AI rarely saves time on a task — it just moves the hours from doing the work to prompting, checking, and fixing it.

This explores how AI changes the mix of activities inside a single task, not just how long the task takes. The short answer from the corpus: AI often doesn't shrink total task time much. It moves that time around. Hours that used to go to drafting, coding or calculating now go to writing prompts, reading outputs and checking whether they're right Does AI really save time, or just change how we spend it?. That's why 'time on task' is a misleading productivity measure. Two people can spend the same 40 minutes on a report while doing very different kinds of thinking.

The same pattern shows up at a larger scale. 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 one Does AI really compress all layers of knowledge work equally?. Deciding and delivering stay the same or even grow, which helps explain why translation and legal work haven't shrunk the way the hype predicted. Labor-market data tells a similar story. When AI touches only a few tasks in a job, workers can move their effort to the tasks AI doesn't handle well, and employment holds up better than you might expect Does concentrated AI exposure enable workers to adapt and reallocate?. The split of time within a task and the split of tasks within a job seem to change in the same way.

The less obvious part is what this does to how the work feels. A set of four experiments with more than 3,500 people found that after working with generative AI, people felt more in control when they went back to working alone. They also felt less motivated and more bored Does AI collaboration drain motivation when workers return to solo tasks?. The explanation: AI had absorbed the engaging parts of the task, and what remained was the dull work. So the change isn't neutral. It can remove the parts of a job people actually enjoyed.

The new 'interacting with AI' time has costs of its own. AI suggestions can break your concentration even when they're correct. You stop to read them, and then you have to rebuild your focus before you can continue Does AI assistance always help reasoning or does it carry hidden costs?. That re-focusing time rarely shows up in productivity numbers. There's also a quieter effect on how people see themselves. Because more of the task is reviewing and accepting AI output, people can start counting the AI's work as their own skill. The corpus calls this the 'LLM Fallacy' How does AI-assisted work reshape how people see their own abilities?.

The takeaway: the useful question isn't 'how much faster?' but 'what am I now spending my time doing, and what does that do to my attention, motivation and sense of my own ability?' The corpus suggests AI moves people from making things toward supervising them. That shift has real effects on learning and engagement, and speed metrics don't capture them.