DEV Community

VoiceFleet
VoiceFleet

Posted on • Originally published at voicefleet.ai

Building an AI Follow-Up System for Missed Dental Appointments

#ai #voiceai #healthtech #automation

Building an AI Follow-Up System for Missed Dental Appointments

Here's an interesting problem I've been working on: dental practices lose thousands per year to no-shows, and the follow-up process is almost always manual and delayed.

The Technical Challenge

When a patient misses an appointment, three things need to happen fast:

  1. Detection — the PMS (practice management system) flags the no-show
  2. Outreach — an automated call goes out within 15 minutes
  3. Resolution — the patient rebooks, cancels, or gets escalated to staff

The timing matters enormously. Follow-up within 15 minutes recovers ~40% of appointments. After 24 hours? Under 10%.

Architecture Overview 

PMS Webhook → Event Queue → AI Voice Agent → Call Patient
                                    ↓
                            Outcome Logger → PMS Update
                                    ↓
                            Staff Dashboard (escalations only)
Enter fullscreen mode Exit fullscreen mode

The voice agent needs to:

  • Sound natural (not robocall-y)
  • Handle interruptions and questions
  • Access available appointment slots in real-time
  • Comply with GDPR (consent, opt-out, data minimisation)

Voice AI Considerations

Modern voice AI stacks (like what we built at VoiceFleet) use:

  • Low-latency TTS for natural conversation flow
  • Intent recognition tuned for healthcare contexts ("I was sick", "I forgot", "Can I come tomorrow?")
  • Slot-filling from live calendar availability
  • Graceful handoff to human staff when the AI hits its limits

The biggest engineering challenge is latency. Patients expect conversational pacing — anything over 800ms response time feels wrong.

Results

In early deployments with Irish dental practices:

  • 30-35% of no-shows successfully rebooked via automated follow-up
  • Average follow-up time: 12 minutes post-missed-appointment
  • Staff time saved: ~45 minutes/day on follow-up calls

Open Questions

Still figuring out:

  • Optimal number of follow-up attempts before giving up
  • Whether SMS follow-up after a missed call improves recovery
  • Best time-of-day patterns for different patient demographics

Would love to hear from anyone else building in the healthcare voice AI space. What are you seeing in terms of patient acceptance of AI callers?

Building at VoiceFleet.ai — AI receptionists for healthcare and hospitality.

Top comments (0)