Case study · Healthcare · 12-clinic GP network · 90 days
How a 12-clinic GP network grew patient bookings 38% in 90 days.
A multi-site general practice was losing patients at the front desk — 240 inbound calls a day, only 60% answered in time. A voice agent picked up the rest. Three months later, the books were full.
Chapter 01 · The problem
Two hundred and forty calls a day they couldn’t answer in time.
When a GP network grows past five sites, the phones become the business. Front-desk staff at twelve clinics were fielding around 240 inbound calls a day. Roughly 60% were answered inside two rings. The rest dropped into voicemail or disconnected — and many of those were new patients trying to book a first appointment. The clinic group was paying to acquire those calls through Google Ads. Each missed one was paid-for foot traffic walking out the door. The Practice Manager had been quietly tracking it in a spreadsheet for nine months.
240 inbound calls a day
Chapter 02 · The approach
We didn’t replace the receptionist. We gave them a co-pilot.
The fix was not a louder phone system. It was a voice agent that picked up calls only after two rings — the moment a human front-desk worker would otherwise drop one. The agent identified the caller, pulled their record from the practice management system, offered the next three available appointments, and handed the call back to a human if anything looked unusual. Boring on purpose. Useful by design.
What we built
A Twilio + Vapi voice agent backed by Anthropic Claude for intent recognition, integrated with the clinic’s practice management system through n8n. The Hermes orchestrator coordinates the voice agent, the booking workflow, and the post-call SMS confirmation.
Build time · 18 days · pilot phase · followed by 8 weeks of iteration with the Practice Manager.
Chapter 03 · The outcome
Three numbers that didn’t move on their own.
- +0%
bookings recovered in 90 days — measured against the baseline missed-call rate from the prior quarter.
- $1.2M
AUD in patient lifetime value reclaimed — based on the clinic group’s own LTV model (12-month horizon).
- 0%
of after-hours calls answered by the agent — without waking a duty receptionist or routing to voicemail.
Numbers verified by the Practice Manager. Anonymised under our standard case-study disclosure: vertical and outcome are real; name, location, and PMS-specific identifiers are not.
Chapter 04 · What we learned
The voice agent was easy. The handoff was the hard part.
The voice agent itself took 18 days to build. The eight weeks that followed were spent tuning when it should hand the call back to a human — too eager to defer and the front desk gets flooded; too confident and the agent books the wrong slot for a complex patient. Most of the value came from a single threshold setting we revisited four times.
The lesson was less about the AI and more about the operating rhythm: we ran a 30-minute weekly review with the Practice Manager, watched ten transcripts together, and shipped a small change every Friday. Eight cycles in, the handoff rate had settled at 7%. The agent was handling the rest in full.
Settled handoff rate · 7%
Chapter 05 · In the client’s words
Raava helped us recover capacity we didn’t know we had. The agent isn’t replacing anyone — it’s catching what we couldn’t catch on the busiest mornings.
Practice Manager · 12-clinic GP network · NSW (anonymised at the client’s request)
Curious whether your clinic has the same gap?
A free 45-minute audit. We look at your call volume, your practice management system, and where time is leaking. You leave with a one-page memo, whether we’d be a fit or not.