Workslop refers to AI-generated deliverables that look professional but lack substance. They create the illusion of completed work while forcing recipients to check, correct and redo, resulting in a net loss of time and money.

How is workslop different from simply sloppy work?

Sloppy work still takes some effort from its author. Workslop takes none: it is enough to accept an AI output and pass it on. The effort is not saved, it is shifted to the recipient, who has to decode, correct or redo it.

How much does workslop cost a company?

According to the BetterUp Labs and Stanford study, the direct cost reaches roughly 186 dollars per month per affected employee, or more than 9 million dollars a year for an organisation of 10,000 people. On top of this come indirect costs: loss of quality, demotivation and eroded internal trust.

Why is workslop a problem beyond the time lost?

Beyond the financial loss, it undermines the sender's credibility. About half of recipients judge the author to be less creative, capable and reliable, and more than four in ten trust them less. This domino effect weakens collaboration and organisational cohesion.

What is the difference between a 'pilot' and a 'passenger' of AI?

The pilot uses AI as a co-pilot: they revise the output and adapt it to context. The passenger simply forwards the output without checking. The first amplifies their added value; the second produces workslop and burdens colleagues with correction work.

Which sectors are most exposed to workslop?

Professional services, under pressure to produce impressive deliverables quickly on complex topics, and the technology sector, where code that compiles without meeting the specification, or generic documentation, easily passes for finished work.

Should you ban AI to avoid workslop?

No. Imposing or banning AI uniformly is counterproductive. What works are explicit guardrails: define use cases, train people in prompting and post-processing, set validation rules and a culture of transparency about AI use.

How do you put 'AI hygiene' in place in an organisation?

By treating AI output to the same review standards as human deliverables: thoughtful use of the tool, systematic validation against context, clear AI and human zones, training in post-processing, and transparency about usage. These habits turn AI into a productivity lever rather than a source of false gains.

Workslop: the hidden cost of AI in business

The use of generative AI has spread massively across business, and yet the return on investment remains elusive for the overwhelming majority of organisations. The MIT NANDA programme, in its State of AI in Business 2025 report, establishes that close to 95% of organisations deploying AI see no measurable gain. A study by BetterUp Labs and the Stanford Social Media Lab, published in the Harvard Business Review in September 2025, offers an explanation as concrete as it is unexpected: workslop. The term describes AI-produced content that has every appearance of careful work but lacks the substance to actually move the task forward. Far from saving time, it costs time — and that time has a price.

For a CIO or CTO, this is not a management anecdote. It is a blind spot in AI governance: we measure adoption, rarely the quality of the deliverables it produces. Yet that is precisely where the value leaks away.

The mechanics of an invisible tax

Workslop is not ordinary carelessness. Sloppy work still demands a minimum of effort; workslop demands none. It is enough to accept an AI output and forward it. Here lies the perverse effect: the effort is not saved, it is displaced. Whoever receives the document has to decode what was produced, spot what is missing, reconstruct the absent context, then correct or redo it. The cognitive load slides from sender to recipient, often without the sender being aware of it.

Monday morning. A project manager opens a fifteen-page report received the day before: structured headings, clean paragraphs, colourful charts. After twenty minutes of reading, doubt sets in — the figures do not match the target market, the competitors analysed operate on another continent, the recommendations are generic to the point of being unusable. The author wanted to “save time”. The result: several hours of rework, a clarification meeting, and sometimes the task handed to someone else.

This dynamic is part of a wider phenomenon, that of producing content with no added value, which AI makes trivially easy by turning the tool into a pretext for avoiding the thinking and contextualisation that quality work demands.

What it actually costs

The study puts a figure on this invisible tax. Each workslop incident costs an average of 1 hour 51 minutes of handling time to the person who receives it — about twenty minutes more than if the sender had done the work themselves. Set against the salaries reported by respondents, this amounts to roughly 186 dollars per month per affected employee (about 170 euros). For an organisation of 10,000 people, given an estimated prevalence of 40% of employees affected each month, the annual bill exceeds 9 million dollars (around 8 million euros) in lost productivity.

