Talking to a human
A visitor can always reach a person. The AI’s very first line says so — “Want a person instead? Just ask.” — and when they do, the conversation moves to your team. This page covers both sides: what visitors see, and how your team gets told and takes over.
Can I talk to a human?
Yes. A visitor just asks — “can I talk to a person?”, in any wording — and the AI hands the conversation off, confirming right in the chat that the team has been brought in and will follow up there.
The AI also hands off on its own, without being asked, when:
- it cannot answer from the Knowledge Base — the conversation is flagged for your team instead of left hanging,
- something goes wrong on the AI’s side — it never leaves a visitor talking to a wall.
If the visitor shared an email address, replies from your team also reach them by email, so leaving the chat open is not required.
Are you AI responding or a person?
Visitors can always tell which is which. AI replies carry a ✨ icon, the assistant’s name, and a visible AI tag — even if you named your assistant something human-sounding. Messages from a person show that teammate’s initials and first name, and no AI tag. Visitors are never left guessing, and the AI never claims to be human. See How the AI answers for the full honesty story.
What visitors see when a human joins
When a teammate replies, the sender line above the message switches from the AI’s name to the teammate’s name and initials. The chat header also carries a live availability line: a green dot with “Typically replies in a few minutes” when your team is around, or an amber dot with “Away right now — leave a message and we’ll reply by email” when nobody is online.
How your team gets notified
- When the AI hands a conversation off, every owner, admin, and member gets an immediate email and push notification — it reads “A conversation was handed to your team”, with the note “The assistant escalated it — open it to take over.” These alerts are deliberately loud: they go out even if a teammate already has the inbox open somewhere.
- When a visitor sends an ordinary message, an email goes to the team only if nobody is in the inbox at the time, and a burst of messages collapses into one email — you are not spammed per message.
- On your phone, install the admin app for push notifications — see Mobile notifications.
Each person controls their own alerts under Settings → Notifications: email and push can be toggled separately, and email is the durable backstop — an overnight message is still in your inbox in the morning.
Finding conversations that need you
On the Chats page:
- A conversation waiting on your reply shows in bold with a count bubble of unread visitor messages (it clears when anyone on the team opens the conversation).
- A conversation the AI handed off shows a hollow amber ring on its row. The Escalated filter pill at the top shows only these.
- Inside an escalated conversation, an amber Escalated chip in the header explains why on hover — for example “AI didn’t have an answer”.
Taking over
Reply. That’s the whole move. The moment a teammate sends a message, the AI pauses on that conversation — the AI pill above the reply box reads “AI · Paused — you’re handling”. The AI does not talk over a human, ever.
Also available:
- Claim — the button in the conversation header assigns the conversation to you, so teammates can see who owns it. Admins can assign conversations to others.
- Return to AI — in the AI pill or the ⋯ menu, when you’re done and want the assistant to resume.
- Automatic hand-back — in Autopilot Workspaces with “Hand back to AI after (hours)” set, the assistant resumes a conversation after that many quiet hours, on the visitor’s next message. Set it to 0 and the AI never takes a conversation back on its own.
One safety net: in an Autopilot Workspace, if you take over and then go quiet for 15 minutes while the visitor is still waiting, the conversation returns to the team queue and the team is alerted (“Nobody followed up — open it to help this visitor”) — a takeover can’t silently strand anyone.
What do the conversation status colors mean?
The badge in the conversation header shows the conversation’s state:
| Badge | Color | Meaning |
|---|---|---|
| New | Blue | Nobody has picked it up yet |
| Open | Amber | Being worked |
| Waiting on visitor | Purple | Your side replied; the visitor hasn’t |
| Resolved | Green | Done |
In the conversation list, rows stay clean and only exceptions get a dot: a purple dot means waiting on the visitor, a grey dot means resolved or closed, and the hollow amber ring means escalated to your team. No dot means a normal open conversation. Hover any dot for its label.
What do the visitor presence dots mean?
The small dot on a visitor’s avatar in the list:
- Green — online in the chat right now.
- Amber — idle: recently active, gone quiet for a moment.
- No dot — we don’t have presence information. No dot never means “offline for sure”; it means unknown.
Inside a conversation, the header shows the same thing with more detail: Online, then Idle after roughly 45 quiet seconds, then “Last seen 3m ago” once they’ve been gone about five minutes. Presence is approximate — it tells you whether a live reply is likely to land, not what the visitor is doing.