Case study snapshot
- Client: Beth Psychiatry, a mental health practice
- Service: AI voice receptionist for new patient intake, appointment scheduling and prescription refill requests
- Stack: voice agent, n8n, Python, Tebra API, Google Apps Script, SMS
- Headline results: 200+ patient interactions handled every week, 60% less manual scheduling work, 300+ refill requests captured every month
The problem
Beth Psychiatry's front desk was fielding the same three kinds of calls all day: new patient questions about insurance and first visits, requests to book, move or cancel appointments, and prescription refill requests. The same staff also greet patients, handle insurance filing and complete paperwork, so calls went to voicemail and patients had to wait for a call back.
Phone calls carry extra weight in a psychiatry practice. A first call can be hard for a patient to make, and a refill that is not requested on time can matter, especially when someone calls on a Friday afternoon about a prescription that runs out over the weekend.
What we built
We built an AI receptionist that answers the phone and handles the routine part of each call, and connected it to the practice's own tools instead of replacing them:
- New patient intake: answers basic questions about the practice and collects name, date of birth, contact details, insurance, what the caller is looking for, availability and whether they prefer telehealth or in-office. It asks no medical questions and passes the next step to staff.
- Scheduling: identifies a known patient, checks Tebra for open slots with the preferred provider, offers the options, books the one the patient picks and sends an SMS confirmation after the call.
- Refill requests: collects the medication, dosage and pharmacy, texts the patient a confirmation and sends the request to staff. The AI never approves a refill. A provider always makes that decision.
Safety and privacy by design
The agent checks for safety first on every call. If a caller mentions suicide, self-harm or another emergency, it points them to 911 or the 988 Suicide & Crisis Lifeline and alerts staff. Controlled medications, medical advice and any request to speak to a person are handed to the front desk. Caller identity is verified before any record is read, only the information needed is collected, and SMS confirmations stay brief.
The stack
The conversation layer is a voice agent. The workflow layer is n8n and Python, which run the intake, scheduling and refill steps. The data and messaging layer is the Tebra API for real patient records and real appointment slots, Google Apps Script for the practice's existing Google Workspace forms, and SMS for confirmations.
Measured results
- 200+ patient interactions handled every week
- 60% reduction in manual scheduling workload
- 300+ prescription refill requests captured every month
- 40% better accuracy from live Tebra record lookups
- 1,000+ automated SMS notifications sent to patients
- 35% higher patient engagement
Read more
We wrote up the full build, including the call flow, safety rules and HIPAA considerations, in Building an AI Receptionist for a Psychiatry Practice. To see how the same approach works for other businesses, visit our AI receptionist page.
