← All case studies
Healthcare clinicVoice AI

Answering every call at a four-site physiotherapy group

Meadowbank was missing a third of inbound calls. A voice agent now books, reschedules and triages, with clinicians reviewing anything clinical.

0
Missed calls
−22%
No-show rate
19 hrs
Reception time saved weekly
Problem

Two receptionists, four sites, and a phone that never stopped

Meadowbank ran reception from a single desk covering all four clinics. During morning peaks roughly a third of calls went unanswered, and voicemails were returned late in the day when most callers had already booked elsewhere.

Approach

A voice agent with a narrow, well-defined job

We deliberately scoped the agent to four tasks: booking, rescheduling, cancellation and directions. Anything touching symptoms, medication or insurance goes to a human, and the agent says so plainly. It reads live availability from the practice management system and writes the booking back with a transcript attached.

Safety

Clinical boundaries enforced in code, not prompts

Escalation is a hard rule outside the model: a keyword and intent classifier routes any clinical content to the on-call coordinator before the agent can respond. Every call is transcribed, retained for 90 days and reviewable by the practice manager. We ran two weeks of shadow mode, where the agent drafted responses that staff approved, before it took a single live call.

Result

Reception moved from the phone to the front desk

Answer rate went to 100% within the first week. The no-show rate fell 22% once the agent began confirming appointments the evening before, and reception staff now spend their morning with patients in the building rather than on hold music.

Before / after
Calls answered
68%100%
Average wait
3m 40s4s
After-hours bookings
031/week
No-show rate
14%11%

Patients tell us the phone situation improved. Almost none of them have noticed why.

MPPractice Manager, Meadowbank Physiotherapy