After-Hours Voice Receptionist for a Local Clinic

This is a representative case study based on the kind of engagement we run for local clinics. Specifics are anonymized; the architecture, approach, and outcome shape are real.

The Problem

A multi-provider clinic in the Flathead Valley was sending every after-hours call to voicemail. Most callers didn't leave one. The team had no idea how many of those callers were urgent vs. routine, how many were existing patients vs. new, or how many they were losing to the next clinic on the search results.

What they did know: the front desk was spending the first 90 minutes of every morning returning voicemails, many of which were just "what time do you open?"

The Approach

We ran this as a five-week Automation Sprint. The plan:

• Week 1: Sit with the front desk team. Review a week of after-hours voicemails. Identify the top five call types and what a "good" response looks like for each.
• Weeks 2–3: Build the voice agent. Anchor it to the clinic's actual hours, services, and provider list. Wire it into the existing scheduling system (Google Calendar through their EHR).
• Week 4: Test extensively. Real callers, real edge cases, real bad audio. Tune the script.
• Week 5: Launch overnight. Front desk has a dashboard showing every call, the AI's response, and any flagged items.

What It Does

• Answers after-hours calls in a friendly, identifiable voice that explicitly tells the caller they're speaking with the clinic's automated assistant.
• Handles routine questions — hours, location, services offered, insurance accepted, basic provider info.
• Books appointments for new and existing patients into the next available slot, then sends an SMS confirmation.
• Escalates urgency — if a caller indicates anything urgent (specific symptom keywords, the word "emergency"), the agent immediately texts the on-call provider and tells the caller to call 911 if life-threatening.
• Logs everything so the front desk can review and follow up the next morning if needed.

Outcome

Six weeks after launch:

• After-hours voicemail volume down roughly 70% — most callers got their question answered.
• Estimated ~3 additional booked appointments per week that would otherwise have been a voicemail or a hangup.
• Front desk morning recovery time down from ~90 minutes to ~25 minutes.
• Zero customer complaints about "talking to a robot." (We were braced for some. They didn't come.)

What We Handed Over

• The running production system on the clinic's accounts (Vapi, Twilio, Railway).
• A written runbook covering every scenario the front desk needs to handle: changing hours, adding/removing providers, updating the FAQ, viewing call logs, escalating issues.
• Two hands-on training sessions with the office manager and a backup.
• Two weeks of post-launch support (included in the sprint).

The clinic now owns and operates the system. We get a check-in every quarter to look at the call patterns and tune the script. That's it.

Tech Used

• Claude for the conversational logic and intent recognition.
• Vapi for the voice agent infrastructure.
• Twilio for the phone number and call routing.
• Google Calendar API for appointment booking.
• Postgres on Neon for the call log and audit trail.
• Railway for the management dashboard.