More than half of desk workers admit sending sloppy AI-generated work to colleagues — but do we know if frontline staff do the same?

This explores how often workers using AI day to day hit three recurring problems: sloppy AI output handed to them by colleagues, AI errors and hallucinations, and fear of job loss. One caveat first. Almost all the measurement in the corpus comes from surveys of desk workers and AI power users, not frontline staff. The closest frontline evidence is customer-service agents taking over from chatbots. Treat the numbers below as office-worker numbers.

Workslop is the best-measured of the three, and it's common. In one survey more than half of desk workers (52.7%) admit to sending AI-generated content that lacks substance. Recipients lose about 3.4 hours a month to it, and it lowers their trust in whoever sent it How much do desk workers actually suffer from AI workslop?. A related survey puts the share of incoming work that is unhelpful AI content at about 15%. Each instance takes nearly two hours to deal with, which is longer than if the sender had just done the work themselves How much work that employees receive is actually unhelpful AI content?. The cost doesn't vanish. It moves from the sender to the recipient.

Hallucinations show a counterintuitive pattern: the more you use AI, the more errors you report. Heavy users (6+ hours a week) report hallucinations three times as often as casual users and spend ten times longer on revisions. The likely reason isn't that the tools get worse. Heavy users attempt harder tasks and hold the output to higher standards Do heavy AI users actually encounter more hallucinations?. Part of the problem is also built into the models. Even frontier models are much worse at meeting needs users leave unstated, which is exactly where real work gets messy Why do AI models struggle with unspoken user needs?. For frontline service, the Taobao field experiment shows what happens after the AI fails. A human taking over fixes technical mistakes, but not a customer who is already frustrated, and late handoffs don't recover at all When human workers take over from AI, does it actually fix the problem?.

Job-loss fear also rises with use, but not in a straight line. Gallup's panel of 30,000 workers finds daily AI users report more than twice the job-elimination fear of infrequent users Does frequent AI use make workers fear job loss more?. Anthropic's survey of 81,000 Claude users finds a U-shape. Worry peaks both among people AI slows down and among people it speeds up the most, and it is lowest for people who see no change Does AI productivity gain always ease job displacement fears?. Exposure is also unevenly spread. In female-dominated occupations it reaches lower-paid, lower-skilled workers, who have the fewest resources to adapt Does AI exposure hit low-wage workers harder in some fields?. That group is closer to the frontline population this question asks about.

What ties the three together is that management changes how much of each problem workers feel. Supportive managers cut the job-fear gap by 6 to 11 points Does frequent AI use make workers fear job loss more?. Microsoft attributes about two-thirds of AI's impact to organizational factors such as culture and incentives, rather than individual effort Why do ready workers struggle to transform their work?. Meanwhile 61% of leaders see employee anxiety rising while investment in reskilling lags Why aren't organizations investing in reskilling despite AI job displacement fears?. The open gap: nobody in this corpus has surveyed frontline workers (retail, warehouse, care, field service) on all three problems directly. If that's the group you care about, the honest answer is that it hasn't been measured yet.