Conversation AI (AI assistant)
Set up the AI assistant that suggests or sends replies to incoming messages, with safety rules and human handover.
3 min read
Conversation AI is the assistant that uses AI to reply to WhatsApp, Instagram, Facebook and other messages arriving in Conversations. It can answer a patient writing from London at midnight within seconds, ask for photos and book an online consultation. Set up well, it takes pressure off your coordinators; set up badly, it can give patients wrong information. So take it step by step.
Before you start
- You need the Admin role.
- What the assistant knows comes from the Knowledge base. Fill that in first.
- If you'll use the booking goal, have a consultation calendar ready: Create a calendar.
Create the bot and choose its mode and channels
- Open Settings → Conversation AI and create a new bot (or edit an existing one). Give it a clear name (e.g. Aura Assistant).
- Choose how it works under Bot Mode:
- Off: the bot is inactive.
- Suggestive: the bot drafts a reply in the message box; a coordinator reviews, edits and sends it.
- Auto-Pilot: the bot sends replies on its own.
- Select the channels the bot works on (WhatsApp, Instagram, Facebook, live chat, SMS…) and save.
Important
Run in Suggestive mode for the first 2–4 weeks. As your team reads and corrects the suggestions, you'll spot the gaps in your knowledge base. Once answers are consistently right, switch to Auto-Pilot for out-of-hours or a single channel first.
Goals, human handover and safety rules
- Bot Goals: define what the bot tries to achieve in a conversation. Typical goals in medical tourism:
- Book Appointment: guide the patient to the online consultation calendar.
- Collect photos: a smile photo, X-ray or photo of the hair-loss area. Photos stay in the patient's conversation and record.
- Collect information: country, treatment of interest, planned travel dates.
- Human handover: make the bot stop when a team member replies, and assign the conversation to a coordinator when the patient asks to speak to a person.
- Instructions: spell out what the bot must never do (see the rules below).
Safety rules for medical tourism
Always add these rules to the bot's instructions:
- No diagnosis, no treatment recommendations, no guaranteed results. "Our doctor will decide which treatment suits you after reviewing your photos."
- No final prices. If you want to share price ranges, write them clearly in the knowledge base; otherwise the bot should point to a consultation.
- In an emergency (severe pain, bleeding, fever, post-op complications), the bot directs the patient to local emergency services straight away and hands the conversation to the team.
- No medication, dosage or post-op medical instructions; only a doctor or nurse shares those.
Heads up
AI can produce wrong information. Read the bot's replies regularly, especially in Auto-Pilot mode. To the patient, what the bot says is your clinic's word.
Example flow
At 23:40, Jonas Weber from Germany writes on Instagram: "How long do I need to stay for implants?" The bot explains the general timeline from the FAQ in your knowledge base, says the exact plan depends on the doctor's assessment, asks for a panoramic X-ray and offers online consultation slots for the next day. In the morning a coordinator takes over the conversation.
Common issues
The bot doesn't reply. The mode may be Off, the channel may not be selected, or the bot stopped because a team member replied in the conversation.
The bot gives irrelevant or outdated information. Update the knowledge base sources and delete old documents.
The bot doesn't reply in the patient's language. Add the rule "Reply in the language the patient writes in" to the instructions.
Related articles
Didn't solve your problem?
Ask the assistant or message our support team.