A study published in Harvard Business Review found that generative AI made work more intense at the company it observed. Instead of freeing time for higher-value work, the tools led to bigger workloads, cognitive fatigue and burnout.
The productivity gains from AI are real. This study shows the costs that come with them, which most organisations do not plan for.
How the study was done
Aruna Ranganathan and Xingqi Maggie Ye of UC Berkeley's Haas School of Business spent eight months inside a US technology company with about 200 employees. They observed staff in person twice a week, tracked internal communications and conducted over 40 in-depth interviews across engineering, product, design, research and operations.
They expected to see people doing the same work in less time. What they recorded were three kinds of intensification that built on each other over the study period.
Three ways the work grew
Task expansion. When AI made some tasks faster, employees did not use the saved time to rest or think. They took on work outside their roles. Product managers started writing code, and researchers took on engineering tasks. One engineer summed it up: "You had thought that maybe you could work less. But then really, you don't work less." Expectations of each role grew to fill the time saved, and often more.
Blurred boundaries. People began working during breaks, over lunch and in meetings. The researchers described work as becoming "ambient, something that could always be advanced a little further." Because AI makes small tasks easy, any spare moment can turn into work. The study found that downtime no longer gave people the recovery it used to. Many UAE workplaces already expect long, intense hours, so this pattern needs watching here. A tool that makes "one more quick task" easy during personal time wears people down over months.
Constant juggling. Workers ran several threads at once, which the researchers described as "cognitive load and a sense of always juggling." In practice this meant continual switching of attention, frequent checking of AI outputs and a growing pile of open tasks. Fragmented attention of this kind leads to errors and exhaustion.
What it cost the company
The researchers documented cognitive fatigue, burnout, and workload creep that happened without any explicit request from management. They also recorded weaker decision-making, lower-quality output despite early productivity gains, and higher staff turnover.
Turnover changes the return on an AI investment. Short-term productivity numbers can hide the longer-term loss of experienced people and what they know.
Habits that helped
Some employees developed what the researchers call "AI practice": deliberate habits for managing the pace that AI tools create. Management did not mandate these habits. Workers who saw the pace was unsustainable built them for themselves.
- Intentional pauses. Set moments to step back, assess and reflect, instead of always pushing the next task forward.
- Sequencing. Batching notifications and protecting focus time, so AI outputs do not become a stream of interruptions.
- Human grounding. Keeping regular conversation and social contact in the work, so human judgement and relationships stay central.
The study does not argue against AI. Many employees welcomed the tools because they "enabled us to accomplish more, made our work feel more familiar, and made our work more rewarding." The trouble was that nobody managed the side effects on workload.
The UAE's AI strategy is ambitious, and adoption is growing fast across government and the private sector. Any team rolling out AI tools should track work intensity alongside productivity from the first month: hours logged, after-hours activity, how often people switch between tasks, and wellbeing survey scores. If those numbers climb while output climbs, the gains are being borrowed from your staff, and the habits above are a reasonable place to start correcting it.
Sources: