Copilot Chat is often the first part of Copilot people try. A broad first prompt can produce an equally broad answer, which makes the tool easy to dismiss. Better results start with a few practical checks: which account is active, what evidence Copilot can reach, and whether the answer holds up against that evidence.
The four lessons follow the way the work tends to unfold. You'll start a controlled chat on desktop, web, or mobile, then decide whether a question calls for work data, the public web, or text supplied in the prompt. You'll also learn to point Copilot to a particular file, meeting, or person. Later, you'll clean up prompts worth keeping, prepare them for scheduling or sharing, compare visible model settings fairly, and use Create for assets such as posters, forms, and short videos.
Copilot can ground an answer only in material it can reach. Choose the source, add or reference it when the task needs one, and check the entire response against that source before you use it or send it on.
What you'll walk away with
Open Copilot Chat with the right work identity, whether you're on the desktop app, the web, or your phone
Decide whether an answer should be grounded in work data, the public web, or text you paste in yourself
Reference an exact file, person, or meeting so the source is explicit
Reuse, schedule, and share your best prompts without leaking anything private
Compare models and response styles fairly before you trust one over another
Turn a single prompt into an image, video, infographic, poster, or form in the Create hub
