‘AI workslop’: Zapier survey finds AI delivers gains as well as costly corrections

Zapier has published survey findings that point to a gap between reported productivity benefits from workplace AI and the time staff spend correcting its output.

The survey of more than 1,100 US enterprise AI users found that 92% of workers said AI boosted their productivity. At the same time, the average employee reported spending 4.5 hours per week revising, correcting, or redoing AI-generated work.

Zapier described the issue as "AI workslop". The company defined it as output that appears polished but lacks substance, precision, or context. Zapier said the rework created a hidden layer of effort in many organisations.

The research found that only 2% of respondents said they generally did not need to revise what AI produced. It also found that 58% of enterprise workers spent time revising AI outputs.

Most respondents reported knock-on effects. Zapier said 74% had experienced at least one negative consequence from low-quality AI outputs. The survey listed work rejected by stakeholders at 28%, security incidents at 27%, and customer complaints at 25%.

Where time goes

The survey results suggested that cleanup demands varied by task type. More than half of respondents, 55%, said data analysis and visualisations required the most revision work. Writing tasks followed at 46%.

The findings also pointed to differences by department. Engineering, IT, and data roles reported an average of five hours per week fixing AI outputs. In those functions, 78% said they had experienced negative consequences.

Finance and accounting respondents reported the highest rate of negative consequences at 85%. They also reported an average of 4.6 hours of cleanup work per week.

Zapier linked heavier cleanup workloads with commercial outcomes. Workers who spent five hours or more per week on AI cleanup were more than twice as likely to report lost revenue, clients, or deals. The survey put that figure at 21% for the group with five hours or more of cleanup time, compared with 9% for other respondents.

Training gap

The survey highlighted a difference between trained and untrained staff. Zapier said trained employees were six times more likely to see productivity gains. The survey also found that workers without AI training were six times more likely to say AI made them less productive. It put the figures at 6% for untrained workers and 1% for trained workers.

Reported productivity benefits also differed. The survey found that 69% of untrained workers said AI helped them, compared with 94% of trained workers.

Zapier said trained employees spent more time with AI and more time fixing AI output. It linked that behaviour with heavier use and use in higher-stakes contexts.

One context sentence before a quote: Zapier attributed the reported difference in outcomes to governance and investment decisions inside organisations.

"The productivity gains from AI are real. 92% of workers feel them. But so is the cleanup work," said Emily Mabie, Senior AI Automation Engineer, Zapier. "The companies seeing the best results aren't the ones avoiding AI. They're the ones who have invested in training, context, and orchestration tools that turn AI from a sloppy experiment into a managed process."

Tools and context

The survey asked respondents about access to what Zapier called orchestration tools. Zapier reported that 97% of workers with access to such tools said AI boosted their productivity. The survey also reported the same figure, 97%, for workers with access to AI orchestration tools.

Zapier also pointed to documentation and standard materials inside organisations. It said workers with "comprehensive company context", including internal documentation, brand guides, and project templates, reported a 96% productivity boost.

Training resources also correlated with reported benefits. Zapier said 95% of workers with access to prompt libraries and ongoing training said AI made them more productive.

"The solution isn't fewer tools, it's better infrastructure," Mabie added. "Orchestration, training, and proper context convert AI from a vague experiment into a managed process where the extra cleanup is the cost of doing more meaningful work faster, rather than the cost of pretending you are."

Enterprise steps

Zapier set out recommendations for employers based on the survey. It said companies should make AI training mandatory and build "AI-ready context" into tools. It also called for formal review flows for higher-risk outputs and for organisations to track time spent on cleanup as an internal metric.

The company also recommended investment in orchestration platforms and tighter workflow governance. Zapier said such measures could change how AI gets used across functions where errors can trigger security issues, rejected work, or customer complaints.

The survey was conducted by Centiment for Zapier between November 13 and November 14, 2025. Respondents worked at companies with 250 or more employees, and all were screened as US AI users. Zapier said the margin of error was approximately plus or minus 4% for the overall sample at a 95% confidence level.