not sufficient on its own, because the incentive to save time by offloading effort remains; norms and accountability matter as much as a checkbox.
The fifth is that measuring workslop is impossible. It is hard, not impossible: rework rates, time spent fixing received work, and how often documents bounce back are all visible signals.
Risks and boundaries
This is a recent term, coined in September 2025, and its usage may still shift; this article was written in September 2026. It is easy to over-apply. Not every imperfect document is workslop, and calling colleagues' work workslop can become a way of dismissing anything AI-touched. The concept specifically names polished output that lacks substance and transfers effort to the receiver, not any work that happens to have used AI.
The term also does not settle the productivity question by itself. The claim that AI is destroying productivity is stronger than the evidence supports; what the research shows is a mechanism by which gains can leak away, not proof that they always do. The honest reading is that workslop is a real and recognisable failure of how AI is used, and one worth managing, rather than a verdict on AI as such.
What to do next
Set a simple norm for AI-assisted work: say what was AI-generated, check it before sending, and own the content as if you had written every word. The person who presses send is accountable for quality, full stop.
Replace blanket mandates to use AI with guidance on where it helps and what good looks like. A target of more AI usage invites workslop; a standard of useful, checked work does not.
Watch the right numbers. Track rework, review time and how often received work has to be redone, rather than raw output volume. Rising activity with flat results is the signature of workslop.
Model it yourself. If leaders forward polished but empty AI drafts, staff learn that this is acceptable. If leaders visibly review and take ownership of their own AI-assisted work, that standard spreads.
FAQs
Who coined the term workslop?
Researchers from BetterUp Labs and Stanford's Social Media Lab, in a Harvard Business Review article published on 22 September 2025.
Is workslop just another word for AI slop?
It is the workplace version. Slop is low-value AI content in general; workslop is specifically the polished-but-empty work product that a colleague sends you, shifting the effort onto you.
Can accurate work still be workslop?
Yes. Workslop can be factually correct and still useless, because the defining problem is that it fails to advance the task and pushes the real work downstream.
How much does workslop cost?
The coining research reported that recipients spent an average of 1 hour and 56 minutes per instance, and that 40 percent of the 1,150 surveyed workers had received it in the previous month.
How do we stop workslop without banning AI?
Set norms that the sender checks and owns the content, guide staff on where AI genuinely helps, and measure rework rather than output volume.
Does workslop hurt teams beyond wasted time?
Yes. The research found 42 percent of recipients saw the sender as less trustworthy and about half as less capable, which damages collaboration and is slow to repair.
How do we measure workslop?
Look at rework rates, time spent fixing received work, and how often documents are sent back for redoing.