How much work that employees receive is actually unhelpful AI content?

BetterUp Labs and the Stanford Social Media Lab report a September 2025 survey of 1,004 full-time U.S. desk workers. The excerpt defines workslop as "AI-generated content that looks polished and complete — but is actually unhelpful, low-quality, or off the mark." The headline measures are what respondents believed or estimated: 40% of employees believe they received workslop in the last month, and on average workers "estimate that 15.4% of the work they receive is AI workslop." Managers report more exposure (54%, against 38.5% of individual contributors), and 53% admit that at least some of the work they send may be workslop. The excerpt presents this as circulation "in every direction," broken down as sideways to peers (40%), up to managers (19%) and down from leaders (16%). It does not say what those percentages are of.

The cost measure is also self-reported. Employees say they spend "an average of 1 hour and 51 minutes dealing with each instance of workslop, about 20 minutes longer than if the sender had done the work themselves." The excerpt turns that into an "invisible tax" of $186 per month per employee, and for an organization of 10,000 workers, "over $9 million in lost productivity every year." Its only line on mechanism is "Workslop spreads when AI work loses context and accountability," which describes how workslop moves. The excerpt does not test it.

The time side is the recipient's version of a shift that Does AI really save time, or just change how we spend it? describes for the person using the tool: time moves away from the task and toward evaluating and parsing AI output. Workslop moves that kind of time onto whoever receives the output. It also parallels Does AI turn freelance work into validation instead of creation?, where validating replaces producing. The survey counts hours lost, not skill lost, so the two findings address different costs of the same move.

The excerpt does not establish how workslop was identified. The 40% and 15.4% figures are perceptions, not counts from a content audit or a detector, and the excerpt gives no sampling method, question wording or response rate. It does not explain the dollar conversion (wage rate, hours, instances), so the $186 and $9 million figures cannot be checked from the text and should be read as the authors' illustrative estimate. The 20-minute comparison with the sender's own time is also unexplained, and the excerpt does not say what "dealing with" an instance involves. What the source supports is narrower than its headline: one 2025 survey in which respondents estimate that 15.4% of what they receive is workslop, and that dealing with each instance takes them longer than the sender doing the work would. Whether that estimate matches the true share of work is something this source cannot answer.