AI agents for dental clinics: recall, insurance prep and follow-up

The unscheduled-treatment problem

Most practices hold a list of patients who were told they need a crown, a filling or a hygiene visit and never booked. Another list holds patients who have not been seen in 18 months. Both lists are worth money and goodwill, and both get worked only when the front desk has a quiet afternoon, which is rarely.

A follow-up agent reads those lists from your practice software, drafts a polite and specific message for each patient, sends it through the channel the patient prefers once a staff member approves the template, and offers real appointment slots. Replies that need a person, such as a question about pain or cost, go straight to the desk.

Jobs for the agent, and jobs it must not do

Anything that sounds like a symptom is routed to a human within the working day, and urgent keywords such as swelling or severe pain trigger an alert to the on-call contact.

Patient privacy is the design brief

In the US, a practice is a covered entity under HIPAA, so any vendor handling patient information on your behalf needs a business associate agreement, and the model provider's terms must allow it. The HHS website explains the rules. In the UK, patient data is special category data under UK GDPR, the ICO publishes guidance, and the General Dental Council sets standards for professionals. Australian practices work under the Privacy Act and AHPRA advertising rules, and Irish practices under GDPR.

Our build choices follow from that: the agent sees the minimum fields needed for each task, records stay in your own systems, nothing is used to train a model, and the log shows every record touched. Review requests also need care. Rules on incentivizing reviews and on how healthcare services may be advertised vary by country, so check with your professional body before wording them.

What connects to what

US practices commonly run Dentrix, Eaglesoft or Open Dental. UK practices often run systems such as SOE or Software of Excellence products, and Australian practices use tools like Best Practice or Core Practice. Integration options vary by product and by your plan, and some vendors charge for API access. We check this before quoting, because it is the item most likely to change the price.

When no direct integration exists, a safe fallback is an export-based flow: the agent reads a scheduled report you drop into a protected folder and writes back to a task list. It is slower than a live connection but avoids brittle screen scraping.

Phone calls are a separate channel, covered by our page on AI voice agents for dentists. This page covers messaging, forms and back-office tasks.

Timeline and price

1. Pick the listUsually overdue recalls. We review the wording your team already uses for reminders.
2. Build on test dataNo live patient records are used until the privacy agreement is in place.
3. Supervised sendStaff approve a batch each morning. We measure replies and edits.
4. ExtendAdd treatment-plan follow-up, then insurance prep.

Projects start at $1,500 and our rate is $30 an hour, with a fixed quote within one working day of a short call. A single-list recall agent sits near the small end; adding insurance verification and several software connections raises hours. Compare scope with our general guide, AI agent development, and see what an AI agent costs.