How to use Copilot Voice and save your next steps
Start a spoken conversation with Copilot, ask a focused question, and keep the useful conclusions and next steps when you finish.
Updated 5 min read
To use Copilot Voice in the experience Microsoft documents, choose Start a new voice chat in the Message Copilot prompt box and allow microphone access. Its voice and vision instructions (opens in a new tab) give that starting route; check the instructions for the account and app you use before following it. Start with one question you can discuss aloud, then write down the useful next step. Tools like fenbs keep that work so you can pick it up on another device or with another connected AI.
Where do you start a spoken conversation?
- Open Copilot and find the Message Copilot prompt box. These steps follow the desktop route in the Microsoft instructions linked above.
- Choose Start a new voice chat. On Windows, Microsoft also documents the Copilot key or
Win+Cas a starting route. - Allow microphone access when your device asks for it. A spoken session needs that permission.
- Say what you are trying to accomplish, then ask one specific question. For example: “Help me prepare three questions for a website review.”
- Listen to the answer and ask a follow-up that narrows it. Before leaving the conversation, ask for a short recap and check that it reflects what you actually agreed.
Keep the first attempt small. You are checking whether the conversation helps you think through the problem, so choose a question that does not need private files or a complicated setup. Write down the result yourself if that is the simplest way to keep it. A spoken exchange is useful only if you can explain the next action afterward.
What should you say in your first question?
Give the conversation a destination. Say what you need to finish, who it is for, and the one constraint that matters most. “Help me plan a review” leaves a lot open. “Help me prepare a ten-minute review of a restaurant menu website, focused on phone readability” gives you something you can judge.
Ask for one question at a time if you want a discussion. Ask for a short answer if you need a quick explanation. These are ways to make your request easier to follow, rather than special settings you must find in a menu. You can also tell Copilot to distinguish ideas worth exploring from steps you have chosen to take.
- Name the goal: “I want to leave with three questions for the designer.”
- Name the boundary: “Do not assume we can change the checkout system.”
- Name the format: “Give me one suggestion, then let me respond.”
- Name the finish: “End with what I should do next and what still needs checking.”
For more ways to structure the request, use the separate Microsoft Copilot prompts guide. Save a useful spoken opening alongside your next steps in fenbs so you can reuse both; how fenbs works explains what you can keep.
Do voice and visual sharing have the same requirements?
Do not use a visual-sharing prerequisite as a blanket rule for every spoken conversation. Microsoft’s Vision getting-started guide (opens in a new tab) names specific account and subscription requirements for visual help during voice. It documents desktop or web screen sharing on Windows and camera sharing in the mobile app. Check those requirements if you want Copilot to see something as well as hear you.
For this workflow, begin with the voice control available in your account. Adding a shared screen is a separate action. If your question depends on a diagram or an interface, describe what you need before adding visual input, and show only material you are allowed to share. The separate Copilot Vision guide covers that visual workflow.
How do you check a spoken answer?
Treat a fluent answer as a suggestion you still need to evaluate. Repeat the key claim in your own words and ask what evidence would confirm it. If the conversation concerns a website review, for example, distinguish “the text looks difficult to read” from “we tested the text with customers.” The second statement needs evidence beyond the discussion.
Before acting, write a short list with three parts: what you observed, what Copilot suggested, and what you decided. Check any names, dates, measurements, or instructions against the source you would normally trust. If you cannot check something now, keep it as an open question instead of turning it into a fact.
This is also where a direct voice session differs from processing a team meeting. A one-to-one conversation can help you prepare or think, while a meeting has other people’s agreements to preserve. For that separate job, see Copilot for meetings and follow-ups.
What should you check if the session goes wrong?
First check the permission that the documented workflow requires: microphone access. Then check that you started a voice conversation rather than simply writing in the message box. Keep your next attempt to a short question whose answer you can recognize. That makes it easier to tell whether the problem is input, understanding, or the complexity of the request.
When visual sharing is involved, selecting Stop sharing leaves voice running, according to Microsoft’s usage guide (opens in a new tab). Check whether the voice conversation is still active when you finish. For a session that will not start or has no sound, use the separate Copilot Voice troubleshooting guide.
What is worth keeping after you talk?
Keep the outcome that someone can use without hearing the whole exchange: the goal, the answer you checked, the option you chose, and the next step. Include one unanswered question if it blocks progress. A page of polished conversation is less useful than five accurate sentences about where the work stands.
If you want editable written content inside Copilot, eligible web users can create persistent Pages through the routes in Microsoft’s Copilot Pages guide (opens in a new tab). Keep a reviewed summary separately instead of assuming every spoken exchange was archived. For finding earlier conversations, use the Copilot history guide.
Hand work from one AI to another
When one AI stops, save where it got to in fenbs. The next AI you open picks up right there, so nothing is done twice.
Create a free accountFree for up to 5 people and 50 open tasks · No credit card · Works with Claude, ChatGPT, Cursor and more
How to keep spoken next steps with fenbs
Keep the useful answer and your next step so you can carry on from exactly where you left off, on any device and in another connected AI. Copy your checked conclusions into fenbs yourself. The GitHub Copilot connection below belongs to a separate coding assistant.
- Create a free account.
- Write the useful result in plain words: what you were trying to do, what you checked, what you chose, and what you will do next.
- Add a task for the next action. Keep an unanswered question beside it so the next person does not mistake a suggestion for a decision.
- Connect GitHub Copilot to fenbs if you also use that coding assistant. Ask it to read your saved next step before continuing, then save what it did before you stop.
GitHub’s guide to connecting outside tools (opens in a new tab) documents that separate setup. Those instructions apply to GitHub Copilot, rather than a control inside ordinary Copilot Voice. fenbs is useful when the discussion creates work you need to resume; a quick question with no lasting next step may need only the answer. Browse more practical workflows in AI at work.