Next Gen System
Case 02 Boston HEALTHCARE · VOICE AI
Voice AI Appointment Scheduling Assistant
Production voice agent that books, reschedules, and hands off cleanly—with humans in the loop.
01 — Context
The Client’s Story
Front-desk lines stacked every morning. Callers waited on hold to book or move appointments while staff juggled walk-ins, insurance questions, and chart prep at the same desk.
After-hours coverage was thin. Missed calls turned into portal messages that sat overnight, and patients often called again the next day—doubling the load.
Leadership wanted containment on routine booking work without losing a clear human path for clinical or policy questions. Trust was fragile after a prior IVR rollout that felt robotic and dropped calls.
02 — Ambition
Objectives
Handle booking, reschedule, and cancel intents by phone with confirmation SMS, using the live calendar—not a shadow spreadsheet that drifts from the EHR.
Escalate to humans with a full transcript the moment policy or clinical questions appear, so staff inherit context instead of restarting.
Prove quality with evals on recognition, task completion, and handoff rate before any clinic-wide expansion.
03 — Friction
Challenges
Live calls demand low latency and clean barge-in. A slow or talk-over agent feels broken even when the booking logic is correct.
PHI controls and audit logging had to satisfy compliance review—every AI turn and escalation needed a durable trail.
Staff distrust after the earlier IVR failure meant change management was part of the delivery, not an afterthought.
04 — Approach
Solution
We built a Voice AI assistant with speech dialogue, calendar tools, and hard stop conditions. Complex or clinical intents escalate immediately with transcript and intent tags.
Confirmation SMS closes the loop for successful bookings. Evals cover recognition quality, task completion, and handoff rate before wider rollout across clinics.
Runbooks and listen-ins gave front-desk leads a way to spot failure modes early. The same gate that blocks unsafe automation also protects the brand when the model is unsure.
05 — Outcomes
Results
Booking call types cleared faster, after-hours coverage improved without new headcount, and every AI-handled call left an audit trail ops could defend.
- Material drop in hold time on booking call types.
- Higher after-hours coverage without extra headcount.
- Clean audit trail for every AI-handled and escalated call.
06 — Forward
What’s Next
Expand to prescription refill status with the same human-approval gate, so status lookups stay automated while clinical decisions stay with licensed staff.
Ready to ship something similar?
We’ll map scope, risks, and a first production slice in one working session.
