Even when no one gets laid off, AI can quietly make everyone's workday faster, denser, and more exhausting.

This explores why AI often makes work feel heavier rather than lighter, even when nobody loses their job. The most direct evidence is behavioral. ActivTrak tracked what employees actually did as AI adoption rose sharply. People spent more time in work apps, worked more weekend hours, and their daily focus time fell to a three-year low Does AI adoption actually reduce the work that employees do?. The report concludes that AI speeds work up and packs it more densely; it doesn't shrink it. Time saved on one task gets spent on more tasks.

The corpus offers a structural explanation for where that extra work comes from. Narayanan and Kapoor split knowledge work into three layers: deciding what to do, executing it, and delivering it. They argue AI mainly compresses the middle layer. Deciding and delivering stay the same size or grow Does AI really compress all layers of knowledge work equally?. So when drafting gets cheap, people produce more drafts, and then have more to judge, revise, coordinate and ship. Firm-level data points the same way. When AI touches only a few tasks, workers shift their effort to the tasks AI can't do instead of leaving Does concentrated AI exposure enable workers to adapt and reallocate?. An Alibaba customer-service experiment shows this happening in practice. Agents whose routine chats were handled by AI put the freed attention into the harder chats they still owned, without any change in pay or rules Does supervising AI free up worker attention for other tasks?. In that study the outcome was good, since quality went up. But it shows the general pattern: freed capacity rarely becomes free time. It moves to the next queue.

There is also a quieter cost in what is left of the job. In four experiments with more than 3,500 people, those who had just worked with generative AI felt more in control when they went back to working alone. They also felt less motivated and more bored, because AI had taken over the engaging parts of the task and left the routine remainder Does AI collaboration drain motivation when workers return to solo tasks?. Work can get more intense and less rewarding at the same time. Another study adds to this. AI's productivity gains show up when people apply skills they already have. When people use AI to learn something new, the gains disappear and their learning suffers When does AI actually boost worker productivity?. Faster output today can come at the cost of the skill growth that would make tomorrow's work easier.

Supervising AI is itself new work. Across 22 tested models, even the best one broke compliance rules about once in every eighteen tries under realistic workplace pressure, so none can safely run unsupervised in regulated settings Can large language models follow compliance rules under workplace pressure?. Someone has to check. Making that checking rigorous adds friction that people dislike. Design features that force users to stop and think before accepting AI suggestions did reduce over-reliance, but participants rated them lowest, and the human-AI teams still did worse than the AI alone Do forcing functions reduce AI overreliance without harming team performance?. AI can also create work by escalation. In hiring, applicants use AI to send more applications and slip past filters, and 34% of recruiters now spend half their week filtering out the resulting spam Are job applicants and employers locked in an escalating AI arms race?.

What you might not expect is that whether this intensity feels like empowerment or a squeeze may depend less on the technology than on who captures the time it saves. Brynjolfsson argues that AI built to automate rather than augment shifts bargaining power toward whoever controls the technology Does automating labor concentrate power more than augmenting it?. That would explain why saved time turns into more output instead of shorter weeks. One caveat: the strongest direct evidence of intensification here is a single trace-data report. Most of the rest explains the mechanisms behind it rather than measuring workload over time.