AI voice agent cut appointment no-shows by 38%

Medicare Hospital, a 220-bed multi-specialty hospital in Chittagong, was losing patients and frustrating staff due to an overwhelmed front-desk phone system. Their 4 receptionists handled roughly 4,200 inbound calls per month — appointment booking, rescheduling, doctor availability enquiries, report-ready confirmations, and emergency triage. During peak hours (9 AM–1 PM), 31% of calls went unanswered, and patients would either call back multiple times or simply go to a competing hospital. The manual ledger-based booking system also meant double-bookings happened 8–10 times a week, and no-show rates were sitting at 22% because there was no systematic reminder system. Elderly patients who only spoke Bangla struggled with the English-only IVR menu, and after-hours calls (5 PM–9 AM) were completely unhandled. Hospital management estimated they were losing 30–40 OPD patients per week to competitors purely because of phone accessibility issues, and the receptionists were reporting burnout at an unsustainable rate. The hospital's existing IT vendor had quoted ৳18 lakh for an on-premise PBX upgrade with a 6-month implementation timeline — money the management was reluctant to spend on infrastructure that still wouldn't solve the human-staffing gap. They needed a solution that could handle call volume 24/7, in both Bangla and English, without requiring additional headcount, and without disrupting the existing HMIS that staff were already trained on.

We deployed a bilingual (Bangla/English) AI voice agent built on a telephony-integrated LLM stack that now answers every inbound call 24/7 within 2 rings. The agent handles appointment booking by checking doctor availability in real-time against the hospital's HMIS, confirms slots verbally with the patient, and writes the booking directly into the system. It also handles rescheduling, cancellations, and answers FAQs about visiting hours, doctor OPD schedules, report collection, and billing counter locations. For after-hours calls, the agent triages emergencies — if keywords like "emergency", "accident", or "chest pain" are detected, it immediately forwards to the on-call doctor's mobile. We layered on an automated WhatsApp + SMS reminder system that sends confirmations at booking time, reminders 24 hours before, and a final reminder 2 hours before the appointment. Patients can confirm or cancel via a simple reply, and the AI updates the HMIS automatically. A full call transcript and analytics dashboard lets hospital management monitor volume, common questions, and agent performance. The entire integration was completed in 4 weeks without any changes to the existing HMIS database schema — the AI agent talks to the HMIS through a thin REST API wrapper that the hospital's IT team reviewed and approved. A fallback "press 0 to speak to a receptionist" option was preserved during business hours to handle edge cases and build patient trust during the rollout phase.