AI chatbot for clinics and medical practices, what it handles and what it must not
Reception answers the same questions all day while the waiting room fills up. How an AI chatbot handles opening hours, self-pay prices and pre-appointment instructions, where it should deliberately stop, and what we are building for booking.
The phone at reception rings, a patient is waiting at the desk, and five more sit in the waiting room. Most of those calls are not about health. They are opening hours, prices, parking, and whether you are taking new patients. An AI chatbot on your website can take that traffic, so your staff can look after the people standing in front of them.
The questions that repeat every day
Go through one week of calls and the same handful of topics comes back:
- When you are open, when you are closed for holidays, and who covers
- What a procedure costs when insurance does not pay for it
- Whether you accept new patients, and which insurers you work with
- What to bring and how to prepare for a procedure
- Where to park and how to reach you by public transport
- How to book an appointment and how to cancel one
None of that is a medical question. It is practical information you already have on your website, in your price list, or in a pre-appointment leaflet.
What the chatbot does today
- It answers from your own material. Upload your price list, practice rules and preparation instructions, or let the bot read your website, and it answers from those. The method is in how to train your chatbot.
- It answers in the patient's language. In bigger cities a share of your patients write in another language. The bot replies in whichever language it was asked, with no second version of your website.
- It works in the evening and at weekends. That is exactly when people check whether you are open on Monday.
- It hands over to a human. When a question goes past practical information, a member of staff takes the conversation over. Human takeover (Take Control) is included from the Pro plan and during the trial.
- It reaches you on your phone. The Breezaro Companion mobile app sends a push when somebody asks for a person. It is free with every plan.
- It answers on WhatsApp too. From the Standard plan you can add WhatsApp, Messenger and Instagram, all on the same knowledge base.
Where the bot must deliberately stop
In healthcare this matters more than anywhere else. A chatbot that invents an answer is an annoyance elsewhere and a serious problem here. So the setup needs clear limits:
- No advice, no diagnosis. The bot should not assess symptoms or suggest treatment or medication. A medical question belongs to a human, not to a guess.
- No patient data. Nothing about individual people belongs in the knowledge base: no records, no results, no names. The bot should know your price list and your practical information, and nothing more. The background is in chatbots and GDPR.
- No emergencies. Your welcome message should say that anyone with a sudden medical problem needs to call emergency services rather than write in a chat.
- No promised appointments. Until a person confirms a booking, the bot explains how to book, it does not say you are booked.
These limits are not extra restrictions. They are the reason a patient believes the rest of what the bot tells them.
What we are building for booking
The request we hear most often from practices is booking an appointment straight from the chat. Two things lead there, and both are in progress:
- Custom actions let the bot reach into your systems and work with current data instead of static text. How they work is in custom actions.
- Webhooks push an event out the moment it happens, so a request for a human can become a Slack message or a row in a spreadsheet. There is a worked example in when a visitor asks for a human.
Actions that write to a system, which is what creating an appointment does, are not switched on yet. We want certainty about what gets sent and who confirmed it first. So treat booking inside the conversation as something we are building, not as today's state.
Setting it up in an afternoon
- Put the widget on your website and upload your price list, preparation instructions and practice rules.
- In the welcome message, say what the bot helps with and what belongs on the phone, including the line about emergencies.
- Turn on human takeover and phone notifications so a request does not sit unread.
FAQ
Can a chatbot advise patients about symptoms? It should not. Set it up so medical questions go to a human. Its job is practical: hours, prices, preparation and directions.
Does patient data belong in the knowledge base? No. Give it public, practical information only. The bot should not know anything about individual patients or their records.
Can a patient book an appointment in the chat? Not yet. The bot explains how to book and hands you over to staff. Booking inside the conversation is coming, through custom actions.
Will it answer people who do not speak the local language? Yes. It replies in whichever language it was asked, so you do not need a translated website.
A practice does not need a chatbot that acts like a doctor. It needs one that reliably clears the practical questions, steps back from everything else, and calls a human. The trial runs 30 days with no card, and the plans are on the pricing page.
Related: Chatbots and GDPR · Why does an AI chatbot make things up.