The Hidden Cost of AI 'Workslop' in the Workplace

The Invisible Tax of AI-Generated Work

Somewhere in your organization, an employee is submitting work that appears complete but isn't. Elsewhere, another employee is quietly fixing it. In today's workplace, that second person is often a millennial, and their job description almost certainly doesn't mention this responsibility.

This flawed output has a name. Researchers at BetterUp Labs and the Stanford Social Media Lab coined the term "workslop" to describe it: AI-generated content that looks like solid work but lacks the substance needed to move a project forward. Examples include a polished presentation containing an invented statistic, a report that mirrors a document's structure while missing its central argument, or code that executes flawlessly yet produces the wrong result.

The Real Cost of Cleanup

The scale of this problem is far from trivial. A survey of 1,150 full-time U.S. desk workers found that roughly 40 percent received workslop within a single month. Among those affected, they estimated that 15.4 percent of all materials crossing their desks qualified as workslop.

Resolving each incident takes an average of two hours. When researchers priced that time against respondents' reported salaries, the cost came to approximately $186 per month for each affected employee. At a company with 10,000 employees, that translates to more than $9 million annually in lost productivity. The researchers emphasize that this figure represents a conservative floor, not a ceiling, because it doesn't account for employee turnover or the strain on working relationships.

The burden becomes even clearer when measured in time. Sixty-six percent of professionals now spend at least six hours each week correcting AI-generated output. That's most of a standard workday, dedicated to restoring documents to the quality they should have had when they arrived.

Who Bears the Burden

This correction work doesn't fall evenly across the workforce. It lands on those with enough domain expertise to recognize the errors but insufficient positional authority to decline the cleanup task.

In a typical white-collar hierarchy, that means senior individual contributors and first-line managers. These are the people with roughly eleven years of experience who can immediately spot an incorrect figure, often reporting to someone who never reviewed the underlying data. According to the Stanford and BetterUp research, most workslop arrives laterally from peers, though a significant portion also flows upward from direct reports to their managers.

The structural issue is that fixing workslop produces no visible output. The colleague who generated the flawed presentation still has their deck, which shipped and now appears in the official record as completed work. Meanwhile, the person who spent two hours correcting it has a document that now says what it should have said from the start. At performance review time, one employee has a portfolio of deliverables while the other has only the knowledge that the original asset was deficient.

Millennial women, in particular, have spent their entire careers absorbing the coordination work that keeps teams functioning yet never appears in promotion materials. Generative AI has industrialized this dynamic, transforming what was once an occasional inconvenience into a continuous, unrelenting demand on their time.

The Cost Of Saying So

Speaking up about flawed AI output carries a specific reputational risk in today's workplace. With roughly 60 percent of companies now requiring AI use in some form, pushing back on a colleague's AI-generated draft is often interpreted not as feedback on the work itself, but as resistance to the corporate mandate itself. In a labor market where visible enthusiasm for AI functions as a proxy for adaptability, that distinction can be costly.

Yet the frustration has to go somewhere. According to the research, recipients of AI-related criticism reported feeling annoyed 53 percent of the time and offended 22 percent of the time. Approximately half of those surveyed said they viewed the person offering the critique as less creative, capable, and reliable as a result. A further 42 percent rated them as less trustworthy. In effect, interpersonal trust is being spent to keep an adoption metric looking healthy on paper.

The organizational toll is equally measurable. Analysis of employee reviews found that AI-related commentary was significantly more negative than overall review sentiment, and that negativity correlated with lower firm productivity. Separately, Workday research found that employees reported 37 percent of the time AI saved them was offset by time spent correcting, rewriting, or clarifying weak output.

Put The Repair On The Ledger

The individual response is straightforward and worth adopting: log the hours. This is not about keeping a grievance file, but about recording cleanup work as a line item, attributed to its source and presented at performance review as scope. That simple act reframes the work from a personality trait to a measurable workload, and workload is the only frame a budget responds to.

The organizational response belongs to leadership. Someone in a position of authority has to state plainly that a document arriving fast and wrong is worse than one arriving slow and right, and then hold that position when the deadline lands on a Thursday.

Millennials entered this workforce as the generation that was going to be replaced by automation. They were not replaced. They were reassigned, without discussion, to quality control for a technology their employers purchased on the promise that it would not need any.