AI is freeing up hours for workers each week, but much of that time is being wasted.

      The initial appeal of workplace AI was the promise of time: hours returned to employees overwhelmed by routine tasks. However, a new study indicates that these hours are indeed real, yet many companies are quietly losing them once again.

      Research from Workday, which surveyed 3,200 business leaders, shows that 85% of employees are now saving between one and seven hours a week with AI, but nearly 40% of that time is instantly consumed by rework.

      Rework emerges as a subtle antagonist in the narrative of AI productivity. Although the time saved when an employee uses a model to draft a document or summarize data is tangible, if that output requires checking, corrections, and partial redoing, much of that saved time disappears. Workday’s findings quantify an experience familiar to many knowledge workers: AI may be quick, but the time saved is often spent ensuring its accuracy.

      This trend is evident not only in Workday’s research but also in a broader study for 2026. Researchers from Stanford and BetterUp have labeled "workslop," referring to AI-generated content that appears polished but lacks depth, estimating that 40% of US workers encountered this in a recent month and calculating the cost at millions annually in lost productivity for larger organizations. Although the output may increase, the quality does not always keep pace, and someone further down the line bears the consequences.

      The key structural insight revolves around who reaps the benefits of whatever value remains. PwC’s 2026 AI study found that nearly three-quarters of AI’s economic advantages are accumulating to about a fifth of companies, specifically those leveraging AI as a growth catalyst rather than merely a cost-cutting measure. This concentration is significant, suggesting that the technology does not uplift all firms equally but rather provides compounded advantages for organizations that have undergone the more challenging task of reshaping their operations around it.

      The distinction between these two groups lies more in discipline than in tools. Acquiring AI licenses and allowing employees to save fragmented minutes yields the disjointed, reworked, and ultimately unclaimed gains reflected in Workday’s data. To convert those minutes into measurable value, companies must determine how to utilize the freed-up time, and the study implies that most have failed to make that decision.

      As noted in several 2026 studies, the default response is often to demand increased output from the same workforce, redirecting the saved time into more work rather than improving work quality. In some cases, this leads to longer hours and increased burnout.

      None of this suggests that AI fails to save time; all research consistently shows it does, often significantly. The issue lies one step higher in organizational decisions regarding the use of saved time. A tool that provides an employee with an extra hour is only as effective as the organization’s ability to use that hour purposefully. Current evidence indicates that most companies are allowing that time to slip away due to rework, quality issues, and a tendency to simply add more tasks.

      Despite the significant investment in enterprise AI, studies point to a more mundane conclusion: the bottleneck is no longer the technology itself but the surrounding management. Companies that are realizing genuine gains are not necessarily those with the most advanced models but those that determined in advance how to utilize the saved time. The rest are acquiring the hours only to let them go to waste.