Start from a dump, not a plan

Open the day by writing everything you think you should do, in any order, without judging it. Deadlines, errands, the message you have been avoiding, the thing you promised someone on Friday. The list will be longer than the day, which is the useful part: you cannot make an honest plan until you can see the whole demand.

Tell the assistant your real constraints

This is the step most people skip, and it is the step that makes the difference. A useful prompt includes the hours you actually have, the fixed commitments inside them, your energy pattern, and what must not slip.

"I have from 9:00 to 17:30 with a meeting from 11:00 to 12:00 and school pickup at 15:30. I focus best before lunch. The client draft is due tomorrow morning. Here is my list — build a realistic plan and tell me what will not fit."

The instruction to say what will not fit is important. Without it, an assistant will politely arrange everything into the day, which produces a plan that fails by 2pm.

Ask for a shape, then correct it

Treat the first output as a proposal. You know things it does not: that a particular task always takes twice as long, that you cannot do creative work straight after a meeting. Push back explicitly — "move the writing to the morning", "this one takes two hours, not thirty minutes" — and the second version is usually close to right.

Protect two blocks, plan the rest loosely

Detailed hour-by-hour schedules break at the first interruption. A more durable structure is two protected focus blocks for the work that actually matters, and a flexible pool for everything else. If the day goes sideways, you still finished the two things that count.

Name the one thing

Before you start, decide the single outcome that would make the day a success. Write it somewhere visible. When priorities compete later — and they will — this is the tiebreaker that saves you a decision when you have least capacity to make one.

Close the day in five minutes

List what actually got done, including the unplanned things.
Note anything that took much longer than expected, and why in one line.
Move unfinished items forward deliberately — if something has moved three days in a row, either schedule it properly or delete it.
Write tomorrow's one thing before you close the laptop.

The evening review is what makes the morning plan get better over time. Without it you repeat the same estimation errors indefinitely.

What to avoid

Letting the assistant assign priorities it has no basis for — you know what matters, it does not.
Planning in fifteen-minute increments; the overhead exceeds the benefit.
Rebuilding the plan from scratch every morning instead of starting from last night's review.
Treating an unfinished plan as failure. A plan is a forecast, and forecasts are supposed to be revised.