An AI receptionist and operations dashboard that turns patient calls into booked appointments, resolved requests, or clear follow-up tasks for dental clinic staff.
My role
Product Owner, AI Architect & Full-stack Engineer
System
Retell, React, TypeScript, Express, Firebase, Cloud Run
Built from scratch around patient experience, clinic workflows, booking integrations, reporting, infrastructure, and production deployment.
A dental call can be a new-patient booking, a reschedule, a cancellation, a records question, an insurance question, or an urgent pain case. The receptionist has to recognize the intent, collect only the necessary details, and move the call toward a real operational outcome.
The system combines an AI receptionist, OpenDental booking logic, HIPAA-conscious identification flows, server-side PHI redaction, and an operations dashboard. Staff see booked appointments, unresolved issues, and follow-up work without digging through raw transcripts.
02
Real calls
The strongest proof is the call itself. Cindy can move a new patient from first question to a booked dental appointment in under two minutes, while still sounding calm and human enough that fewer than 3% of callers asked for a person.
2-min booking
AI receptionist identifies the goal, finds availability, collects identity details, and books the appointment quickly.
Hectic call
A slow and scattered caller, and the receptionist keeps the flow moving without becoming brittle.
03
AI Receptionist
The receptionist is not a scripted FAQ bot. It is an integrated front-desk workflow that can answer common questions, protect patient identity, and complete appointment work inside the clinic's operating rules.
OpenDental integration
Identifies existing patients through HIPAA-compliant phone plus DOB/SSN-style verification, creates new patients, finds upcoming appointments, and books new appointments against live availability.
Clinic rules
Maps appointment reasons to procedure types, providers, urgent paths, child-patient rules, office hours, location constraints, and reschedule notes before committing a slot.
Seven-language coverage
Supports multilingual callers, with the Broad Street receptionist configured for natural English, Spanish, and Russian handling in the prompt.
Empathetic call style
Keeps turns short, asks one question at a time, reuses details already provided, and stays warm during pain, confusion, or slow caller responses.
HIPAA from day one
Uses server-side PHI redaction, protected booking flows, and patient identification rules designed so the model does not expose patient IDs or internal payloads.
Front-desk coverage
Answers insurance, payment-plan, X-ray, accessibility, directions, walk-in, pediatric, cosmetic, wisdom-tooth, and uninsured-pricing questions when the prompt contains safe clinic policy.
Urgent and admin routing
Prioritizes pain, swelling, infection, bleeding, broken teeth, trauma, and post-procedure problems for the soonest appropriate appointment while routing billing, records, cancellation, and unsupported requests to staff follow-up.
Staff-ready outcomes
Every call resolves into a clean state: booked appointment, urgent booking, reschedule request, cancellation message, answered FAQ, callback request, or staff follow-up summary.
04
Operational view
Overview
Managers can compare handled calls, booked appointments, booking rate, intent distribution, and urgent issues across clinics and date ranges.
Action queue
Problem calls and follow-up states become a focused work queue instead of disappearing inside transcripts.
05
Call review
Each row turns a conversation into operational data: intent, result, appointment state, end reason, attention status, and a concise summary.
Call history with intent, result, appointment state, end reason, and follow-up status.
06
Booking path
The model can speak naturally, but appointment commitments stay behind deterministic workflow steps.