You set up a personal agent, ask it something vague, get something vague back, and decide it’s overhyped.
I’ve watched smart people do this with every AI tool for years. The problem isn’t the tool. It’s that nobody trained it.
An agent is like a new assistant who’s brilliant, fast and knows nothing about you. Generic in, generic out. The fix is a week of real, small jobs, each one briefed clearly and reviewed honestly.
This Muse tutorial is that week. It assumes you’ve already done the basic setup in how to use Muse: least access, confirmations on, payment through one-time cards. As of October 2026, Muse is rolling out in the U.S., so if it isn’t in your account yet, keep this plan for when it is.
The brief template (use it every day)
Every task this week starts with the same five-part brief. Copy it into a note on your phone.
Goal: [WHAT YOU WANT DONE, IN ONE SENTENCE] Context: [WHO, WHERE, WHEN, AND ANY PREFERENCES IT SHOULD KNOW] Limits: [WHAT IT MUST NOT DO — E.G. “DON’T SEND ANYTHING,” “NOTHING BEFORE 9 A.M.,” “KEEP IT MODEST”] Done looks like: [THE RESULT YOU’LL CHECK — E.G. “THREE OPTIONS IN A LIST,” “A CONFIRMED BOOKING”] Check with me when: [THE MOMENTS IT MUST STOP AND ASK — E.G. “BEFORE ANY PURCHASE,” “IF THE TIME ISN’T AVAILABLE”]
Short is fine. Specific is mandatory. “Book dinner” is a wish. “Book dinner for two near home Friday evening, quiet, nothing fancy, check with me before confirming” is a brief.
Day 1: An errand you’ve been dodging
Pick a routine appointment or reservation. A haircut, a car service, a table for Friday.
Brief it with the template. Watch how it handles the confirmation. Review the audit trail afterward: what steps did it take, and did any surprise you?
The point of day one isn’t the haircut. It’s learning how Muse asks, how it reports back, and how much detail it needs from you.
Day 2: Inbox triage, read-only
Connect your personal email at read-only, if you haven’t already. Ask Muse to sort the last week of mail into three buckets: needs a reply from me, has a deadline, safe to ignore.
Then check its work against your own read of the inbox. Where did it miss? Tell it, in plain words: “Anything from the school goes in ‘needs a reply,’ even if it looks like a newsletter.” Muse remembers what matters to you, so this is where your corrections start compounding.
Don’t turn on send access today. Read-only is the lesson.
Day 3: A travel plan
Give it a real trip, even a short one. Dates, who’s going, rough limits in words, must-haves (“aisle seat,” “a hotel with a pool for the kids”).
Ask for options first, not a booking. Compare them to what you’d have found. Then, if you like one, let it book with your confirmation, using the one-time card.
What to watch: whether it respected your limits, and whether it filled in any detail you didn’t give it. Plausible guesses are the most common agent mistake.
Day 4: A bill to lower
Meta’s own example of what Muse can do is negotiating on your behalf, like lowering a bill. Pick one recurring bill that has crept up.
Brief it tightly. Which account. What you want. What it may and may not agree to. And the most important line: “Check with me before agreeing to any change.”
Whatever happens, read the audit trail. You’ll learn more about how the agent works from one negotiation than from a week of chatting.
Day 5: A goal plan
Muse can build plans for bigger goals and keep advancing them. Pick one personal goal: getting back in shape, finishing a book a month, more one-on-one time with each kid.
Brief it with the goal and your real constraints (“mornings before the school run only,” “two evenings a week are family nights”). Ask it to propose a plan and the next three actions, then check in with you on a schedule you choose.
This is where a personal agent stops being an errand runner and starts being a follow-through partner. The plan is only as good as the constraints you give it.
Day 6: Corrections and forgetting
No new task today. Clean up.
- Scroll back through the week’s audit trail.
- List every correction you made. Restate the important ones to Muse as standing preferences.
- Tell it to forget anything it picked up that it doesn’t need: a one-off health detail, a surprise you’re planning, an account number you pasted by accident.
- Review connected apps and access levels. Revoke anything the week didn’t use.
This is the habit that keeps a personal agent useful and safe for the long haul.
Day 7: Weekly review
Set a recurring weekly check-in, the same way I run a Momentum Meeting in my companies, just smaller and personal.
Ask Muse for a short recap: what it handled, what’s waiting on you, what’s coming up next week. Then decide three things:
- One job to keep handing it because it went well.
- One permission to widen (say, send access for a single email account), or none if you’re not ready.
- One thing to take back because it’s better done by you.
Scoring your own agent weekly is how you earn trust in it based on evidence, not hope.
What to keep for yourself
Here’s my “why,” briefly. I build AI into my work and my home so I can be more present with my wife and kids, not so I can outsource being a husband and a father.
So I’m careful about the line. Let Muse handle the logistics around what matters: book the date night, plan the trip, track the practice schedule. Keep the relationships, the conversations, your faith and the family decisions for yourself. An agent can clear your calendar. It can’t show up for you.
After the first week
If the week went well, keep the brief template and the weekly review going. Add one job at a time.
If you want to understand the bigger picture of how agents plan, use tools and hand back to people, read my plain-English AI agents guide. Before you widen any permissions, read the Muse privacy page so you know what’s protecting you and what isn’t. And if you run a company, see how I separate the two lives on personal AI agent for entrepreneurs.
Your next step
Copy the brief template into your phone right now. Tomorrow morning, run Day 1 with one errand and one connected app.
I share my own agent experiments, including the ones that go sideways, on YouTube. If you’d rather start with the basics of directing AI, my guide on how to write AI prompts is the foundation. And if you want to know more about the person behind these picks, here’s my story.
Questions people ask
How long does it take to train a personal AI agent like Muse?
Give it a focused week of one real task a day, then keep correcting it as you go. Muse remembers what matters to you and factors in what you tell it, so the briefs and corrections you give early shape how useful it becomes.
What should I never delegate to Muse?
Anything that’s really about a relationship or a values decision: how you talk to your spouse, what you teach your kids, what you believe. Let it handle logistics around those things, and keep the decisions and the conversations for yourself.
Do I need special prompts for Muse?
No special language. Use a simple brief: the goal, the limits, what done looks like and when to stop and ask. The template on this page works for any task.