Cadence
Get started
All posts
April 30, 2026 · 8 min read · ai-receptionist · open-dental · integrations · operations

Open Dental + AI receptionist: the practical setup guide

30 minutes of work to wire an AI receptionist into Open Dental — bidirectional sync, recall cadence, escalation playbook, and the three pitfalls most vendors won't warn you about.

By Cadence team

Open Dental is the source of truth for ~30,000 US dental practices. If you're running it, the question isn't whether to add an AI receptionist — it's how to wire one up without breaking the workflow your front desk already trusts.

Here's the practical guide. About 30 minutes of work end to end, assuming you have admin access to OD and your phone provider.

What "wired up" actually means

A real Open Dental + AI receptionist integration moves data both directions:

  • Out of OD: appointment changes, new patients, provider schedule updates flow into the AI so it knows what's bookable in real time.
  • Into OD: bookings made by the AI write back to OD as proper appointment rows. Your hygienist sees them in the same OD calendar she uses today. No second tool.

If a vendor pitches you "AI receptionist with Open Dental integration" and the integration is just an email digest you copy-paste, walk away. That's not integration — that's an extra step.

Step 1: enable the OD API

Open Dental ships an HTTP API that's off by default. To turn it on:

  • OD > Setup > Advanced Setup > FHIR / Web Service
  • Enable the FHIR server (or the legacy "Service Center" API depending on your version)
  • Generate a customer key and a developer key for the AI vendor. Treat them like passwords — full read/write access.
  • Note the endpoint URL — usually internal IP + port 80 / 8080.

If your OD is hosted on a server in your office (not cloud), the AI vendor will need either a port-forward or a tunnel (ngrok, Cloudflare Tunnel) to reach it. Don't expose port 80 publicly without auth — use the tunnel.

Step 2: tell the AI about your service menu

The AI doesn't guess. It needs to know your appointment types, durations, and which provider does what. With Cadence, you tell Maya in plain language:

  • "Cleaning is 60 minutes, hygienist only."
  • "New patient exam is 90 minutes with a doctor and ends with x-rays — schedule those at 8:30, 10, 1:30, or 3."
  • "Crown prep is 90 minutes, doctor only, no Mondays (lab pickup day)."

Maya parses these into bookable slot rules and writes them back to OD's definitions. Five minutes of conversation, not an afternoon clicking through a setup wizard.

Step 3: forward the phone

The AI gets its own phone number from the vendor. You forward your existing practice number to it (or port the number over if you want — usually a 2-day process with your carrier).

Forwarding lets you A/B-test the AI without commitment: send 50% of calls to the AI, 50% to your existing front desk for a week, look at the conversion rates side by side. If the AI wins, port the number; if not, you walk away with no churn.

Step 4: configure the recall + reminder cadence

Most practices already run some recall in OD (the "recall list" tool). Decide what the AI owns vs. what stays in OD:

  • AI owns: 6-month recall reminders, 24h + 2h appointment confirmations, no-show waitlist auto-fill, dormant-patient reactivation at 18 months.
  • OD continues to own: the recall list itself (the source of truth for who's due), insurance verification, treatment plans.

Don't double-send. If OD's reminder is firing AND the AI is firing, your patients get two SMSes and one will be ignored. Pick one or the other for each cadence step.

Step 5: train your front desk on escalations

The AI handles ~70-85% of calls without human intervention. The rest — complex insurance questions, complaints, emergency triage — get warm-transferred to your front desk with a full transcript already in the chat tool. Two things to set up:

  • The escalation phone number: usually the owner's cell or a dedicated front-desk line.
  • The escalation playbook: teach the front desk that they're inheriting context, not re-doing the call. The AI summary tells them what the patient already said.

Step 6: monitor the first week

Watch the OD appointment book the first week. Check that:

  • AI bookings appear in OD within seconds of being made.
  • OD reschedules made by your front desk sync back so the AI knows the slot is gone.
  • Confirmation status (the "Confirmed" column) updates when patients reply C/Y to the SMS.
  • The AI isn't double-booking — early on this can happen if your slot definitions are loose.

Cadence ships a sync-health dashboard that flags conflicts (both sides edited the same appointment within minutes). It's rare but worth knowing about.

Pitfalls

  • OD running on a Windows desktop in your office: the AI vendor needs network access to that machine. Either port-forward (with auth) or use a tunnel. Cloud-hosted OD installs are easier — just give the vendor the API endpoint.
  • Custom procedure codes: if your practice uses non-standard CDT or local codes, tell the AI vendor before go-live. Otherwise the AI books generic "cleaning" slots and your billing has to clean up the codes after the fact.
  • Multi-location: if you run 2+ locations on the same OD database, make sure the AI books to the right provider AND location. Most vendors get the provider right; location is the second-most-common bug.

Bottom line

OD + AI receptionist works. The setup is real but bounded — 30 minutes of work, less than half a day of monitoring after go-live. The ROI shows up in the missed-call rate (down to zero), the no-show rate (down 60% with proper reminders), and the recall conversion (up 25-40 points).

If you want to skip the integration build and ship the cadenceagents.ai end-to-end version, Cadence is the AI front desk built for OD specifically — two-way sync, conflict resolution, sync-health dashboard, all in. Five-minute setup; Maya walks you through it in chat.