AI time savings lead to more work

A study of 25,000 Danish employees found that 85% of those who used generative AI redirected the time saved to other professional tasks, rather than reducing their overall workload.

Workload increases without clear benefits

Researchers from the University of California at Berkeley found that employees at a mid-sized American tech company took on more tasks, expanded their responsibilities, and worked longer hours without being formally asked to do so.

Mechanisms of increased work

The study identified three main mechanisms: employees were given more missions, breaks were eliminated, and multitasking became the norm. While efficiency initially improved, long-term productivity gains were limited, and workers experienced fatigue and more errors.

AI does not reduce working hours

The Danish study also found that time saved through AI use was minimal and reinvested in other tasks, with no significant changes in salaries or working hours. Microsoft’s experiment with Copilot showed a reduction in time spent on emails but no change in the number or type of tasks completed.

Economic impact remains unclear

Productivity gains from AI are not clearly visible in traditional economic indicators. A report by economist Torsten Slok found that seven major American tech companies, which sell AI infrastructure and services, showed increased revenue and margins compared to other S&P 500 firms, which did not see similar changes.

New tasks created by AI

The Danish study found that a third of new tasks created by AI involved overseeing the technology, ensuring legal compliance, and reviewing AI-generated reports. A quarter of new tasks related to integrating AI into workflows, and even employees who did not use AI reported new workloads linked to it.

Challenges in AI adoption

Researchers warn that the real challenge for companies is to determine where AI adds real value, taking into account the time spent on oversight, errors, and working conditions. They recommend protecting time for focus, grouping non-urgent requests, and creating space for team reflection to prevent saved time from simply becoming new tasks.