AI assistant for dental practice: bookings, reminders, phone

An average Bulgarian dental practice loses 15-25 percent of potential revenue to no-shows. AI booking assistants change that number to 5-10 percent — measured in real conditions. This article describes what implementation includes, how much it costs, and what the traps are.

What the AI assistant handles

Concretely, in order of biggest benefit:

1. Answers the phone 24/7. A patient calls at 9 PM after work. AI answers with voice, understands the request, checks available slots, books the appointment. In the morning, reception sees the new booking in the system.
2. Sends reminders 24 hours before the appointment. Via Viber, SMS, or email, depending on patient preference. With one-click confirm or reschedule.
3. Confirmation call 4 hours before the appointment. If the patient does not respond to the reminder, AI makes a call and offers: confirm, reschedule, cancel.
4. Handles information inquiries. "How much does root canal cost," "do you have slots tomorrow morning," "do you accept children." Standard answers are handled automatically; non-standard ones go to reception.
5. Builds patient history. Records where the patient is in their treatment plan, when their next recommended visit is, when it is time for a checkup. Proactively sends reminders before they need to ask.

What it does NOT handle

AI is not a human replacement. Cases that must hand off:

• Complaints or dissatisfied patients. AI detects tone and immediately escalates to the manager.
• Emergencies. Pain, bleeding, broken tooth — directly calls the on-duty doctor, does not run a booking conversation.
• Complex medical questions. Treatment plan changes, X-ray interpretation — out of scope.
• Price negotiations or special conditions. Leave these to humans.

Expected effect

Numbers from a typical practice with three rooms and about 60 appointments per day:

• No-shows: from 22% to 8% in the first month, down to 5-6% after 3 months of optimization.
• Reception phone time: down about 60%.
• After-hours bookings: from 0 to about 18% of total volume (evenings and weekends).
• Response speed to inquiries: from an average of 4 hours to under a minute.
• Patient satisfaction in post-visit surveys: up about 12 points.

Technical architecture

Standard implementation has three components:

1. Voice AI agent for phone

Platform like Twilio + AI model (Claude or GPT-4) with calendar tool access. Natural conversation in real time. Bulgarian language works well since late 2024.

2. Chat integration (Viber, Messenger, website)

Same brain behind the phone works with text too. A patient can start in Messenger, continue on the website, finish on the phone — AI remembers context.

3. Integration with existing calendar system

This is the critical part. If you use DentApp, BoolaPay, or a custom solution, AI must read available slots and write bookings in real time, not against a copy. Without this, the assistant does not work.

Expected cost

For a practice with one to three rooms:

• Implementation: 5,000 - 9,000 EUR (depends on the calendar system and phone volume).
• Monthly infrastructure (Twilio + AI): about 200-400 EUR for a practice with 60-100 daily visits.
• Maintenance: 100-300 EUR monthly (minimum — security updates and fixes).

Payback: typically 2 to 4 months. The no-show reduction alone covers the monthly cost in the first month.

Traps

1. Bad-sounding Bulgarian voice. Until mid-2025, synthesized Bulgarian voices sounded clearly robotic. Now there are several solutions that pass for human in normal conversation, but not every provider uses them. Hear a demo before signing.
2. Mid-conversation handoff. AI must pass the full context to reception (name, purpose, what was said so far), not transfer with "hello, what is your problem." This kills the convenience.
3. GDPR. Recording calls — requires an explicit announcement at the start and patient consent. Without it, you are inviting a CPDP complaint.
4. Slight error rates. AI will occasionally book the wrong time or confuse two names. Reception must review new bookings daily, at least for the first month.

Frequently asked questions

Won't patients give up if they realize they're talking to AI?

Our experience: no. Patients want a fast answer and a booking at a time that works for them. If the AI assistant does that job without issues, they're not bothered by the fact that it's a machine. What kills convenience is bad sound quality or incorrect booking — not the AI itself.

Does it work with older patients?

Depends on the channel. Phone calls — yes, older patients often prefer phone over chat. Viber works surprisingly well with the 55+ generation. Expect 5-10% of patients to insist on a human — leave them that option without closing the line.

What happens if internet or power fails?

AI infrastructure is in the cloud — if you have internet, it works. If you don't, the phone forwards to a mobile reception number automatically. This is one of the settings configured at implementation, not left for later.

How many hours of training does staff need?

About 4 hours split into 2 sessions. The first covers what AI handles and what it doesn't; the second covers reviewing new bookings and passing context back to AI for complex cases. Small compared to onboarding a new reception hire.

Have a practice and want a concrete estimate for your case? Describe it in the chat on the home page: number of rooms, daily visit volume, current calendar system. In a minute you get an initial plan.