AI is making workers more productive, but it could also be burning them out, according to a new study by researchers at the University of California at Berkeley.

The revolution and the skyrocketing productivity AI promised is already taking hold in corporate America, including at an unnamed 200-person U.S. tech firm studied by the Berkeley researchers, an article about the in-progress research published in the Harvard Business Review shows.

Over the course of eight months and with the help of 40 “in-depth” interviews across engineering, product, design, research, and operations, the researchers found employees using AI tools increased both the work they could complete as well as the variety of tasks they could tackle—even when they weren’t forced to adopt the technology.

Yet, as employees’ productivity increased, so did the amount of work they took on, in part because AI made it easy to begin tasks. Soon, some workers were using up what previously had been natural breaks during the day to prompt AI, eventually filling most of their time at the office with tasks.

This type of implicit pressure paired with a lack of time to recharge could lead workers to be less productive, Rebecca Silverstein, a licensed clinical social worker and the program director at Brooklyn-based Elevate Point, told Fortune.

When workers take up every part of their day with tasks and sacrifice their breaks, they give up the interpersonal relationships that are just as important to a person’s work life as their actual work.

People also need these breaks, either during the day or after work, to recharge and have the capacity to work effectively, she added.

“Just focusing on that productivity mindset, in the long term, is super harmful for someone,” Silverstein said.

And as one worker who was interviewed by the Berkeley researchers put it: “You had thought that maybe, ‘Oh, because you could be more productive with AI, then you save some time, you can work less.’ But then really, you don’t work less. You just work the same amount or even more.”

The researchers warned that while the idea of workers taking on more tasks voluntarily could seem ideal, nonstop work has the potential to lead to problems down the line, including blurring the boundary between work and nonwork, as well as burnout and cognitive fatigue.

Worse yet, employees’ focus on supercharging their productivity could potentially lead to lower-quality work, the researchers found.

In AI, workers described having a “partner” that helped them take on a larger variety of tasks, and yet, doing so led to more multitasking and task-switching, which has been shown in previous studies to decrease productivity.

When workers found that each of them was doing more work with the help of technology, this created implicit pressure that weighed on them mentally, the researchers found.

How to overcome AI overload

To battle the trend of AI overload, the UC Berkeley researchers recommended organizations take the time to be intentional. They suggested incorporating pauses into work to better evaluate decisions or reconsider assumptions, as well as organizing work so as to protect employees’ windows of focus without interruption. Companies should also prioritize human connection and social exchange, the researchers said.

Josh Cardoz, who advises organizations on enabling people in the AI era in his work as chief creative and learning officer at Sponge, told Fortune organizations also need to make sure that by encouraging AI use either explicitly or implicitly, they are not sacrificing work quality.

These changes have to come from the top, he said.

Company leaders need to define explicitly what AI fluency means for employees depending on their role. When they make decisions about AI strategy, they should encourage employees’ input. Those workers who are already making the most out of AI should also be uplifted by the company, he said.

Most importantly, when it comes to this rapid change in the workplace, Cardoz said companies need to get back to basics by encouraging employees to adopt the new technology, but also assuring them to help decrease the fear and anxiety that accompanies the unknown.

“You need to remember that there’s a human factor in all of this,” he said.