I'm building Zirco.ai — an AI administrative employee for dental practices.
The problem: dental front desks spend 2–3 hours every day manually verifying insurance benefits through carrier portals. On top of that, they're handling inbound scheduling calls, sending reminders, coordinating referrals, and managing new patient intake — all manually, all repetitive, all expensive. A single front desk employee costs $40–50K/year and turns over at 40% annually.
What I built: an AI that handles the full front desk workflow end-to-end. It answers inbound calls with a voice AI (built on Vapi), verifies insurance automatically by integrating with carrier APIs and using browser automation for portals that don't have them, books appointments directly into the practice's existing software (Dentrix, Open Dental, Eaglesoft), and sends personalized SMS/email reminders. Everything the front desk does, handled autonomously.
The technical parts I found most interesting:
- Insurance verification requires logging into 10+ different carrier portals with different data formats and session handling. Playwright handles the ones without APIs. Availity covers most of the rest.
- Voice AI for healthcare needs to handle interruptions, insurance jargon, and multi-step scheduling logic simultaneously. Getting this reliable enough for a medical practice took significant prompt engineering and fallback design.
- Multi-tenant HIPAA compliance from day one meant thinking carefully about row-level security, audit logging, and BAA agreements with every vendor that touches PHI.
Stack: Python + FastAPI, Next.js, PostgreSQL + pgvector, Vapi.ai for voice, Claude for reasoning, Playwright for portal automation, Twilio for SMS, deployed on AWS with HIPAA-eligible services.
Currently in beta. Have spoken to 30+ dental practices during discovery. Happy to discuss the technical architecture, the insurance verification approach, or anything else.