AI Booking Chatbots for NZ Clinics: A Step-by-Step Guide

Reception teams in physiotherapy, dental, GP and allied health clinics answer the same questions all day: Are you open on Saturday? Where do I park? Do you treat this? How do I book? A chatbot on your website can answer those at any hour and point people to a booking.

A clinic chatbot is a good fit for:

• opening hours, location, parking and accessibility;
• which services you offer and which practitioner does what;
• prices, if you publish them, and your cancellation policy;
• what to bring to a first appointment;
• helping someone find and make a booking.

It should not give clinical advice or try to triage symptoms. Anyone describing an emergency should be told to call 111, and anyone with a health question should be pointed to a clinician. Write those rules down before anything is built; they decide everything else.

In New Zealand, health information collected and held by health agencies is covered by the Health Information Privacy Code 2020, which takes the place of the Privacy Act's information privacy principles for the health sector (Office of the Privacy Commissioner).

Before launch, be able to answer these questions:

• What does the chatbot collect? Ask only for what a booking needs. A chatbot answering "where do I park?" needs no personal details at all.
• Where does it go? Most chatbots send each message to an AI provider's servers, which may be outside New Zealand. Know which provider, where it processes and stores data, and what its terms say about retention and training.
• Who can read transcripts, for how long are they kept, and how are they deleted?
• What do you tell people? Say plainly in the chat window what is collected and why, and link your privacy statement.

Check your answers against the code itself, or get privacy advice, before you go live.

There are two common ways to handle bookings:

1. Link to your existing online booking page. The chatbot answers questions, then hands the person a link to book in your current system. It is the simplest option, and patient details never pass through the chatbot.
2. Integrate with your practice management system's API. If your system offers an API, the chatbot can check availability and create a booking inside the conversation. Read the vendor's API documentation and terms first: what is available differs between systems, and some charge for API access.

Either way, let your booking system send the confirmation and reminder messages it already sends, so there is one source of truth for appointments.

A chatbot is only as good as what it has been given to work from.

• Collect the source material: service descriptions, practitioner profiles, prices, policies, and the questions reception actually gets (ask them to jot these down for a fortnight).
• Approve the answers. Have someone at the clinic sign off on what the chatbot may say, especially anything about prices, funding or ACC.
• Design the handovers. During opening hours, offer a phone number or a callback. After hours, take a name and contact details and say when the clinic will reply.
• Set the tone: friendly, short and plain, in the clinic's own voice.

Test with the people who know the questions best: your reception team.

1. Everyday questions: hours, parking, services and prices. Check every answer against your approved material.
2. Things it must refuse: symptoms, medication questions, requests for another patient's information.
3. Emergencies: any mention of chest pain, breathing trouble or self-harm should get the 111 message straight away.
4. Made-up answers: ask about a service you don't offer, or a price you don't publish. It should say it doesn't know and offer a handover, not invent one.
5. Off-topic and hostile messages, and attempts to make it ignore its instructions.
6. Phones and screen readers: check the widget works on a small screen and with assistive technology.

Fix what you find and test again before you switch it on.

Decide what you will track before launch, so you have something to compare against:

• conversations started, and how many end with a booking link clicked or a booking made;
• how often it hands over to a person, and why;
• questions it couldn't answer (each one is a gap in the material);
• phone and email enquiry volume at reception, compared with the same period before launch.

Review a sample of conversations regularly, within the limits of your privacy settings, and update the approved answers as services, prices and staff change.

• Hosted chatbot platform: quick to start, usually a monthly subscription, and integrations limited to what the platform supports. Check where it stores data and whether it can link to your booking system.
• Custom build: fits your own systems and booking flow, and you own it. It costs a build fee plus the AI provider's usage charges.

Infynt builds the custom kind. Our AI Customer Support Chatbot starts from $1,250, includes a Privacy Act 2020 compliance review and human handover, and AI usage is paid at cost, estimated for your traffic before setup. See current packages and prices.