AI receptionist for a dentist office. The missed-call math.

The hygienist is mid-cleaning. The front desk is on the other line. The phone rings, rolls to voicemail, and the new patient hangs up. That call was worth $1,800 in lifetime value to the practice. You'll never know it happened.

This is a field note on what an AI receptionist for a dentist office actually has to do, booking, emergency triage, HIPAA, Dentrix integration, the boring stuff that decides whether the rig is real or a demo. The framing is dentist-specific. The underlying rig is the same one in our after-hours answering service spec, tuned to a dental protocol.

§ 01The dental missed-call leak.

Three patterns show up in every dental practice we audit. None of them are surprising once you look at the call log. All of them are expensive.

• The lunch hour. 12:00–1:30pm. Front desk is at lunch, hygienists are in chairs, the phone goes to voicemail. About 8% of weekly inbound calls land in this window. Most are existing patients confirming or rescheduling, high-intent, low-friction, easy money.
• The 5pm cliff. Practice closes at 5pm. Inbound calls keep coming until 9 or 10pm, partners calling on the drive home, parents calling after dinner, the patient who just chipped a tooth on a pretzel. Of these, fewer than 6% get a callback before 9am the next morning.
• The Saturday gap. Most general dental practices are closed Saturday. Emergency calls still happen. The patient calls three practices in a row, lands a callback from whichever office answers first. If you don't, your competitor does.

§ 02What the leak costs a working practice.

Take the conservative numbers. We're going to under-count on purpose, the real loss is bigger.

Assumptions, written down

• 180 inbound calls per week (a mid-sized general practice)
• 20% land outside front-desk hours = ~36 missed calls / week
• Of those, ~30% are new-patient inquiries or active patients with a real ask
• New-patient lifetime value: $1,200–$2,400 over 3 years (ADA benchmark for general dentistry)
• Industry close rate on cold inbound dental inquiries: ~25%

That puts the floor of the leak at roughly two to three new patients lost per week, or $10,000–$25,000 per month in deferred lifetime value for a single under-capacity practice. The rig that plugs this is $2,000, fixed, once. The math isn't close.

The voicemail box is the most expensive employee in a dental practice. It works 24/7, never asks for a raise, and loses you a new patient every other day. From a practice-manager interview, April 2026

§ 03The on-call protocol an AI receptionist actually needs.

Most "AI receptionist for small business" demos answer the phone in a friendly voice and then either book or take a message. That's not enough for a dental practice. The protocol has four branches.

Branch 1 · Routine booking

New patient wants a hygiene appointment. Existing patient wants to reschedule. The rig reads the open-slot calendar from Dentrix, Eaglesoft, or Open Dental, offers two slots inside the provider's rules (new patient blocks, hygiene vs. operative chairs), and writes the appointment back. The front desk sees it Monday morning as a normal booking.

Branch 2 · Emergency triage

The rig listens for emergency language, knocked-out tooth, abscess, swelling, severe pain, post-extraction bleeding) and routes to the on-call dentist's phone in seconds with the patient's name, callback number, and a 30-second summary. The patient gets a holding response with first-aid instructions (rinse with warm salt water, keep the tooth in milk, ibuprofen if no contraindications). The dentist gets the page.

Branch 3 · Prescription / clinical question

"Can I take ibuprofen with this antibiotic?" "Is it normal for the gum to be swollen three days post-extraction?" The rig does not answer clinical questions. It captures the ask, flags it as clinical, and routes to the on-call dentist with a transcript. The patient gets a holding response that promises a callback inside 30 minutes.

Branch 4 · Logistics

Hours, address, insurance accepted, parking, whether you treat kids, whether you do Invisalign. The rig answers from a practice-specific knowledge base. No escalation needed. Most calls land here, about 55% of after-hours volume in the practices we audited.

§ 04The rig, in five pieces.

The minimum viable AI receptionist for a dentist office is five pieces. We've shipped this pattern. We know what's load-bearing.

1. The front number. Either the practice's existing main line (forwarded after-hours) or a dedicated AI receptionist line. Twilio with a HIPAA BAA.
2. The voice + SMS stack. Voice for callers who want to talk, SMS auto-response within 60 seconds for the missed-call leak. Same protocol on both channels.
3. The practice-management integration. Read open slots, write appointments. Dentrix, Eaglesoft, or Open Dental, via API or a thin server-side adapter where the vendor doesn't expose a clean API.
4. The on-call escalation. A direct ring to the on-call dentist's phone for triaged emergencies. Pushover or a regular SMS with a tap-to-call link.
5. The morning runbook. A 7am email to the practice manager summarizing every call the rig handled overnight: bookings, triage escalations, logistics questions, anything that needs human follow-up.

// rough shape of the dental on-call webhook
on inbound.call →
  classify(intent) →
    EMERGENCY  → page dentist + first-aid holding SMS
    BOOKING    → read calendar, offer 2 slots, write appt
    CLINICAL   → capture ask, route to dentist, hold SMS
    LOGISTICS  → answer from KB, no escalation
  log to morning_runbook (Sheet)

§ 05Dentrix, Eaglesoft, Open Dental.

Most general practices in the US run on one of three systems. Each one has its own quirks for an AI receptionist integration.

• Dentrix. Henry Schein. Most common. No clean public API; we use the Dentrix Developer Program (DDP) and a server-side adapter that reads the local database. Appointment booking lands in under a minute.
• Eaglesoft. Patterson Dental. Similar pattern, adapter against the local PostgreSQL. The provider-rule logic (which chairs are hygiene, which are operative) is more tedious to model.
• Open Dental. Open-source friendly, has a real REST API. Cheapest and cleanest integration of the three. If the practice is on Open Dental, the rig ships about three days faster.

§ 06HIPAA, in one paragraph.

An AI receptionist for a dentist office is handling protected health information the moment it captures a patient's name and reason for calling. That makes HIPAA non-optional. The rig ships on HIPAA-eligible infrastructure (AWS or Azure regions that support healthcare workloads), routes SMS through a carrier with a signed BAA (Twilio with the healthcare-tier contract), encrypts transcripts at rest, and ships with an access log and a documented breach-response runbook. The BAA paperwork and the compliance work are included in the $2,000 fixed fee. We don't sell a non-HIPAA version.

§ 07Takeaway, in one line.

The leak is a settings problem, not a strategy problem. An AI receptionist for a dentist office that handles routine booking, emergency triage, clinical-question routing, and logistics is two weeks of work. The math pays back inside the first month on any working practice. The longest part of the project is the practice owner deciding to do it.

---

If you run a practice and you've read this far, the next step is the same as everywhere else on the site. Fifteen minutes. One question: where is the phone going quiet?