Workers are wasting half a day each week fixing AI ‘workslop’

AI 'workslop' is forcing employees to work an extra four and a half hours each week to clean up mistakes, according to new research.

A survey of more than 1,100 US enterprise AI users from Zapier found that while 92% of workers say AI boosts their productivity, the average employee spends more than half a workday revising, correcting - and sometimes completely redoing - AI-generated outputs.

Three-quarters of respondents reported at least one negative consequence from low-quality AI outputs, including work rejected by stakeholders (28%), security incidents (27%), and customer complaints (25%).

Zapier noted that just 2% of respondents don’t need to revise what AI produces.

A key factor here lies in poor training, researchers noted. Employees without AI training are six times more likely to say AI makes them less productive.

While untrained workers spend less time on AI cleanup, they also report fewer productivity gains: just 69% say AI helps, compared with 94% of trained workers.

“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 at 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.”

The worst AI workslop offenders

Data analysis tops the workslop list, according to Zapier, with 55% saying data analysis and visualization projects require the most cleanup, followed by writing tasks at 46%.

Meanwhile, engineering, IT, and data roles average five hours per week fixing AI outputs, with 78% reporting negative consequences. Finance and accounting teams face the highest rate of negative consequences at 85%, averaging 4.6 hours of cleanup per week.

The time lost in fixing AI-generated outputs has a significant impact on bottom lines, according to Zapier. Workers spending more than five hours a week on AI cleanup tasks are more than twice as likely to report lost revenue, clients, or deals.

Zapier said better data quality and more robust infrastructure could go a long way in helping to improve the situation.

The study found that those with access to AI orchestration tools and comprehensive company context – so internal documentation, brand guides, project templates, or prompt libraries – said the technology does have a big positive impact on productivity.

“The solution isn’t fewer tools, it’s better infrastructure,” said Mabie. “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.”

AI workslop is here to stay

The rise of AI workslop has become a recurring pain point for enterprises ramping up adoption of the technology.

A report from MIT researchers last year found that more than 40% of US-based workers had been given AI-generated content that “masquerades as good work but lacks the substance to meaningfully advance a given task”.

This, the study noted, was destroying productivity and harming perception of the technology in the workplace.

Certain professions are experiencing acute issues with this trend, such as those working in software development. A recent study from CodeRabbit, for example, shows that AI makes 1.7 times as many mistakes as human programmers.

The use of AI in software development has been one of the leading use-cases for the technology over the last three years, with developers reporting significant productivity boosts from AI code generation.

Research from Harness in early 2025, however, found that these productivity gains are offset by the fact developers are having to drop tools and manually remediate faulty code, slowing down processes.

FOLLOW US ON SOCIAL MEDIA

Make sure to follow ITPro on Google News to keep tabs on all our latest news, analysis, and reviews.

You can also follow ITPro on LinkedIn, X, Facebook, and BlueSky.