For years, the corporate promise of artificial intelligence (AI) was simple: Let machines handle grunt work so humans can focus on high-value creativity.
But new research published in the Harvard Business Review shows that promises are morphing into a grueling reality. Instead of offloading tasks, AI is “intensifying” work, leading to a silent creep in responsibilities that could trigger a wave of burnout.
Meanwhile, a team of researchers from the Berkeley Haas School of Business recently concluded an eight-month field study of a U.S.-based tech company with about 200 employees. Their findings suggest that AI doesn’t just speed up tasks but changes the very nature of the workday. While the company did not mandate AI use, employees flocked to the tools voluntarily.
The result? A self-reinforcing cycle of intensification. Because AI makes starting a project easier, workers took on a broader scope of tasks they previously would have outsourced or deferred. Product managers began writing code; researchers took on engineering roles. So-called job creep didn’t just affect the amateurs; experts like engineers found their workloads doubling as they spent more time vibe coding or fixing the AI-generated outputs of their colleagues.
Perhaps most alarming is how AI erodes the boundaries of the traditional workday. Researchers observed that because prompting an AI feels like chatting rather than formal labor, workers slipped tasks into moments previously reserved for rest.
“Downtime no longer provided the same sense of recovery,” the Harvard study found. Employees reported sending “one last prompt” before lunch or engaging in AI during evenings and weekends. This constant state of “ambient work” led to a workday with fewer pauses, creating a feeling of being perpetually “on,” researchers said. While workers initially felt a “momentum” from having an AI partner, that feeling quickly evolved into cognitive fatigue and a sense of always juggling too many open threads.
On paper, it all looks like a productivity explosion. However, researchers warn that these gains are often unsustainable. A presumed “silent increase” in workload can mask a decline in decision-making quality and an increase in errors. When supervisors mistake unsustainable intensity for genuine efficiency, they set the stage for long-term turnover and burnout.
To combat this, researchers argue companies must move past passive adoption and establish a formal AI practice. This involves setting intentional norms to protect human workers through implementing structured intervals to assess alignment and reconsider assumptions before moving forward at AI speed, protecting focus windows and batching notifications to prevent constant context-switching that drains mental energy, and carving out space for human dialogue to counter the isolating effects of solo AI work.
