What Is Workslop? Why AI Output Is Breaking Collaboration Between Colleagues
Workslop is AI-generated work that looks finished but does not move the task forward, so the colleague who receives it has to interpret, fix, or redo it. Four in ten US full-time employees received workslop in the past month (Niederhoffer et al., HBR, 2025). It is not a technology failure. It is a failure of collaboration between people.
TL;DR
- Workslop is polished-looking AI output that shifts the real work onto the person who receives it.
- 40% of 1,150 surveyed US employees received workslop in the past month (Niederhoffer et al., HBR, 2025).
- The larger cost is social: about half of recipients saw the sender as less capable afterwards, and 42% as less trustworthy (same study).
- Workslop damages all three conditions of collaboration described by Victor Wekselberg: aligned goals, compatible attitudes, and mutual knowledge of competencies.
- The fix is organizational: named ownership, standards set by the receiver, and metrics that count accepted work, not volume.
What is workslop?
Workslop is AI-generated work content that looks complete but leaves the real thinking undone. Researchers from BetterUp Labs and the Stanford Social Media Lab, who coined the term, define it as AI-generated content that "masquerades as good work, but lacks the substance to meaningfully advance a given task" (Niederhoffer et al., HBR, 2025). Think of slides that support no decision, or summaries that never reach a conclusion.
The defining feature of workslop is not low quality. It is displaced effort. The sender saves time by skipping the thinking, and the receiver pays by decoding, checking, or redoing the work.
How common is workslop, and what does it cost?
Workslop is common, and it flows in every direction. In a 2025 survey of 1,150 US full-time employees across industries, 40% said they had received workslop in the past month (Niederhoffer et al., HBR, 2025). Most of it travels between peers (40% of cases), but 18% is sent up to managers and 16% comes down from managers to their teams.
The direct cost of workslop is time: nearly two hours per instance. The researchers estimate the loss at $186 per employee per month, or more than $9 million a year for a 10,000-person organization (Niederhoffer et al., 2025; BetterUp, 2025). That is a model-based estimate, not an audited figure.
The indirect cost of workslop is trust, and it lasts longer than the lost hours (Niederhoffer et al., 2025):
- 53% of recipients felt annoyed, 38% confused, and 22% offended.
- About half saw the sender as less creative, capable, and reliable.
- 42% saw the sender as less trustworthy, and 37% as less intelligent.
- 32% were less likely to want to work with that person again.
These are US data. We know of no comparable published measurement for Poland or Central and Eastern Europe, so treat them as a reference point.
Why does AI make people busier instead of saving time?
AI lowers the cost of producing output, but not the cost of absorbing it. When output becomes cheap, the bottleneck moves to the person who reads, checks, and decides.
An eight-month ethnographic study of about 200 employees at a US technology company found that generative AI intensified work rather than reducing it (Ranganathan & Ye, HBR, 2026). Employees worked faster, took on a broader scope of tasks, and stretched work into more hours, often without being asked. Engineers, in turn, spent more time reviewing and correcting AI-assisted work from colleagues: the downstream burden behind workslop, observed inside a single firm.
In the EY 2025 Work Reimagined Survey of 15,000 employees in 29 countries, 64% reported a perceived increase in workload over the past year, while only 5% used AI in advanced ways that transform their work (EY, 2025). The survey does not attribute the workload rise to AI. It shows that high AI use and heavier workloads now coexist.
Why is workslop a collaboration problem, not an AI problem?
Collaboration is a social process between people. In the cooperative theory of groups developed by organizational psychologist Victor Wekselberg, it rests on three conditions that must hold at the same time: aligned goals, compatible attitudes, and mutual knowledge of competencies, that is, knowing what others can actually do (Wekselberg, 1999). Workslop damages all three at once.
| Condition of collaboration | What workslop does to it | What to measure |
|---|---|---|
| Aligned goals | The sender optimizes for looking productive; the receiver needs the task to move. Goals split at the handoff. | Do goals reward accepted outcomes or produced volume? |
| Compatible attitudes | Receivers lose trust. Teams divide into those who treat AI output as finished work and those who treat it as a draft. | Attitudes toward AI-assisted work across roles and levels |
| Mutual knowledge of competencies | Polished output hides what the sender actually knows. About half of recipients downgrade their view of the sender's capability. | How accurately people know who can verify what |
Wekselberg also separates collaboration from exchange. In an exchange, the parties agree on little beyond keeping the exchange going, and every other goal stays private to each side (Wekselberg, 1999). Workslop is exchange dressed up as contribution: a file changes hands, and the real work stays where it was.
AI did what it was asked to do: drafting and formatting, which is coordination-layer work (see Where does AI fit without replacing human judgment?). The failure happened between two people: one who forwarded the output and one who had to absorb it.
What causes workslop in organizations?
Workslop is a predictable response to how organizations introduce AI, not a character flaw. Four causes recur:
- Mandates without quality standards. The researchers point to blanket mandates to use AI for everything, paired with little guidance on what good output looks like (Niederhoffer et al., 2025).
- Activity metrics. When dashboards count prompts, documents, and tickets, people produce prompts, documents, and tickets.
- Diffused ownership. "The model wrote it" becomes a way to send work that nobody fully stands behind.
- Standards set by the sender. Nobody asks the receiver what "ready to use" means.
These are change-management failures. Polish AI leaders rate change management at 2.42 out of 5, the lowest of five dimensions of AI maturity (AI_Managers™ 2026). Tools get deployed; norms do not.
