Employees spend a lot of time checking AI output and correcting mistakes. Photo: Shutterstock
Generative AI (genAI) tools are enabling employees to work faster, take on more tasks, and extend their workdays, but new research suggests this change is unsustainable, increasing employees’ overall workloads and putting them at risk of burnout.
Researchers at the Hass School of Business at the University of California, Berkeley, spent eight months observing the work habits of a U.S. technology company with about 200 employees.
They regularly observed and interviewed engineering, product, design, research, and operations staff who had voluntarily started using genAI in their work.
Employees found the tools empowering them. AI has given them the confidence to take on tasks outside of their normal roles, creating what researchers call an “enhancement” of their jobs. For example, product managers started writing code, and researchers took on engineering tasks, often guided by AI feedback rather than their peers.
This made employees feel empowered when taking on “newly accessible” projects with the support and feedback of an AI rather than a colleague. For example, a product manager started writing code and a researcher took on engineering tasks.
Many employees are now using AI to get more done at the same time. For example, what used to be your break time at work is now filled with a series of AI prompts to perform various small tasks. Meanwhile, some employees have put in extra time, such as an engineer helping a coworker learn to vibe code.
Author Xinqi Maggie Ye and Aruna Ranganathan, an associate professor at the Hass School of Business at the University of California, Berkeley, write that new small projects and additional work responsibilities that were “just experimenting” with genAI “added up to meaningfully expand the scope of my work.”
“These actions gave little sense of additional work, but over time they created a working day with fewer natural breaks and more continuous engagement,” they found, citing “a new rhythm in which workers manage multiple active threads simultaneously.”
AI efficiency paradox revealed
While GenAI is often touted as a way to increase productivity and make workers’ lives easier by automating menial tasks like coding and freeing up time for higher-level thinking, results from observational studies suggest that the actual results are very different.
“As the excitement of the experiment wears off, workers notice their workload quietly increasing and suddenly feel overwhelmed with everything they have to do,” they found.
The results support similar studies around the world that have found that, despite the rhetoric, the outcomes of AI deployment remain widely variable and, as the CSIRO puts it, the evidence for its benefits is “equivocal.”
Companies such as Procter & Gamble and Boston Consulting Group have reported increased productivity after implementing AI, but similar to PwC’s analysis, one survey of 300 Australian employees found that 30 per cent did not see an increase in productivity.
A recent Workday survey of 3,200 employees found that even though AI saves them 1 to 7 hours a week, employees still spend nearly 40% of their time checking that output and fixing mistakes.
Only 14 percent of these workers say they consistently see clear and positive end results from AI, and 77 percent admit they spend as much or more time reviewing AI-generated work than humans do.
One observer calls this a “crisis of rework that no one can measure,” and developers in particular report that coding with AI tools feels 20% faster, but actually takes 19% longer. This is because the AI-generated “workslop” is not ready for production.
Management doesn’t know about employees’ actual AI experience
Misconceptions about the benefits of genAI are distracting executives, with the technology still proving to be a blind spot, with UNSW Business School’s Frederic Anthiel warning companies to stop focusing on “easily measurable outcomes”.
“Employees learn to navigate these systems, focusing more on looking busy than on creating real value,” he writes.
No matter how you look at it, management supports this: A recent report from the AI Consulting Firms section AI Proficiency ReportFor example, it surveyed 5,000 knowledge workers in the US, UK, and Canada and warned that “executives are in the dark” about how genAI is helping them and their employees.
While approximately 69% of workers are “AI experimenters” who use genAI for basic tasks, 28% are still “AI novices,” 97% of workers say they are not using AI well or at all, 25% say AI does not save them time, and 40% are satisfied they will never use AI again.
“Employees know how to use LLM, but if they can’t think of a use case for it, they quickly stop using it,” the report says, with less than a third of respondents saying they save more than four hours a week by using AI.
“While executives said their company has a clear AI strategy, adoption is widespread, and employees are encouraged to experiment and build their own solutions, the rest of the workforce disagrees,” said Section COO Taylor Malmsheimer.
