Every company seems determined to integrate AI, but acquisitions may not have a long shelf life. After an “initial increase in productivity,” employees who use AI report worse workdays and less work-life balance, and they produce lower-quality work overall, according to an ongoing study first. PUBLISHED this week in the Harvard Business Review.

Researchers from the University of California, Berkeley, studied the habits and behaviors of about 200 people who use generative AI in their work at a technology company for eight months. The company offers enterprise-level subscriptions to employees of its AI products. Employees are not required to use AI, but many workers do. What happened next was exactly what AI companies hoped would happen: Employees who used AI worked faster and took on more responsibilities. But there are unintended consequences that show the limitations of current AI tools used in the workplace.

One of the biggest selling points of AI in the workplace is that it can help employees manage tasks that may not be within their skills or expertise. Non-developers, for example, can now vibe code almost any project. Employees in the study did this, taking on work that could have otherwise been delegated or avoided, the authors said. So employees are inadvertently creating more work for themselves, putting more on their plates and struggling to balance it all.

Read More: Is AI putting Jobs at risk? A New Survey Finds Important Differences

We also know that AI as a job hack is not without its failures. AI outputs are rarely ready to go without first being reviewed by a real human. A September 2025 study found that employees spend hours each week dealing with their colleagues and their own low-quality or flawed AI work, sometimes called “worklop.” A 2025 business report from OpenAI said that employees only save an average of 40 to 60 minutes per week, with even more hours saved for power AI users.

That time saved by AI probably hasn’t made a measurable difference to work-life balance. Employees in the UC Berkeley study actually ended up working longer hours. The always-on and easy-to-use nature of AI makes it simple for them to run a query during their lunch break or ask a quick question after logging off.

Although employees have the feeling of having a digital partner, their cognitive loads do not necessarily decrease, and there are expectations to deliver results quickly because they use AI to help. This is why researchers at UC Berkeley say that AI is more likely to “intensify” work than reduce it.

Authors Aruna Ranganathan and Xingqi Maggie Ye offer culture-centered solutions and rules that companies can adopt to avoid AI-driven burnout. This includes protecting time for human connection, prioritizing quality results over speed, and ensuring employees have blocked time to focus without AI distraction. Intentional use of AI — both on and off the job — is one of the best ways to avoid misuse and make work less sloppy.

Across industries, workers worry that advances in AI will wipe out their jobs. Anthropic CEO Dario Amodei just said AI could result in “unique disease” that will soon disrupt the workforce. And the latest fights Amazon’s thousands of layoffs are clearly done because the company expects AI to fill the gaps and help the remaining employees do more with fewer resources. But we’ve seen a lot of evidence that while AI can help you do some tasks, it can’t make it happen. whole paper in most industries.