How a 22-Site Clinic Network Cut Missed Appointments by 38%
4.8× faster patient call routing
About this case study: results are from a real Salamtalk deployment. The customer’s name is a pseudonym and their logo is withheld under a confidentiality agreement. The metrics below are reported as measured.
The situation
MedCore Clinics runs 22 outpatient locations, each with its own front desk handling appointment bookings, prescription-refill requests, and general enquiries. Call volume across the network exceeds 15,000 patient calls a month, concentrated into two peaks: first thing in the morning and the hour after lunch.
Front desk staff were doing two jobs at once: greeting patients in person and answering the phone. The result was predictable.
- Long hold times at peak hours. Patients calling during the morning rush regularly queued behind in-person check-ins.
- Missed appointments were rising. Patients who couldn’t get through to reschedule or confirm simply didn’t show up.
- After-hours calls went to voicemail. Refill requests and routine queries left outside clinic hours sat until the next business day.
What they needed
MedCore was explicit from the outset about what an automated system would and would not be allowed to do. Any solution had to route appointment scheduling and prescription-refill requests automatically, but it could not be permitted to assess, advise on, or make decisions about clinical or medical matters. Every call needed an immediate, obvious path to a human, and any after-hours call that sounded urgent had to be escalated rather than queued.
Alignment with HIPAA handling of patient information was a baseline requirement, not a negotiable feature.
What we deployed
| Requirement | Implementation |
|---|---|
| Call intake | Salamtalk AI Receptionist answering first, across all 22 sites |
| Appointment booking | Automated scheduling and rescheduling against each clinic’s calendar |
| Prescription refills | Structured intake and routing to the correct clinic’s pharmacy liaison |
| Clinical safety | No symptom assessment or medical advice; immediate transfer on request or on urgency keywords |
| After-hours handling | Triage handoff to on-call staff for anything flagged as urgent |
| Data handling | HIPAA-aligned handling of patient information, encrypted in transit and at rest |
The AI Receptionist was configured narrowly. It books, reschedules, cancels, and routes refill requests. It does not ask about symptoms, does not offer guidance on medication, and does not attempt to triage clinical severity. Any caller who says they want to speak to a person, or whose call is flagged for a possible emergency, is moved to a human immediately — at any site, at any hour.
How the deployment ran
- Mapped each clinic’s call flows, booking rules, and refill-request process separately, since scheduling policy varied by location.
- Built the escalation rules first, before any booking logic, so the handoff-to-human path was tested and confirmed before go-live.
- Piloted at three clinics for two weeks, with front desk staff reviewing a sample of calls each day.
- Rolled out to the remaining 19 sites in three batches, a week apart.
- Tuned after-hours escalation thresholds once real call patterns were visible, since the initial settings over-escalated to on-call staff.
Front desk staff kept a visible override at every site: any call could be pulled to a human at any point, and staff were trained to do so without hesitation if a patient sounded distressed or the request fell outside booking and refills.
The results
- 4.8× faster patient call routing, from first ring to the right destination
- 38% fewer missed appointments, attributed to easier rescheduling and confirmation calls
- 15,000+ monthly patient calls handled across the network without added front desk headcount
- After-hours refill and booking requests now captured immediately rather than queued to the next business day
The missed-appointment reduction was the clearest clinical and operational win. Most of it came from patients being able to reschedule in a single call instead of hanging up during a queue and not calling back.
What was harder than expected
Getting the escalation threshold right took longer than the booking logic. Set too conservatively, nearly every after-hours call got escalated to on-call staff, defeating the purpose. Set too loosely, there was a real risk of a caller with an urgent need being left in the automated flow too long. MedCore’s clinical leads reviewed escalation logs weekly for the first month and adjusted keyword and context rules twice before settling on a version they were confident in.
Staff trust had to be earned, not assumed. Front desk teams were initially cautious about letting a system handle scheduling for their own patients, and a few sites kept manually re-confirming AI-booked appointments for the first two weeks. That extra checking dropped off once staff saw the booking accuracy hold up, but it meant the measured efficiency gains understated the first month’s results.
If MedCore ran this again, clinical leadership says they would involve front desk staff in defining the escalation keywords from day one, rather than after the pilot.
Next steps: see AI Receptionist for the booking and routing capability, security for how patient data is protected, or talk to our team about a healthcare rollout plan.