Even correct AI suggestions can derail your focus — the cost of an interruption isn't whether it's right, it's that it breaks your train of thought.

This explores what happens to workers' concentration when AI steps into their work, and whether AI only drains focus or can sometimes give it back. The corpus has a clear answer on one point: being right is not enough. An AI suggestion that is correct can still make someone reason worse, because it pulls them out of deep focus and they have to rebuild their train of thought before they can go on Does AI assistance always help reasoning or does it carry hidden costs?. That means judging an assistant by how accurate each suggestion is misses the real cost, which only shows up when you look at the whole task.

If interruptions cost something, the useful question is when and how much AI should interrupt, not just what it should say. One line of work breaks AI support into three separate dials: type (what kind of help), timing (when it arrives), and scale (how much of it) When and how much should AI interrupt human reasoning?. Most explainable-AI research tunes only the first dial and leaves timing and scale on default settings, even though those are where much of the harm or benefit happens. Ideas from older automation research point the same way. Simkute and colleagues propose steady feedback, keeping the task stable, and clearly dividing who does what as ways to protect workers' attention, though these remain proposals that haven't been tested yet Can Human Factors design fixes restore GenAI productivity losses?. Timing matters outside of thinking tasks too. In customer service, humans who took over early from a struggling AI rescued the conversation, but late handoffs, after customers were already frustrated, did not When human workers take over from AI, does it actually fix the problem?.

The less obvious finding is that AI changes what workers focus on more than it cuts their total effort. Time saved on doing the task moves into writing prompts and checking what the AI produced Does AI really save time, or just change how we spend it?. That evaluating work tends to drop out of sight: finished output gets treated as proof the work happened, while the effort and attention behind it go unrecorded Which workplace cues survive AI mediation and which disappear?. Over longer periods, EEG evidence suggests that leaning on an LLM is linked to weaker brain engagement and to people having trouble remembering their own recent work Does AI assistance weaken our brain's ability to think independently?. Learning new skills seems especially fragile. Productivity gains disappear when workers use AI to learn something new instead of applying skills they already have When does AI actually boost worker productivity?.

The surprising counterpoint is that AI can also free up focus. In an Alibaba experiment, workers who supervised an AI agent handling some chats did better on the separate chats they still handled entirely on their own, with no change in pay or rules. Their freed-up attention flowed to the work the AI wasn't touching Does supervising AI free up worker attention for other tasks?. A similar pattern shows up across whole firms: when AI touches only a few of a job's tasks, workers can shift toward the rest Does concentrated AI exposure enable workers to adapt and reallocate?. Taken together, the evidence suggests the difference lies in how AI enters the work. AI that breaks into a task you're in the middle of costs focus, while AI that takes whole chunks of work off your plate can return it.

One gap: the corpus has little direct research on everyday workplace interruptions like notifications or always-on copilots breaking up the workday. Most of the evidence comes from reasoning tasks and customer-service settings, so treat the broader claim as a well-supported pattern rather than a settled finding.