Workslop: when AI-assisted work shifts the effort onto the recipient

A polished document can still be unfinished work

A specialist sends over twenty slides with headings, charts and crisp conclusions. The material looks great. The problem only emerges when the director wants to make a decision: it is unclear where the figures come from, what period the comparisons cover and which recommendations matter. She has to reconstruct the basis of the analysis herself. The author saved time, but the recipient did the missing part of the work. It is exactly this shifting of work onto the recipient that is now sometimes called workslop.

It is not a scientific diagnosis of a person, nor a synonym for any content generated by AI. The problem existed before chatbots came along: in reports padded with flourishes, presentations that were too general, or notes with no decisions in them. AI may simply make it easier to produce such material at scale. That is why quality has to be judged by whether the recipient can use the output to take a specific action, not by the tool used to prepare the first draft.

Time saved by the author is not time saved by the whole team

In a 2023 experiment by Noy and Zhang, participants who completed specific writing tasks with ChatGPT worked faster on average, and the quality of their output was rated higher than in the control group. That is important evidence that AI can help. It does not mean, however, that every company presentation automatically saves time for everyone further down the line. The study looked at specific tasks under experimental conditions, not at any strategic report you care to name.

In a company, costs also arise during handover, review and revisions. If the person writing a report finishes in twenty minutes but three recipients then spend an hour hunting for sources, one person’s saving can mean extra work for the whole team. So the point is not to write every word by hand. The point is to check quality at the stage where the author knows the context best and can verify the content before burdening others with it.

What’s usually missing is an answer to: “What am I supposed to do with this?”

Workslop tends to show up in three places. First, the material describes the topic but doesn’t answer the recipient’s question. Second, it presents figures or claims without sources that can be checked. Third, it sets out recommendations without selection criteria or a person responsible. In every case, the recipient has to go back to the beginning: define the problem, check the facts or clarify the decision. That is not a minor stylistic correction.

A simple test before you hit send: “Does the recipient know what decision they need to make on the basis of this material, where the data comes from and what the next step is?” If the answer is unclear, the work isn’t finished. For example, a report on three suppliers should apply the same comparison criteria, be open about missing information and justify its recommendation. An elegant slogan such as “Choose a partner for the future” is no substitute for price, risk or implementation terms.

Agree what “ready to use” means before you start producing material

It is worth agreeing as a team that material for a decision contains five things: a purpose, an audience, sources, verified conclusions and a recommended decision with its limitations. For simple tasks a shorter version will do; for more important ones you need dates, the scope of the analysis and how the data was checked. Not every file needs a multi-page report on how AI was used. The standard should be proportionate to the consequences of getting it wrong. Sensitive data and the rules for using tools must be respected however good the writing is.

A principle of accountability works well: the person whose name is on the material is responsible for having checked the key claims and for understanding their own recommendation. If several people are working on it, agree who signs off the figures, who signs off the content and who signs off the message to the client. That way, no one can dilute responsibility with “that’s just what the AI came up with”. Technology can support content creation, but it doesn’t change the fact that someone has to decide to use that content.

Feedback should be about the gaps, not about labels

A manager reads an incomplete analysis and says: “You’ve dumped AI text on me again – you couldn’t even be bothered to think.” The employee starts defending how they work instead of filling in the data. It is better to say: “In the report, I can’t see sources for two of the figures, and I don’t know why you’re recommending option B. I need that information before I can make a decision.” Feedback like this addresses the real problem, not the author’s supposed intentions.

If the problem keeps recurring, agree together on a minimum standard for handing work over and a deadline for improvement. You can ask what makes it hard for the author to check their work thoroughly: too little time, no access to the data, an unclear goal or the mistaken belief that volume is what counts most. Some of the causes lie in management, not in one person’s attitude. Instructions like “use AI for everything” are particularly damaging when the company does not, at the same time, explain what it considers good-quality output.

Check how much extra work the material demands of the recipient

Instead of counting only the slides generated or the minutes spent writing, look at how many questions the recipient has to ask before they can act. If every presentation needs a separate meeting to explain it, the problem may lie in the handover standard. After a few weeks, assess whether the corrections concern style, data, purpose or the logic of the decision. This is not an off-the-shelf scientific measure of workslop, but a practical tool for improving the process.

It is also worth distinguishing a draft from a final version. You can deliberately send a “working outline for discussion” and ask others to help refine it. The problem arises when such a draft is presented as a finished result. Clearly labelling the stage of the work reduces tension: colleagues know whether they are meant to assess the line of thinking or verify a finished recommendation. AI can support both stages, but confusing them erodes trust.

Workslop is not caused by the use of AI itself. The problem is offloading onto others work that the author should have done. Material that looks finished should have a purpose, a verified basis, a clear recommendation and an accountable author. Before you criticise the technology or the person, name the gaps and agree a handover standard. Real time savings only become visible when the recipient can act without having to reconstruct the entire analysis.

How Empatyzer helps you talk about workslop without accusations

A team leader receives her third report this month in which attractive language doesn’t translate into a decision that can be checked. Instead of publicly criticising the employee for using AI, she describes the specific gaps to Em: no sources, an unclear selection criterion and the extra questions she had to ask. She asks Em to help prepare feedback that shows how the problem affects the work without judging the author’s intentions. In the context of this relationship, Em can take into account the available tips on how best to deliver criticism. She then plans a short team conversation about what “ready to use” means for an analysis, a presentation and a message to a client. Em helps her phrase rules that make sense to people with different working styles, as well as questions about the real obstacles. Em also helps her set out clear requirements, give factual feedback and agree how analyses will be verified. This way, the time the author saves can turn into a real benefit for everyone involved in the work.