Sending messages and getting answers Guide
Learn the core chat loop: type a request, send it, watch the assistant reply in real time, and follow live status as it works.
Last updated July 16, 2026
On this page
The assistant is where you get work done: you type a request in the composer, send it, and the reply streams back in real time. Beyond answering questions, the assistant can carry out multi-step tasks and shows you live status while it works.
Send your first message
The composer sits at the bottom of the chat. To send a request:
- Click into the composer and type your question or instruction.
- Press Enter or click Send.
- Watch the reply appear word by word as the assistant works. You do not need to wait for the whole answer before reading along.
If you are not sure where to begin, the welcome screen offers starter prompts grouped into categories such as Act, Create, Build an agent, Research, Plan, and Analyze. Select one to drop a ready-made prompt into the composer, then edit it before sending.
Tip: Be specific about the outcome you want. "Draft a two-paragraph follow-up email to a client who missed our meeting, friendly but brief" gives you a far better result than "write an email."
Watch the assistant work
Simple questions come back as a straight answer. For anything that involves a tool — searching the web, checking a connected app, drafting a document — the assistant shows live status as each step runs, so you can see what it is doing rather than staring at a blank screen.
When the assistant produces a deliverable such as a document, spreadsheet, image, or chart, it appears inline in the conversation where you can preview and download it.
Stop or regenerate a reply
You are always in control of a response in progress.
- Stop a reply: If an answer is heading in the wrong direction, click Stop while it is still generating. The assistant halts immediately, keeping whatever it produced so far.
- Regenerate: If the last answer was not quite what you wanted, ask for another version. You can also add a follow-up message telling it what to change — "make it shorter," "use a more formal tone," or "turn this into a bulleted list."
Steering the format is often faster than starting over. Tell the assistant the length, tone, or structure you need and it will adjust.
Use slash commands
Type / at the start of the composer to open quick commands that jump straight to a common action. Available commands include:
/task— create a task/reminder— set a reminder/note— capture a note/schedule— schedule something/image— generate an image/excel— create a spreadsheet/pdf— create a PDF/search— search the web and live sources
Pick a command, add your details, and send as usual.
Copy answers and code
You can copy any reply to reuse it elsewhere. When the assistant produces a code block, a dedicated copy control lets you grab just that block cleanly, without the surrounding text.
Why some actions ask for confirmation
The assistant can take actions that reach the outside world — sending an email or message, deleting something, or making a payment. Before any of these happens, it shows an Approve card summarizing exactly what it is about to do.
Nothing leaves your account until you approve it. Review the details on the card and confirm, or decline if it is not what you intended. This is why you will occasionally see a pause before something is sent or deleted — it is a deliberate safeguard, not an error.
Note: If the assistant says it completed an action but you want to be sure, check the Approve card in the conversation. An approved card confirms the action went through; if you never approved one, the action was not carried out.
Pick up where you left off
Your conversations stay in sync across the web app and any phone channels you have connected. You can start a request on your computer and continue it from your phone, and the assistant keeps the full context of the conversation either way.