Workslop Is Not an AI Problem. It Is a Leadership Problem.

I have spent more than 20 years running technology organizations, and I have never seen a tool adopted faster than generative AI. I have also never seen one so easily mistaken for progress.

Here is what I mean. A document arrives. It has clean headers, a confident summary, and a neat set of recommendations. It takes ten seconds to open and forty minutes to realize it does not actually move anything forward. Someone now has to send it back, fix it, or start over.

That is workslop. In my experience, it is almost never the fault of the tool.

What workslop actually is

Researchers at BetterUp Labs and Stanford's Social Media Lab introduced the term in a September 2025 Harvard Business Review article. My working definition is simple: workslop is AI-generated output that looks finished but lacks the context and judgment needed to be useful.

What matters is the mechanic. Workslop saves the sender time by moving the real work to the recipient. The sender gets to mark the task done. The recipient inherits the job of working out what is missing, what is wrong, and what was actually meant. The effort does not go away. It just lands on someone else's desk.

That is why I do not see it as a quality problem. Weak work has always existed. What has changed is how cheap it is to produce something that looks like strong work, and how invisible the cost is to the person producing it.

How big the problem is

The HBR research put numbers on what most leaders already sense. Of about 1,150 full-time U.S. employees surveyed, 40% said they had received workslop in the past month, and each incident took close to two hours to resolve. The researchers estimated the cost at about $186 per employee per month, which comes to roughly $9 million a year for a company of 10,000 people.

The hidden cost is reputational. People who received workslop saw the senders as less creative and less reliable. The sender saves an hour and spends credibility they did not know they had on the line.

Why "use AI more" is not a strategy

The usual response is to treat workslop as a training gap: better prompting, a style guide, a lunch-and-learn. Those help at the margins, but they miss the cause.

In most organizations, workslop is the predictable result of a management decision. Leadership says "use AI more" and changes nothing else. Not what good work looks like. Not how work gets reviewed. Not what gets rewarded.

Teams then do what any system trains them to do. They optimize for what is measured, and the easiest thing to measure is volume. More drafts, more decks, more summaries, faster. Adoption numbers climb. What quietly disappears is the part nobody was measuring: the thinking, the checking, the judgment. The researchers pointed to exactly this, naming blanket "use AI everywhere" mandates without guidance as a key driver.

I have implemented AI frameworks, tools, and processes across multiple client organizations, and the lesson has been the same every time. Giving people access to AI is easy. Redesigning work so they use it responsibly is the hard part.

At a direct-to-consumer collectibles company, the ChatGPT integration into our Microsoft tools was the quick part. The real work was sitting down with the general counsel to build AI and data governance processes that told people what the tool was for and what it was not for. At a higher-education services firm, we used AI to improve how client data was ingested. That worked because it sat inside a redesigned process with clear owners, not on top of an old one. Both times, it was the management design, not the technology, that decided whether AI made the work better or just made more of it.

Four management moves that reduce workslop

None of these require a new tool. All of them require a leader to make a decision.

1. Name which deliverables require human authorship. Not everything should be AI-first. A board memo, a client recommendation, or a performance review carries judgment that someone has to own. Say so explicitly. When everything is fair game, nothing is protected.

2. Make reviewer time a visible cost. Workslop survives because its cost lands on someone else's calendar. If a manager spends two hours reworking a draft, that time should show up in a retro, a project review, or a direct conversation. Once people can see the transfer, they stop treating it as free.

3. Require the sender to state what they verified. One line at the top of a deliverable changes behavior. For example: "I checked the figures against the Q2 close. I did not verify the vendor claims." The sender either does the verification or admits they did not. Either way, the recipient knows what they are holding.

4. Build AI into how work is scoped, not how it is graded. The time to decide how AI fits into a piece of work is at the start. Decide what a model can draft, what needs a person, and where the checkpoints sit. If AI only comes up when someone asks "did you use AI on this?" after the fact, the chance to design the work well is already gone.

The quieter risk: polish suppresses override

There is a second-order effect that concerns me more than the lost hours.

Polished output does not just waste reviewer time. It changes how people act around it. When something looks finished, with confident language, clean formatting, and tidy conclusions, people defer to it. They are less likely to push back or ask whether it is actually right.

I have written before about how convincing AI-generated explanations can sound whether or not they are correct. The same thing happens here. The better the surface, the more it discourages the human override that good organizations depend on. A rough draft invites challenge. A polished one discourages it.

So the real danger is not only the time workslop wastes. It is the portion that gets through because it looked too finished to question.

The standard, not the policy

Most companies respond with a tool policy: approved platforms, acceptable use, data rules. Those matter, and I have written plenty of them. But they do not fix workslop, because workslop is not misuse of the tool. It is ordinary use of the tool inside a system that rewards the wrong thing.

What fixes it is a leadership standard. That means being clear about what good work looks like, making the cost of review visible, expecting people to own what they send, and designing AI into the work from the start instead of bolting it on at the end.

AI can make a team dramatically more effective. It will not decide what "good" means for you. That was always a leadership job, and it still is.