Artificial intelligence is saving office workers time, but a large share of that time is being spent fixing what the AI produced, according to research released by Workday on January 14. The human resources and finance software company found that roughly 40 percent of the hours employees save with AI tools are absorbed by rework: correcting errors, rewriting content and verifying results.
The study surveyed employees at 3,200 companies across North America, Europe and Asia, each with annual revenue of at least $100 million, Workday said.
Time saved, then spent again
On the surface the numbers look encouraging. About 85 percent of respondents said AI saves them between one and seven hours a week. But only 14 percent reported consistent, clearly positive results from using it. For most, the saved time has become a new kind of obligation.
Gerrit Kazmaier, a senior executive at Workday, described the pattern as a productivity paradox: the employees who use AI most are also the ones spending the most time checking, correcting and reorganizing its output. Speed has increased, in other words, but trust has not kept pace.
Workday estimates that about 37 percent of the time saved goes directly into repairing AI-generated work. Even so, more than 90 percent of daily AI users said they are optimistic about the technology’s future, while 77 percent said they scrutinize AI output as closely as they would a colleague’s work, and sometimes more closely.
The cost of “workslop”
The findings add weight to a term that gained currency in 2025: “workslop,” meaning AI-generated material that looks polished and professional but is thin or wrong underneath. Workday’s analysis suggests that in a single large organization this kind of output can cost millions of dollars a year, because employees spend their time improving work that appears finished but does not hold up.
Organizations have not caught up
Workday attributes much of the problem to a lack of organizational preparation. In 89 percent of the companies surveyed, fewer than half of job roles had been updated to reflect what AI can now do. Although 66 percent of leaders said skills training should be a top priority, only 37 percent of the employees carrying the heaviest rework burden had received practical training.
The age breakdown is striking. Workers aged 25 to 34 are the heaviest AI users and also account for 46 percent of the rework burden. By contrast, among employees who reported genuine benefits from AI, 57 percent redirected their saved time toward higher-value work and 79 percent had received formal training.
Lessons for employers
The results echo earlier academic work. Studies from MIT and Harvard Business School have found that AI can lower productivity in the early stages of adoption until organizations redesign workflows, responsibilities and decision-making around it.
Workday’s conclusion is that adding AI as a tool is not enough. Without training, clear governance and new ways of working, the speed it offers can turn into an additional burden. The problem, the report argues, is not the capability of the technology but unplanned use of it. Companies that fail to close that gap risk inventing a new way to waste time rather than saving it.
