Teach a teammate to defend their calendar by walking them through one real week of their own schedule, not a demo. Sit together, find the meetings that could shrink or vanish, and coach them to write prompts that reflect their priorities. Explaining it out loud is also the fastest way to sharpen your own system.
I've handed people polished prompt libraries before and watched them go unused. What actually works is teaching the reasoning behind the prompts, so they can adapt when their week changes. Here's the sequence I use.
Step 1: Start with their real calendar, not a hypothetical
Open their actual next week. Made-up examples don't stick, because nobody feels defensive about an imaginary 2pm sync. Their real Thursday, with the three overlapping meetings and the block that got eaten by "quick chats" — that's where the learning lands.
Ask two questions and write the answers down:
- Which two hours this week matter most for real work?
- Which meetings could be shorter, async, or skipped entirely?
Don't touch AI yet. You're building the judgment the prompts will serve. If they can't name their most important hours, no prompt will save them.
Step 2: Show the thinking before the prompt
Most people skip this part. Before you type anything into an AI tool, say out loud what you want it to do and why.
For example: "I want to protect Tuesday and Wednesday mornings, so I'll ask the AI to scan next week, flag anything landing in those windows, and suggest where each conflict could move." That's a decision, stated plainly. The prompt is just that decision written for a machine.
Let them narrate their own version
Now have them do the same for their week, out loud, before writing anything. When they say "I want it to draft a polite decline for meetings without an agenda," they've already done the hard thinking. The prompt becomes obvious. This narrate-first habit is the whole skill. Teach that and the rest follows.
Step 3: Write one prompt together, then critique the output
Write a single prompt as a pair. Keep it small — one job, one week. Run it. Then read the result together and ask: did it protect the right hours? Is the tone something they'd actually send?
The critique matters more than the draft. When the AI suggests moving a meeting they'd never touch, that's the teaching moment: the prompt needed their constraint stated up front. Show them how adding one specific line changes everything. That's them learning to steer, not just copy.
Step 4: Give them a starting set, then make them modify it
People learn faster editing a working example than staring at a blank page. Give them a few solid prompts to begin with — for scanning conflicts, drafting declines, blocking focus time — but require them to change at least one thing in each so it fits their voice and role.
If you'd rather hand over a road-tested set than build from scratch, the Deep Work Defender AI prompt pack for focus and calendar control is what I give people who want to skip the trial-and-error. It's organized around the same decisions above, so it doubles as a teaching outline: you talk through why each prompt exists, they adapt it, and they're running by the end of the session.
Step 5: Set a one-week check-in
Learning fades without a follow-up. Book fifteen minutes a week later and ask three things:
- Which prompt did you actually use?
- Where did it get in the way?
- What did you stop doing that you used to do by hand?
That third question is the real measure. If nothing changed in their behavior, the prompts were decoration. Usually one prompt sticks and the rest get quietly dropped, and that's fine. One reliable habit beats five you have to remember.
Why teaching this makes you better at it
Here's the quiet payoff. You can't teach calendar defense without stating your own rules out loud, and most of us have never done that. Explaining why you protect Tuesday mornings forces you to decide whether you actually do — or whether you've been letting it slide.
Every time I coach someone through this, I catch one lazy prompt of my own I've been tolerating. Teaching is just auditing with a witness. Do it twice a year with someone new and your own system stays honest.
FAQ
How long does it take to teach someone?
One focused hour to set it up, plus a fifteen-minute check-in a week later. Longer than that and you're doing it for them instead of teaching them. Keep the first session to one calendar and two or three prompts.
What if my teammate uses a different AI tool than I do?
The method transfers. The thinking — name your important hours, state your constraint, critique the output — works in any chat-based AI. Prompts may need small wording tweaks, but the decisions behind them don't change.
Should I just send them the prompts instead of meeting?
You can, but adoption drops sharply. People use prompts they helped shape. A shared prompt set is a great starting point, but pair it with even a short live walkthrough so they understand why each one exists and how to change it.
What's the most common mistake when teaching this?
Handing over too many prompts at once. Overwhelmed, people revert to their old calendar. Start with the single prompt that solves their most painful recurring problem, get one win, then add more only if they ask.