How Workslop Became the Office Version of AI Slop

AI slop was supposed to be out there.

In the feed. In search results. In the strangely glossy Facebook image of a seven-fingered grandmother holding a cake shaped like Jesus.

Then it got a job.

It started showing up in inboxes, slide decks, reports, project plans and documents that looked finished until somebody actually tried to use them.

That is the idea behind workslop: AI-generated work that has the appearance of completed work without enough of the thinking underneath it.

The awkward part is that workslop can look pretty good.

That is exactly why it works.

Slop puts on a blazer

The term AI slop became useful because it captured something people were already seeing online: generative AI had made content cheap to produce, and cheap production made it easy to flood a space with material that was technically content but not necessarily worth anyone’s attention.

Workslop moves the same complaint inside an organization.

In September 2025, researchers from BetterUp Labs and Stanford Social Media Lab introduced the term in Harvard Business Review for AI-generated work that appears polished but does not meaningfully advance the task.

The important point was not simply that AI sometimes produces bad work. Offices managed that before artificial intelligence.

The interesting part was where the unfinished thinking went.

The sender saved time generating the thing.

The receiver inherited the job of figuring out what the thing actually meant.

That is the defining workslop move.

A 20-slide presentation arrives with clean headings, tidy bullets and suspiciously even paragraph lengths. It looks like somebody spent an afternoon on it.

Then you read slide four.

Then slide seven.

Then slide twelve.

And slowly realize that nobody has actually answered the question.

The deck is dressed for work. The thinking stayed home.

The problem is not that AI wrote it

This distinction matters.

An email does not become workslop because someone used ChatGPT to help write it.

A report is not workslop because Claude summarized a pile of notes.

Code is not automatically workslop because an AI model generated part of it.

If the output is accurate, useful, reviewed and appropriate for the task, the term does not fit very well.

Workslop describes a different pattern: AI makes it easier to produce something that looks ready before the person sending it has done enough work to make it useful.

That is why polished presentation can make the problem harder to spot.

Before generative AI, unfinished work often looked unfinished.

Rough notes were rough. A half-formed idea had gaps. A first draft sounded like a first draft.

Large language models are very good at removing those warning signs.

They can add headings, smooth transitions, fill empty space and turn three vague thoughts into 900 confident words.

The document becomes more finished-looking without necessarily becoming more finished.

AI did not invent shallow thinking.

It just got very good at typesetting it.

The person receiving it gets the homework

BetterUp and Stanford’s original U.S. survey involved 1,150 full-time desk workers. About 40% said they had received workslop in the previous month, and respondents reported spending roughly two hours dealing with an incident.

More than half also admitted that at least some of the work they themselves sent could qualify as workslop.

That last finding is revealing.

Workslop is not a story about a small group of uniquely lazy coworkers.

People can dislike receiving it and still send it.

The incentive is easy to understand.

You have a document due in twenty minutes.

The AI can produce something respectable-looking in two.

The cost to you is small.

The cost of understanding, checking or rebuilding the output may be paid later by somebody else.

That is where workslop stops being merely a quality problem and becomes a social one.

The receiver may have to work out which claims matter, reconstruct missing context, verify facts, ask follow-up questions or redo the task altogether.

In the BetterUp research, recipients also reported frustration and declines in how capable, reliable and creative they considered the sender.

That makes workslop different from ordinary AI slop on the internet.

If a generated article in your feed is useless, you can scroll past it.

If your manager sends you a useless AI-generated strategy document at 4:47 p.m., the scroll-past option is somewhat less robust.

Productivity for whom?

This is why the word caught on.

It turns a broad argument about “AI productivity” into a sharper question:

Whose time was actually saved?

Imagine one employee uses AI to turn a few minutes of thought into a five-page memo.

Their personal productivity looks excellent.

But if several coworkers each spend half an hour trying to understand what the memo is asking them to do, the organization may not have removed much work.

It has redistributed it.

That idea continued beyond the original 2025 research. Harvard Business Review returned to workslop in January 2026, examining why employees produce it and how pressure to use AI without enough guidance can encourage low-effort, polished output.

The Guardian and Financial Times also used the term in later workplace coverage.

A new social rule seems to be taking shape:

Using AI to save your own time is fine. Saving your own time by quietly spending somebody else’s is less impressive.

Workslop is really about missing judgment

The obvious response to workslop might be: use less AI.

That misses the more useful point.

The researchers behind the term distinguish low-agency AI use from more deliberate use in which people remain responsible for the output.

The difference is judgment.

Did you decide what matters before asking for the deck?

Did you check whether the answer addresses the question?

Did you remove the paragraphs that sound intelligent but do nothing?

Did you give the next person enough context to act?

Those tasks existed before generative AI.

The machine simply makes it easier to skip them without leaving obvious fingerprints.

That is why workslop is such a useful piece of language.

It does not mainly describe an AI failure.

It describes a human handoff failure made easier by AI.

From content pollution to cognitive pollution

AI slop and workslop are close relatives, but they pollute different spaces.

AI slop makes the public information environment noisier. There is more to scroll through, more to filter and more material competing for attention.

Workslop does something similar inside a team.

It fills shared space with documents that require interpretation but do not repay it.

Both come from the same new imbalance:

generating something has become much cheaper than deciding whether it deserves to exist.

On the open internet, the result is another post.

At work, the result may be another meeting.

Which, depending on your office, may be the more serious threat.