Bring Autopilot into your Team's conversations
Priya's Autopilot has been on the Alpha launch for two weeks. It flags decisions that stall, spots dates that move, and brings her a short summary each morning. The noticing is handled.
The questions aren't. "Did the vendor sign-off move?" "Is the security review still blocking?" "Which date are we using for beta?" Her teammates still bring them to Priya, and they wait while she sits in back-to-back reviews. Most of the answers are already sitting with her Autopilot.
This unit shows you how to let your team ask your Autopilot directly, and how to decide who can reach it and what it can share.
From an out-of-office reply to someone covering for you
When you're away, an out-of-office reply tells people you can't help right now. A colleague covering for you answers the question. Bringing Autopilot into your team's conversations moves you from the first to the second.
Your Autopilot already has its own name, avatar, and presence in Teams. You can chat with it directly, and you can bring it into a group chat or channel where people @mention it by name. Because it's been working on your brief, it answers from the context it's already built up, not from a cold start.
For Priya, that means the engineering lead can ask in the launch chat and get an answer in minutes instead of waiting for her review to end.
Bring it into a conversation
Getting started takes three steps.
- Turn on team collaboration. This access is off by default. Your Autopilot works with other people only after you turn it on.
- Add it to the conversation where the work happens. Start from your one-to-one chat with your Autopilot, create a group chat, and add teammates. If channels are available in your environment, you can add it there too.
- @mention it by name. Type @, the name you gave your Autopilot, and then the request. Teammates in the conversation can do the same, within the limits you set.
Priya named her Autopilot Rio. In the launch chat, the engineering lead types:
"@Rio, summarize where we landed on the launch checklist and list anything still waiting on a decision."
Rio replies in the conversation under its own name. It draws on the checklist, the decision log, and the launch channel it's already watching. It can reply directly in shared conversations within the limits you set. Consequential actions still wait for your approval.
Note
Autopilot is in preview. Where you can add it and how it behaves in shared conversations might change.
Decide who can reach it and what it can share
In the previous unit, one of your access and sharing decisions was People: can it work with colleagues on your behalf, and which ones? Inviting your Autopilot into a conversation is where that decision becomes real.
Your Autopilot works within your permissions. Before you invite it in, decide what it should answer from, the same way you'd brief a colleague before they covered for you.
| Decision | Question to answer |
|---|---|
| Who | Which conversations should it join, and who can @mention it? |
| What | Which topics and sources should it answer from? |
| How far | What can it answer on its own, and what should it bring to you first? |
Put your answers in the brief, in plain language. Priya adds this to hers:
"When the launch team asks about Alpha, answer from the launch checklist, the decision log, and the launch channel. Don't share anything from my email or calendar. If a question is about budget or a vendor contract, bring it to me before you reply."
It names the sources, draws a line around what's private, and sets a point where Autopilot stops and asks. It's the same shape as the brief in the previous unit, pointed at people instead of signals.
You can review these choices at any time in the Autopilot side pane, under Working with other people. That's where you control how your Autopilot shares work and acts on your behalf. Access is off by default. Before you enable shared replies, decide which sources Autopilot should use and what information it should share. Review its replies to check that they stay within your intended scope.
Important
The protections you already know still hold in a shared conversation. Autopilot can't reach anything you can't reach yourself, and it asks before consequential actions, such as sending a message or sharing a file.
It's your Autopilot in a shared space
When Rio answers in the launch chat, everyone sees its name, not Priya's. That makes its replies easy to tell apart from hers. It doesn't make them someone else's responsibility.
Autopilot is a personal agent. It can answer your teammates within the limits you set, but it isn't a shared team agent. It works for you. That changes three things:
- You own what it says. For the first few weeks, read its replies in the conversation. If it answers outside the scope you meant, narrow the brief.
- Your teammates should know what it's for. When you add it, post a short note saying what it can answer and what still comes to you. A minute of setup saves a week of confusion.
- Every answer counts. Autopilot usage consumes Copilot Credits, and higher-volume shared use adds up quickly.
Tip
Start with one conversation and one kind of question, such as status on a single project. Broaden it once you trust the answers.
Think about your own work.
What question do people ask you every week that you answer from the same few sources? That's usually the first one worth handing to your Autopilot.