Chatbot for patient service at the HOF clinic

Answering messages takes time, and the patient who writes at night wants an immediate reply. It is natural for many orofacial harmonization (HOF) clinics to think about a chatbot: an automatic reply that welcomes, informs and routes. The technology helps, but there is a clear boundary. The chatbot can handle the operational side; the clinical decision and the relationship with the patient remain human.

This guide separates what is worth automating from what requires a person, shows how to design a safe flow, points out privacy and ethics risks and offers a rollout checklist. For the channel context, see WhatsApp Business for the HOF clinic and response time on the clinic's WhatsApp.

What is a chatbot and what types exist?

A chatbot is a program that chats by message. There are two families, with different risks:

For a clinic, a rule-based flow is usually the safest starting point. If you use generative AI, restrict what it can answer and review the conversations. See AI in the HOF clinic: uses and limits.

What is worth automating?

Simple, repetitive tasks with no clinical decision:

• Opening hours and address, with a map link.
• Appointment confirmation and reminder, with a reschedule option. See automatic reminders for HOF patients.
• First welcome outside business hours: "we received your message and will reply soon".
• Assessment scheduling, within the schedule's rules. See online scheduling for the HOF clinic.
• Administrative FAQs: payment methods, parking, what to bring.
• Collecting basic data for registration (name and contact).

What requires a person?

Everything that involves assessment, risk, emotion or decision:

• Clinical questions: "is this procedure right for me?", "can I have it with this condition?".
• Concerns or complications: a patient with pain, swelling, color change or any worrying sign needs immediate human attention. See patient calls in the middle of the night with a complication.
• Negotiating prices and terms.
• Complaints or dissatisfaction. See dissatisfied patient: how to handle it.
• Sensitive health data: medical history, allergies, medications, which should be collected in person or through a secure channel.

How do you design a safe flow?

The best flow is short, honest and has an easy exit to a person. One possible sequence:

Test the flow as if you were a patient, on your phone, during and outside business hours. Ask things outside the script and see whether the bot gets lost or says something inappropriate.

What about advertising rules and ethics?

The chatbot is a communication channel of the clinic and is subject to the same rules as the rest of the material. Precautions:

1. No promise of results in any reply.
2. No price or promotion that goes against your council's rules. See how much does it cost in DMs and advertising the price of harmonization.
3. No procedure recommendation before the assessment.
4. No comparison with other professionals.
5. Identification of the professional and the clinic, when required.

Review the bot's texts with the same rigor you would apply to a brochure.

What LGPD precautions does the chatbot require?

Every conversation generates data: name, phone, sometimes a photo and a health account. To stay in compliance (the LGPD is Brazil's General Data Protection Law):

• State the purpose and ask for consent before collecting data.
• Collect the minimum. Health data should not be requested by chatbot without need and without security.
• Define where conversations are kept and who reads them. See data security in the HOF clinic.
• Vet the vendor: where it stores data, whether there is a contract and how it deletes.
• Let the patient delete their data or opt out of contact.
• Be careful with photos: if the patient sends a photo through the bot, treat it as sensitive data. See sending patient photos over WhatsApp and the LGPD.

How do you roll it out without creating problems?

How do you know whether the chatbot helps?

Track a few numbers: first-response time, how many conversations were resolved without a human, how many were escalated, how many bookings resulted and how many complaints came up. Read a sample of conversations every month: that is where you find the errors the numbers hide. If patients complain about "not being able to talk to anyone", the flow is too closed. See the sales funnel of the HOF clinic to fit the channel into the whole.

What does Harmy have to do with this?

Harmy is not a chatbot and does not serve the patient: it is a professional's tool, used during the consultation. The chatbot handles the front door (information and scheduling), and the app comes in afterward, when you present the illustrative projection to the patient.

What common mistakes sink a clinic chatbot?

The first is a menu that is too long, which tires someone who only wanted a time slot. The second is having no exit to a human, leaving the patient stuck in a loop of options. The third is answering off-topic with confidence, typical of an unrestricted AI. The fourth is asking for health data on an insecure channel. The fifth is abandoning the bot after setting it up: outdated texts, wrong hours and broken links give an impression of carelessness. Set aside one day a month to test the whole flow and update what changed.

Is it worth it for a small clinic?

Not always. If the clinic receives few messages and the front desk answers well, a chatbot can be cost and complexity with no return. In those cases, saved quick replies in the messaging app, an away message outside business hours and a good service script already solve much of the problem. Consider a chatbot when the volume of repetitive questions is high, when the team wastes time on simple information and when there are many messages outside business hours.