How I Built an AI Receptionist for My Physical Therapy Clinic Using Hushed, Vapi, Zapier, Gmail, and Google Sheets

Introduction

As a solo physical therapist running my own clinic in New York City, I wanted something simple:

I didn’t want to miss phone calls.

I didn’t want to hire a full-time receptionist.

I wanted every caller to receive a professional response — even when I was treating a patient.

After testing multiple AI voice platforms, I built an AI-powered receptionist using:

• Hushed
• Vapi
• Zapier
• Gmail
• Google Sheets

The entire system works automatically.

A missed phone call becomes a structured summary in my inbox within seconds.

Here’s exactly how it works.

System Overview

The caller never knows multiple services are working behind the scenes.

Everything feels like a normal phone call.

Step 1 — Hushed

Hushed provides the public business phone number.

Instead of answering calls directly, the number forwards every incoming call to Vapi.

Responsibilities:

• Business phone number
• Call forwarding
• SMS support (optional)
• Privacy protection

The patient only sees a single business phone number.

Step 2 — Vapi

Vapi is the brain of the system.

It answers every incoming call immediately.

The AI receptionist can:

• Detect English or Chinese automatically
• Ask why the patient is calling
• Collect:
• Name
• Phone number
• Reason for calling
• Insurance or self-pay
• Preferred appointment time
• Answer common questions
• Check appointment availability (optional)
• End the conversation naturally

The prompt is designed to sound like a real front desk receptionist rather than an AI chatbot.

Step 3 — Structured JSON

When the conversation finishes, Vapi sends structured data through a webhook.

Example:

{
  "patient_name":"John Smith",
  "phone":"(555)123-4567",
  "reason":"Neck pain for three weeks",
  "insurance":"Aetna PPO",
  "preferred_time":"Thursday afternoon",
  "summary":"New patient seeking evaluation for neck pain.",
  "urgent":false
}

Instead of parsing raw transcripts, structured JSON makes automation far more reliable.

Step 4 — Zapier

Zapier receives the webhook and performs multiple actions simultaneously.

Action 1

Create a new Google Sheets row.

Columns include:

• Date
• Caller Name
• Phone Number
• Insurance
• Reason
• Appointment Preference
• AI Summary

This automatically creates a searchable call log.

Action 2

Send an email notification.

Example:

Subject:

New Patient Call

Body:

New Caller

Name: John Smith

Phone: (555)123–4567

Insurance: Aetna PPO

Reason: Neck pain

Preferred Time: Thursday afternoon

Summary: New patient requesting an evaluation for chronic neck pain.

No manual documentation is required.

Action 3 (Optional)

Additional automations can include:

• Slack notification
• Microsoft Teams
• CRM
• Airtable
• Notion
• Google Calendar
• SMS

The webhook can feed almost any system.

Why Google Sheets?

Many people ask why I don’t use a database.

Google Sheets has several advantages:

• Extremely easy to search
• Accessible from anywhere
• Works with Zapier instantly
• Simple backup
• Easy export to Excel

For a small clinic, it is more than sufficient.

Why Email?

Email remains the fastest notification method during patient care.

While treating someone, I can immediately see:

• Who called
• Why they called
• Whether the call is urgent

No need to open another application.

Future Improvements

This system is only the first version.

Future upgrades include:

• Automatic appointment booking
• Insurance verification
• CRM integration
• Follow-up reminders
• Intelligent patient prioritization
• AI-generated SOAP note drafts

The goal isn’t replacing humans.

It’s allowing clinicians to focus on treating patients while routine administrative tasks happen automatically.

Final Architecture

A phone call lasting only a few minutes automatically becomes structured information, notifications, and records — without requiring a receptionist.

For a solo healthcare practice, this has been one of the highest return-on-investment automations I’ve implemented.