Skip to main content
When a customer texts your ServiceBooked number, the AI Receptionist reads it, drafts a response, and sends — all within seconds. You can intervene any time.

​
How a typical conversation goes

  1. Customer texts your number with a question or request.
  2. AI receives it via webhook — instantly, no polling delay.
  3. AI checks context: who is this person? do we have history? what services apply?
  4. AI generates a reply using your business hours, services, knowledge base, and conversation history.
  5. AI sends the reply within ~5 seconds, from your business number.
  6. Customer replies again — back to step 2.
  7. At any point you can take over — type a manual reply in the inbox, and the AI Receptionist defers to you.

​
The AI’s safety nets

A few cases where the AI Receptionist explicitly doesn’t reply:
  • STOP / opt-out — customer texts “STOP” or “UNSUBSCRIBE”, we honor it instantly. They’re added to your opt-out list and we won’t text them again until they reactivate.
  • HELP — customer texts “HELP”, we send the standard “ServiceBooked helps service businesses respond to leads. Reply STOP to opt out” disclosure.
  • Outside business hours (configurable) — for some orgs, the AI Receptionist replies anyway with a disclaimer (“we’re closed but I can take info”). For others it stays silent until you reply manually. Toggle in Settings → Business → AI hours behavior.
  • No subscription / past_due — if your billing isn’t current, AI replies pause. Customers get a polite “we’re unavailable right now” auto-message; you fix billing and the AI Receptionist resumes.

​
Reading a conversation

Click any SMS conversation in the inbox to see the full thread, colour-coded by side:
  • Gray (left) — the customer’s incoming messages.
  • Blue (right) — messages sent from your number. This covers both the AI Receptionist’s auto-replies and manual replies you typed — they sit on the same side of the thread.
The AI’s reasoning isn’t shown in the thread (we don’t want to clutter the UX), but if you want to see why the AI Receptionist said something specific, the conversation log on the back-end has reasoning attached. Talk to support if you need to debug a specific exchange.

​
Taking over the conversation

​
Type a manual reply — the AI steps back

Just type into the message box at the bottom of the conversation and hit Send. Your reply goes out from your business number, and the AI Receptionist automatically pauses for that conversation — the moment a real person replies, it stops auto-answering so it can’t talk over you. This works from either place you can text a customer: the conversation view, and the reply box in an Inbox thread. Both put the message into the same thread and pause the AI the same way. You’ll see an amber “AI paused — you’re handling this conversation” banner above the reply box. It stays paused for that thread until you decide otherwise — the AI won’t quietly jump back in.

​
Resume the AI

When you’re done and want the AI Receptionist to take the thread again, click Resume AI in that banner. From then on it auto-answers new incoming texts as usual. Resuming is always a deliberate choice — nothing brings the AI back on its own.

​
Mark resolved / close conversation

When the conversation is done, click Close. Future texts from this customer will start a fresh conversation thread.

​
AI reply mode (org-wide)

The steps above intervene on a single conversation. To change the default for every inbound message — SMS, social DMs, and email — set your AI reply mode at the top of Settings → Automations:
  • Auto — the AI Receptionist replies automatically (the default described above).
  • Suggest — it writes the reply but doesn’t send it. Every draft waits for you under AI drafts.
  • Off — it never auto-replies; you handle every message.
Use Pause AI for a one-off thread; use the reply mode when you want the change to stick across the whole account.

​
Reviewing drafts before they send

In Suggest mode, everything the AI writes lands on the AI drafts page in your sidebar — SMS, social DMs, email replies, and website form replies, oldest first, with a count beside the menu item so you can see at a glance how many are waiting. Each one shows who it’s going to, which channel it will go out on, and the full message. Edit the text if you want to change a line, then Send — or Discard if it shouldn’t go at all. Nothing on that page has been sent to anyone. Two things worth knowing:
  • Drafts never send themselves. If nobody reviews one, the customer hears nothing — so keep an eye on the count. You’ll also get a notification each time a draft is written.
  • Live chat and phone calls are never drafted. Someone is waiting on the line, so the AI Receptionist answers those directly whatever your reply mode says.

​
Throttling and delivery

We rate-limit outbound SMS per-customer to prevent spam loops (e.g., a confused conversation where the customer keeps replying “What?” and the AI Receptionist keeps trying). Maximum ~10 messages per side per hour to a given recipient. If a delivery fails (carrier filtering, dead/invalid number, the recipient opted out, etc.), we mark the message Failed in the thread. In Canada that’s usually a bad number or carrier filtering. If you text US recipients, a common cause is missing A2P 10DLC registration (US-only — see A2P 10DLC).

​
Language

The AI replies in whatever language the customer wrote in (when it can detect). For best results in non-English markets, set your default voice/AI language in Settings → Voice → Language.

​
Group SMS / MMS

Not currently supported — only one-on-one SMS. MMS messages (with images attached) are received but the AI Receptionist doesn’t process the image; it’ll reply based on the text-only message.

​
Next

Web chat

The same AI but on your website widget.

Voice calls

Phone calls handled by the AI Receptionist.