What’s the ROI of AI?
That’s the question dogging finance and HR teams around the world. Even as tech giants tout the time-saving potential of large language models and companies continue pouring more money into the nascent technology, its purpose in the finance function is still to be determined.
Now, new data indicates that in some cases, the technology is forcing workers to spend hours fixing faulty outputs.
That’s one of the key findings from a new report released last week by enterprise cloud software provider Workday.
Though 85% of the workers surveyed said they’re saving between one and seven hours a week using AI tools, they’re also losing some of that time correcting or validating the technology’s outputs. By Workday’s calculations, 37% of the time employees saved using AI tools was lost to “rework,” which the company defined as “correcting, clarifying or rewriting low-quality AI-generated content.”
“For every 10 hours of efficiency gained through AI, nearly 4 hours are lost to fixing its output,” Workday officials wrote in the report. The company described this as the “AI tax on productivity.”
Workday surveyed a total of 3,200 respondents, with half comprising leaders and the other half employees. The company partnered with Hanover Research to conduct the online survey, which was administered in November.
The report, much like a similar study issued by Harvard Business Review in September and another by MIT in July, calls into question the long-heralded proposition that artificial intelligence will free people up from mundane tasks and enable them to focus on higher-level, strategic endeavors.
As Workday’s report puts it: “Instead of reallocating time toward judgment, creativity and decision-making, employees spend it correcting low-quality output. At scale, this pattern compounds, translating into millions of lost hours each year in large organizations.”
If using AI tools requires some unanticipated additional labor, it probably doesn’t come as a surprise that the most frequent users “feel the most strain,” as Workday put it in the report. Among workers who use AI every day, 77% say they review AI-generated work “just as carefully as work done by humans, if not more,” according to the report.
For business leaders, Workday’s report suggests “rethinking how AI productivity is measured” by “accounting for both time saved and time lost to rework.” Contrary to the steady drumbeat to adopt AI at all costs, Workday’s report calls instead for shifting focus away from simply “accelerating usage and more on improving how AI-supported work is measured, supported and designed.”
The report also recommended that leaders update job descriptions to “clarify where AI is expected to assist, where human judgment is essential and how success is measured.” Almost nine in 10 of the organizations surveyed said that fewer than half of their roles were updated to include AI-related skills, according to Workday.
“AI has been layered onto roles that were never updated to accommodate it — forcing employees to use 2025 tools within 2015 job structures,” the report stated. “Rather than reducing effort, this mismatch often increases it, as employees are left to reconcile faster production with unchanged expectations around accuracy, judgment and accountability.”
And while most leadership respondents (80%) agreed with the notion that businesses should reinvest productivity gains into workforce development, their actions to date don’t fully reflect that belief. Per Workday, organizations are allocating about 39% of cost savings to “technology and infrastructure” compared to 30% to workforce development
“Time savings follow a similar pattern: leaders prioritize increased work volume and digital infrastructure over employee development,” Workday officials said.
