# 'Workslop' isn't productivity. It's a tax.

AI was supposed to do the busywork. In a lot of teams it does the opposite: it generates plausible-looking output that a human downstream has to detect, decode, and redo. Researchers named it 'workslop,' and the numbers are ugly — 53% of desk workers say they've received it, each instance costs ~2 hours to fix, and it quietly poisons trust between coworkers. It's not a productivity gain. It's a productivity transfer — and someone downstream is paying the bill.

- **Author:** Fedor Chishchin
- **Published:** 2026-07-01
- **Tags:** business, methodology
- **Source:** https://fedorthinks.com/blog/workslop-is-a-tax

---

The pitch for AI at work was simple: it takes the busywork off your plate. In a lot of teams, the
reality is the exact inversion — AI puts _more_ work on someone else's plate, dressed up so it looks
done. Researchers at Stanford and BetterUp gave it a name: **workslop.**

## What workslop is

Workslop is AI-generated output that looks polished but is hollow — a report, a slide, a PR
description, a message that _reads_ finished but doesn't actually do the job. It passes the glance
test and fails the read test. So the person who receives it has to stop, figure out what's real,
reconstruct the missing substance, and often redo it. The sender felt productive. The work didn't
disappear; it moved.

The [Harvard Business Review study](https://hbr.org/2025/09/ai-generated-workslop-is-destroying-productivity)
puts numbers on the transfer: **52.7% of US desk workers admit to sending workslop, 38% report
receiving it, and each instance costs about 1.9 hours to sort out.** Scaled up, [one estimate pegs
the cost at ~$9M a year for a 10,000-person organization](https://futurism.com/future-society/ai-productivity-research)
in lost time alone.

> A "productivity gain" that just relocates the work downstream isn't a gain. It's an invoice you
> mailed to a coworker.

## The part that doesn't show up in the numbers

There's a second bill, and it's worse: trust. Per the same research, **53% of people feel annoyed and
22% feel offended** when they get obvious AI slop — and it lowers how capable and reliable they judge
the sender to be. That's the expensive part. Speed you can recover. But once your colleagues learn
that your output needs re-checking, every future thing you send starts from suspicion. You spent your
credibility to look fast.

## Why it happens — and how to not be the source

Workslop is what you get when the goal quietly shifts from _done_ to _done-looking_. The model is
happy to produce confident, complete-seeming text about anything; if "produced output" is the bar, it
clears it every time. The fix is to move the bar back to where it belongs:

- **Ship outcomes, not artifacts.** The deliverable is a decision made, a bug fixed, a question truly
  answered — not "a document exists." If it doesn't move the work forward, it's slop no matter how
  clean it looks.
- **You are the reviewer, not the forwarder.** If you didn't read it closely enough to defend every
  line, you're not sending work — you're forwarding a draft and calling it done.
- **Ground it.** Slop thrives where there's no source of truth to check against. Tie the output to
  real facts, real data, a real spec — the same discipline that keeps
  [an AI honest instead of merely confident](/blog/grounding-is-a-constraint).
- **Optimize for the reader's time, not yours.** The whole win of AI is supposed to be less total
  work. If you saved ten minutes and cost the next person two hours, you didn't win — the team lost.

This is the same trap as [cheap code being the most expensive code](/blog/cheap-code-is-expensive):
output got free, so people started measuring the wrong thing. Volume of AI text is not throughput.
Work that someone else has to redo is negative throughput.

## The bottom line

Workslop is the tax you pay for confusing "the AI produced something" with "the work is done." It
feels like leverage to the sender and lands as a bill on the receiver — in hours, and in trust.

**Don't send work you haven't made real. AI that just moves the effort downstream isn't productivity —
it's a tax with your name on the return address.**