AI
Four chat modes. AI first is right for almost everybody.
The setting lives in Dashboard → Settings → AI & chat behaviour → Chat mode, and it applies to your whole account — every website and every connected channel. The difference between them is who answers the visitor and who has to be available. Both are below, mode by mode.
At a glance
Which mode, and what it costs you in staffing
| Mode | What the visitor gets | Who must be available |
|---|---|---|
| AI only | An AI answer to everything. The assistant never decides on its own to fetch a person. | Nobody. The visitor can still press Team in the chat box, and then somebody should reply. |
| AI first default | An AI answer, and a person when the conversation needs one. | Somebody reachable during your hours. Not every chat — only the handed-over ones. |
| Human first | One AI answer, then a person takes the conversation. | Somebody on every single conversation, quickly. |
| WhatsApp only | No AI at all. Their message is forwarded to your WhatsApp; the widget goes quiet. | Whoever holds that WhatsApp phone. Nothing is answered until they reply. |
Changing the mode takes effect on the next visitor message. A conversation a teammate has already taken over keeps its own state — the mode does not pull it back.
Mode by mode
AI only
The assistant handles everything and never hands over by itself.
- A low-confidence answer, a complaint or an angry message does not pull in a person. In the other modes it would.
- The visitor is not stranded, though: the chat box always shows an AI / Team switch, and pressing Team hands the conversation over and alerts your team. That button is never blocked by the chat mode.
- The button is the route, not the words. In this mode, typing “I want to talk to a person” as a message does not hand over — the assistant simply answers it. Only the Team switch does. If you choose AI only, say so in your welcome message and point visitors at that button.
- Good for: a business with no one to staff chat, an out-of-hours default, or a site where the assistant genuinely can answer nearly everything from your knowledge base.
- The risk: nobody is watching for a chat that is going badly. If you pick this, make sure your alerts actually reach a phone, so the Team requests do not pile up unseen.
Mode by mode
AI first — recommended
The assistant answers, and brings in a person when there is a real reason to. This is the default for new accounts.
- The reasons that trigger a hand-over are: the visitor asking for a person, a complaint, clear frustration, anger or urgency, and someone reporting they cannot get into their account. On accounts still on the older reply behaviour, an answer the assistant is not confident about also hands over.
- A plain gap in your knowledge base is deliberately not a hand-over on newer accounts. The assistant says it does not have the detail and points at the right page instead of dragging a human into a question nobody can answer either.
- Important, and often misread as a bug: a hand-over only takes the conversation away from the AI when somebody is actually reachable. With nobody online, your team is still alerted and the lead is still recorded, but the assistant stays on the line rather than leaving the visitor talking to an empty room.
- Good for: almost everyone. It is the only mode where being short-staffed for an hour degrades gracefully instead of going silent.
Mode by mode
Human first
The assistant gets one answer, then the conversation belongs to a person.
- On the website widget: the assistant answers, and once it has answered the conversation is marked for a person — in practice, from the visitor's second message onwards the chat is a human conversation.
- On WhatsApp, Telegram, Instagram and Messenger: the assistant answers once and then stays quiet. Your team is alerted so somebody picks it up.
- One exception, on purpose: if the visitor has explicitly chosen to keep talking to the AI, that choice is respected and the automatic hand-over is skipped. The visitor's wishes outrank the setting.
- Good for: high-value enquiries — clinics, law firms, anything where a person closes the deal and a bot answer is worse than a slower human one.
- The risk, stated bluntly: this mode assumes somebody is there. If nobody is, the visitor gets one AI answer and then a wait. Only choose it if your alerts land on a phone somebody is holding.
Mode by mode
WhatsApp only
No AI. The widget becomes a way into your WhatsApp.
- The visitor types in the widget; the message is forwarded to the WhatsApp number stored on your account. The assistant writes no reply at all. The chat box switches itself to Team and posts one short notice — “You're now chatting with the team — they'll reply here shortly.” — and after that the widget deliberately says nothing else, because the conversation continues on your WhatsApp and a chatty widget on top of that would talk over you.
- The number used is the one set when you connected WhatsApp under Channels. There is no separate field for it in the dashboard today.
- The failure mode to know about: if no WhatsApp number is stored, or the WhatsApp connection is not actually working, the forward is dropped. The visitor sees that same notice and then silence, so nothing on their side looks wrong — and nothing tells you the message existed. Send yourself a test enquiry after switching to this mode, before you rely on it.
- Your teammates also get no private AI suggestions in this mode — the assumption is that they are on a phone, not in the portal.
- Good for: a business that already runs entirely on WhatsApp and wants the website to feed it. Everyone else is better served by AI first with WhatsApp connected as a channel.
What outranks the mode
Three things that override whatever you chose
- A teammate claiming the chat. Once a person takes a conversation, the assistant stops answering the visitor in it. It may still send that teammate a private suggestion, which the visitor never sees.
- The visitor's own choice. The AI / Team switch in the chat box is a real control, not a label. Pressing Team works in every mode, and switching back to AI works too.
- The safety guard. Attempts to manipulate the assistant into ignoring its instructions are deflected in every mode, including while a human is handling the chat. Your teammate is not asked to forward an attack.
How
Change the mode
- Sign in at
app.orcalinq.com. - Sidebar → Account → Settings.
- Find the AI & chat behaviour card and open the Chat mode dropdown. The line under it describes the mode you have selected.
- Press Save changes at the bottom of the page. One button saves the whole page.
No Settings item in your sidebar? Settings access is granted per person by the account owner. Ask them, or sign in as the owner.
If that did not work
When the mode does not behave as expected
- Saved the mode but the conversation you are watching has not changed. The mode applies from the next visitor message, and a chat already handed to a person stays with that person. Test in a fresh private window.
- WhatsApp only, and nothing arrives on the phone. Open Channels → WhatsApp and use Send a test message to a number you hold. If the test fails, the mode cannot work — fix the connection first.
- Human first, and visitors are waiting. Nobody was reachable. Check that at least one teammate has an alert channel that reaches their phone, and that their working hours cover the time visitors actually write.
- AI first, and nothing ever reaches a person. Hand-over needs a reachable target. If no teammate has a working alert channel, the assistant keeps answering by design — that is the safer failure, but it looks like the setting is ignored.
- AI only, and a chat was handed over anyway. The visitor pressed Team, or the assistant deflected an unsafe message. Both are intended and neither is affected by the mode.
- AI only, and a visitor said they wanted a human but nothing happened. They typed it instead of pressing Team. Typed requests do not hand over in this mode — that is what the mode means. If you want typed requests honoured, use AI first.
Still not matching what you see? Write to support@orcalinq.comwith the time of the conversation and the mode you have set.
Related pages
Keep exploring
Start on AI first. Change your mind later.
The mode is one dropdown and takes effect on the next message — nothing to reinstall.