Put AI agent work on your calendar as three separate blocks, not one: a short brief block when your head is clear, an unattended run window while you do other things, and a review block before the deadline. Your AI tasks don't slip because of the prompt. They slip because they never got a real slot in the day, so they compete with everything else and lose.
If you run a calendar-first day—where a task isn't real until it has a time—then AI delegation has to play by the same rule. Here's how to place agent work so it actually lands.
Why floating AI tasks never finish
Here's the trap. You tell an agent to draft something, then you sit and wait, or you forget it entirely. Either way you've broken the one thing calendars are good at: protecting attention. Waiting on an agent is dead time. Forgetting means the output rots unreviewed and you redo it by hand at 11pm.
Delegation only pays off when both the handoff and the check-in are scheduled. Treat an agent like a capable contractor. You wouldn't hire one and then hover in the doorway all afternoon. You'd brief them, leave, and come back to inspect the work.
The three blocks, and where they go
1. The brief block (10–15 minutes, high-energy)
This is the only part that needs your full brain, so give it your best hour. Briefing is where you set the goal, the constraints, the format, and what "done" looks like. A weak brief gives you a weak result no matter how long the agent runs.
Put this block early, right after the first thing you do that's already sharp. Don't brief an agent at 4pm when you can barely form a sentence. A good brief names the reader, the output shape, and an escape hatch for uncertainty—something like: "Audience is skeptical finance leads. Output a 5-bullet summary, no adjectives, flag anything you're unsure about."
2. The run window (unattended, low-value time)
Once briefed, the agent works while you don't watch. Schedule this over time you'd never spend on deep work anyway—your commute, a standing meeting where you're mostly listening, lunch, a batch of shallow admin. The whole point is that the run window overlaps with something else. If you're staring at a spinner, you've placed it wrong.
Give it more runway than you think. If review is at 3pm, brief by 1pm, not 2:45.
3. The review block (15–30 minutes, before the deadline)
This is the block people skip, and it's why delegation feels untrustworthy. Schedule review with a real buffer—enough time to send it back once if it's wrong. Review isn't "read and approve." It's checking against the "done" definition from your brief: Did it hit the format? Did it flag its uncertainties? Is anything confidently wrong?
Leave only five minutes before the thing is due and you'll rubber-stamp bad work, because you've got no other option.
How to lay it out on the actual calendar
- Color-code agent blocks differently from your own deep work. You want to see at a glance how much of the day is you-work versus agent-work.
- Never stack brief and review back-to-back. The gap between them is the run window doing its job. If they're adjacent, you're just doing the task yourself with extra steps.
- Batch your briefs. Three agent tasks? Brief all three in one morning block, let them run in parallel, review in one afternoon block.
- Protect the review block like a meeting. It has a deadline attached. Treat a no-show to your own review the way you'd treat missing a client call.
A worked example
Say you owe a partner a project update by end of day.
- 9:15–9:30 – Brief block. Hand the agent the raw notes, the audience, the format, the "done" bar.
- 9:30–1:00 – Run window, overlapping your morning meetings and email. You don't touch it.
- 1:00–1:20 – Review block. Check against the bar. Send back once if it needs it.
- By 2:00 – Sent, with hours of buffer left if something surprises you.
The task took maybe 35 minutes of your attention, spread across a day where you also did everything else.
The hard part is the brief
Placement is simple once you accept the three-block rule. The part that takes practice is writing a brief good enough that the run window produces something reviewable—instead of mush you have to redo. That's a skill, and it's the same handful of inputs every time: role, goal, constraints, format, uncertainty flags. Building those from scratch takes longer than most people expect.
If you'd rather start from a proven structure, the Delegate Like a Boss AI Agent Instruction Pack gives you reusable briefing frameworks designed to make your brief block faster and your review block lighter. It's the shortcut for the one part of this that's genuinely hard.
FAQ
How long should the run window be?
Longer than you expect, and always with buffer before review. A good rule: if review is at 3pm, brief before 1pm. That extra runway means one round of corrections doesn't blow your deadline.
What if I don't have low-value time to overlap the run window with?
You almost certainly do—meetings where you mostly listen, commutes, lunch, admin batches. The run window's job is to sit on top of time you weren't going to spend on deep work anyway.
Can I skip the review block if the agent is usually reliable?
No. That's exactly when a confident wrong answer slips through. Keep review short if you trust the setup, but keep it on the calendar with real buffer.
Should I brief multiple agent tasks at once?
Yes. Batch briefs into one high-energy block, let tasks run in parallel, and batch reviews into one afternoon block. It keeps context-switching down and protects your deep work.