Stop 10 Minutes of AI From Spending Your Focus Block
Treat AI help like a recurring meeting, price it in minutes, and keep a tool-free stretch for the work that needs persistence.
Count the calendar before you defend it
Your calendar is full, and the work that needs persistence keeps losing to small asks that never look like meetings. Price the asks, including a 10-minute AI slot that can spend a focus block.
List recurring invites. The result is a short inventory of standing asks that survive the week without a review. Price each in hours. You are done when the total shows which meetings actually take the block.
Count the hours, not the titles. A standing meeting that keeps returning is a recurring cost, and a cost that nobody reviews stays hidden. If a meeting keeps returning, count the hours it would take to cancel, reschedule, or replace it with a note. The calendar should show the cost, not just the invite.
Cancel gracefully when the inventory is clear. Send a short note, offer a deadline, and move the ask to a written update. If the invite is a standing habit, say the habit is being reviewed, not the person. A recurring meeting that nobody reviews takes time from the work. A full calendar hides the cost of the work. Make the cost explicit before you ask for more effort.
Treat AI help as a scheduled cost
Answer-on-demand help can quietly impair the independent work that focus blocks are supposed to protect. The research is blunt. Researchers ran three randomized controlled trials that together recruited 1,222 online participants: 354 and 667 in fraction trials, and 201 in a SAT-style reading trial.
In the first fraction trial, the AI-assisted group solved 57% of the unassisted test problems, while the control group solved 73%. In the second fraction trial, the AI-assisted group solved 71% of the test problems, while the control group solved 77%; the effect size was 0.19. In the third trial, which used SAT reading questions, the AI-assisted group solved 76% of the test questions, while the control group solved 89%.
The assistant was removed once participants got to problem 12. In the second trial, 61% of the AI-assisted participants said they mainly asked the assistant for answers. The observed effects in the trials appeared after approximately 10 minutes of AI interaction.
Add AI help as a scheduled 10-minute line item, and let the visible slot be the only place you ask for answers. The result is a bounded slot, not an open tab that follows you into the work. A small, convenient interruption can take the block before the hard work starts. Answer-on-demand help is convenient, and convenience is the problem when the work needs persistence. The slot makes the help visible, so the block can stay protected.
Protect the stretch where work actually happens
Mark a tool-free persistence stretch. You are done when the window is meeting-free, the assistant is off, and the work stays independent.
The mistake is scheduling AI help inside the persistence stretch. You have not protected the block; you have moved the interruption. If you manage a team, the same rule applies to focus blocks for team members: protect the stretch, not just your own. A team deep work schedule works when the calendar shows the protected stretch.
Put the stretch on the calendar as a meeting-free block, not as a hope. If the team needs a shared window, make it visible to everyone and keep it clear of standing asks. Uninterrupted focus time is the output.
The focus-block cost line is: list recurring invites, price each in hours, add AI help as a scheduled 10-minute line item, and mark one tool-free persistence stretch. Put that line on the calendar. If the stretch is protected, the work gets a fair chance. A protected stretch is easier to defend than a vague intention. Keep the AI slot visible, keep the stretch tool-free, and let the calendar show what the work actually costs.