Restricting AI is not the answer. Restrictions push AI use out of sight, where it cannot be governed (see Shadow AI: The Invisible Half of Your AI Adoption).
How can you tell AI-augmented work from workslop?
AI-augmented work and workslop can look identical. The difference is who did the thinking and who carries the cost.
| Dimension | AI-augmented contribution | Workslop |
|---|---|---|
| Who did the thinking | The sender; AI sped up the drafting | The AI; the sender forwarded it |
| What the receiver gets | An input ready for a decision | A puzzle to decode |
| Verification | The sender checked facts, logic, and fit | No one checked |
| Disclosure | States what AI drafted and what was verified | Silent |
| Effect on trust | Strengthens the sender's reputation for competence | Erodes it |
| What gets measured | Accepted outcomes and rework | Output volume |
How do you stop workslop? A 90-day plan
Stopping workslop means changing the conditions that make it rational. The plan follows the six RECODE Method dimensions.
1. Redesign Work (days 1–15). Run a baseline pulse on how often people receive AI-assisted work they cannot use. Map the ten most frequent handoffs and let each receiver write a one-line "ready to use" standard.
2. Establish Ownership (days 16–30). Make the sender the named owner of everything they send, whoever or whatever drafted it. Add a two-line disclosure: what AI drafted, and what the sender verified.
3. Connect Your Data (days 31–45). Ground AI tools in your templates, past decisions, and approved data, so drafts carry context instead of generic filler.
4. Operationalize Value (days 46–60). Replace activity metrics with downstream ones: first-pass acceptance rate and rework time on received work.
5. Develop Your People (days 61–75). Train editing and verification, not just prompting. Managers model the standard: 16% of workslop comes from them.
6. Engineer to Scale (days 76–90). Build the standards into templates and review steps. Repeat the baseline pulse to see whether prevalence and trust have moved.
Frequently asked questions about workslop
What is the difference between workslop and AI slop?
AI slop is low-quality AI content in public spaces, such as social media feeds. Workslop is its workplace form: AI output that a colleague hands you as finished work. The difference matters because workslop has a sender and a receiver, so it damages a working relationship, not just an information stream.
How do I know if I am sending workslop?
Apply the receiver test. Before sending AI-assisted work, ask whether the recipient can act on it without decoding, fact-checking, or redoing it. If you could not defend every claim in it during a meeting, it is not ready. State what AI drafted and what you verified, so the recipient knows what they can rely on.
Should companies restrict AI use to stop workslop?
No. Restrictions push AI use out of sight rather than improving it, and invisible use cannot be governed. The research points to mandates without quality guidance as a driver of workslop. The fix is clear standards, named ownership, and metrics that reward accepted work, not bans that reward hiding.
Who is responsible for workslop, the employee or the organization?
Both, at different levels. The sender owns every output they forward, whoever drafted it. The organization owns the conditions: mandates without quality standards and metrics that reward volume make workslop the rational choice. Fixing individual behavior while keeping volume-based incentives guarantees that the problem returns.
Can better prompting turn AI into a real teammate?
No. There is no such thing as human-AI collaboration. Collaboration is a social process between people, built on aligned goals, compatible attitudes, and mutual knowledge of competencies. AI can be given goals and context, but it is not a member of the team in the social sense. Better prompting improves drafts. People decide what happens to them.
Sources
1. Niederhoffer, K., Kellerman, G. R., Lee, A., Liebscher, A., Rapuano, K., & Hancock, J. T. (2025). AI-Generated "Workslop" Is Destroying Productivity. Harvard Business Review, 22 September 2025. https://hbr.org/2025/09/ai-generated-workslop-is-destroying-productivity
2. BetterUp (2025). The hidden costs of workslop. https://www.betterup.com/blog/hidden-costs-workslop
3. Ranganathan, A., & Ye, X. M. (2026). AI Doesn't Reduce Work—It Intensifies It. Harvard Business Review, 9 February 2026. https://hbr.org/2026/02/ai-doesnt-reduce-work-it-intensifies-it
4. EY (2025). EY survey reveals companies are missing out on up to 40% of AI productivity gains due to gaps in talent strategy (EY 2025 Work Reimagined Survey), press release, 17 November 2025. https://www.ey.com/en_gl/newsroom/2025/11/ey-survey-reveals-companies-are-missing-out-on-up-to-40-percent-of-ai-productivity-gains-due-to-gaps-in-talent-strategy
5. AI_Managers™ 2026. Self-assessment of AI maturity by Polish AI leaders (N=100+), January 2026.
6. Wekselberg, V. (1999). A "Blueprint" of the Cooperative Theory of Groups. Polish Psychological Bulletin, 30(1), 15–45.
Continue reading
Where Does AI Fit Without Replacing Human Judgment?
AI fits in the coordination layer: gathering evidence, drafting options, monitoring processes. Human judgment stays where a decision is irreversible, contested, or attributable. Three tests, one decision table, and a 90-day plan for drawing the line.
Shadow AI: Why Half of Your AI Adoption May Be Invisible to You
Shadow AI is AI use without your knowledge, approval, or governance. Gallup's undivided 52% hides it, the 47%-vs-25% plan gap feeds it, and only an instrument built on psychological safety can measure it.
What Is Pseudo-Collaboration? How to Tell Whether Your Teams Actually Work Together
Full calendars, friendly teams, smooth handoffs — and still no collaboration. Four look-alikes get mistaken for the real thing, and AI deployed on top of them only automates the illusion.