AI Workslop: Why Your Team's AI Output Is Creating More Work, and the One-Task Fix
If AI was supposed to save time, why are people spending their afternoons fixing documents a colleague generated in two minutes? Harvard Business Review has a name for it: workslop. This page explains what it is, why it happens and how a small firm can stop it, by defining one task at a time and pricing it honestly.
What is AI workslop?
The term comes from a September 2025 Harvard Business Review article, "AI-Generated 'Workslop' Is Destroying Productivity", by researchers at BetterUp Labs working with the Stanford Social Media Lab. They define workslop as "AI generated work content that masquerades as good work, but lacks the substance to meaningfully advance a given task."
In plain terms: a report, slide deck, email or summary that looks finished but isn't. The sender saved time. The recipient now has to work out what's wrong, then either send it back or redo it. The effort hasn't disappeared; it's moved to someone else.
The HBR authors present it as one possible reason so many companies report heavy AI use but little measurable return, a pattern also discussed in our write-up of the MIT GenAI Divide report.
Why workslop happens
- Mandates without standards. The HBR authors point to leaders encouraging AI use broadly while giving little guidance on quality. "Use AI more" isn't a standard.
- Unclear tasks. If nobody has defined what a good client summary or proposal looks like, AI fills the gap with generic content.
- Speed is visible, rework isn't. The sender's time saving shows up immediately. The recipient's rework rarely gets counted.
- Low ownership. When AI writes the first draft, people sometimes feel less responsible for what's in it.
The researchers contrast "pilots", who use AI deliberately to improve their work, with "passengers", who use it to avoid work. Most people are capable of both; the setup of the task tends to decide which shows up.
What workslop looks like in a small firm
- The meeting summary that lists every topic discussed but not the decisions, so everyone has to ask what was agreed.
- The proposal with three pages of generic benefits and no mention of the client's actual problem.
- The research note that cites figures nobody can trace.
- The client email that is perfectly polite and answers a slightly different question.
- The internal report twice as long as last month's, with the one number the manager needed buried on page four.
None of these is a disaster on its own. Together, across a week, they quietly eat the hours AI was meant to save, usually from the most experienced people, who end up as the final filter.
The real cost: it moves, it doesn't vanish
Workslop is a cost-accounting problem. If you measure only the sender's time, AI looks like a big win. Measure the whole task, including the recipient checking, querying and fixing, and the picture can reverse. That's why we always price one finished task: the cost of the work until it's actually usable, across everyone who touches it. Our AI ROI calculator includes review and rework time for exactly this reason.
The one-task fix
You don't fix workslop with a company-wide memo. You fix it one task at a time:
- Pick the task where rework hurts most. Often client summaries, proposals or internal reports.
- Map it. Write the steps, who does each, and where AI is used. See our process mapping guide.
- Define "done". A short standard: what it must include, how long it should be, what must be verified.
- Make the sender the first reviewer. Before passing anything on, the sender checks it against the standard and can explain every part.
- Share a good example and a template prompt. So AI starts from your standard, not a blank request.
- Measure the whole task. Time from start to usable, for everyone involved, before and after.
Once one task is clean, move to the next. Within a few months the habits spread.
Is it workslop? A 30-second check
Use this on a piece of AI-assisted work you've received. It's a prompt for a conversation, not a score to hold against anyone.
What leaders can do
The HBR article's recommendations are worth following: model purposeful AI use yourself, set clear norms and guardrails, frame AI as a collaborative tool rather than a shortcut, and hold AI-assisted work to the same standard as any other. For a small firm, that translates into a short AI policy, a defined standard for each common task, and no praise for speed that creates rework downstream.
Avoid the opposite reaction too. Banning AI because of workslop throws away real gains on tasks where it works well. The aim is fewer, better uses, not none. Our guide to rolling out AI your team won't resist covers how to introduce those norms without resentment.
Not sure whether your team has a workslop problem? Take the AI readiness assessment on one process, or let us map it with you.
Frequently asked questions
Is workslop the same as AI slop?
It's the workplace version. AI slop usually means low-quality AI content online; workslop is the same thing passed between colleagues inside a business.
What is AI workslop?
A term from a 2025 Harvard Business Review article for AI-generated work that looks good but lacks the substance to move a task forward, shifting the effort onto whoever receives it.
Who coined the term workslop?
Researchers at BetterUp Labs, working with the Stanford Social Media Lab, introduced it in a Harvard Business Review article published in September 2025.
Why is workslop a problem for small businesses?
In a small team, every hour of rework lands on a few people, often senior ones. It can erase the time AI saves and strain trust between colleagues.
How do you stop AI workslop?
Define what good output looks like for each common task, make the sender check AI work before passing it on, share templates and measure the whole task including rework.
Should we ban AI to stop workslop?
No. Bans lose the real gains and push use onto personal accounts. Set standards and ownership instead.
How do I measure workslop?
Time the whole task until it's usable, across everyone who touches it, and compare it with the time before AI was used.