Infographic of the invisible tax of workslop: 40% of employees affected each month, 1 hour 51 minutes per incident, about 186 dollars per month per affected employee, more than 9 million dollars a year for 10,000 people, and one employee in two judging the sender less reliable.

These orders of magnitude are calculated on American salaries and working hours. To find out what it represents for your own situation — your headcount, your salaries, your working time — the simplest approach is to enter your own figures into a cost calculator rather than reasoning from an imported average.

The erosion of trust

The financial cost is only the visible part. The study documents a more difficult damage to repair: the deterioration of how colleagues regard one another. About half of those who received workslop judged the sender to be less creative, less capable and less reliable than before. More than four in ten now trust them less, and a third say they want to work with them less.

This distrust is particularly corrosive in professions where human added value is supposed to be at the heart of the offering — consulting, strategy, design. When an obviously AI-generated deliverable arrives for the third time, it is not just a document that is devalued, it is a professional reputation. And trust, once dented, is slow to rebuild.

The most exposed sectors

Two worlds concentrate the risk. Professional services first, under a double pressure: to produce impressive deliverables quickly, on complex subjects. A consultant can generate a fifty-page report on a digital transformation in minutes — one that says nothing relevant about the client’s actual situation.

The technology sector next, a victim of its own enthusiasm. Code that compiles but does not meet the specification, roadmaps disconnected from technical reality, documentation so generic it becomes unusable: the appearance of output masks an absence of real work.

Pilots and passengers: two relationships with the tool

The study distinguishes two user profiles, and the gap between them is decisive. “Pilots” treat AI as a co-pilot: they generate drafts they rework, seek perspectives they evaluate, automate the repetitive to focus on value. Their AI usage is about 75% more frequent than others’, yet paradoxically less visible — because the final result is contextualised and relevant.

“Passengers”, by contrast, see AI as a way to avoid the work. They forward outputs without rereading them, hoping it “passes”. Ironically, they are the ones who create the most work for everyone else. The same tool therefore amplifies two opposite behaviours: it multiplies the rigour of some and the carelessness of others.

How to avoid the trap

The study’s conclusion is counter-intuitive for many leadership teams: imposing AI everywhere does not reduce workslop, it makes it worse. A blind mandate pushes passengers to produce even more hollow content. What works are explicit guardrails and a usage framework. A few principles structure effective “AI hygiene”.

The first is a simple matter of discipline. Think before prompting, by clarifying the real objective. Systematically review the output against context. And take ownership: whoever’s name is on the document answers for it, AI or not. The second is to delineate what belongs to AI and what belongs to humans. First drafts, rephrasing and the analysis of structured data lend themselves well to automation; strategic decisions, personalised feedback and ambiguous situations call for human judgement.

The third principle concerns training. Knowing how to prompt is not enough; you also need to know what to do with the result — assess its relevance, adapt it to context, spot hallucinations and biases. Post-processing is a skill in its own right, too often neglected. The last principle is transparency: rather than hiding the use of AI, which leads straight to workslop, it is better to own it openly. Stating that a text was structured with AI and then adapted to the specific context preserves trust while still benefiting from the tool.

For an IT department, the stake goes beyond training teams: it is about treating an AI output to the same review standards as a human deliverable. As long as generated content escapes quality control, it keeps circulating — and keeps costing.

The hidden cost of convenience

Workslop is a reminder of a truth that the promise of AI tends to obscure: there is no shortcut to quality. AI can speed up processes, relieve tasks, open up perspectives. It replaces neither thought, nor context, nor judgement. It is an amplifier, not a substitute.

The parallel with other forms of hidden debt that weighs on IT budgets is striking: here as elsewhere, today’s apparent saving turns into tomorrow’s sinkhole. Investing in AI without investing in governing its use is to create the perfect conditions for a workslop epidemic. At the scale of a large organisation, that is a luxury few can afford.