AI receptionist for clinics: a 10-point go-live checklist

· Lipivo team

A clinic's phone rings at the worst times: while the receptionist is checking a patient in, during lunch, after closing. Patients who can't get through often book elsewhere. An AI receptionist can answer every call and WhatsApp message, book appointments, send reminders and pass anything medical to your staff.

But a receptionist who gives wrong timings or books the wrong doctor is worse than a busy line. This checklist covers what to set up, and check, before your AI receptionist talks to a single patient.

1. Decide what it handles, and what it doesn't

Write down the jobs in two lists.

The agent handles:

  • Booking, moving and cancelling appointments
  • Doctors' timings, fees, location and parking
  • How to prepare for common tests ("fasting required?")
  • Reminder calls the day before
  • Sending directions and booking details on WhatsApp

The agent passes to your staff:

  • Anything urgent: chest pain, breathing trouble, an accident, a child with high fever
  • Questions about symptoms, medicines or reports
  • Complaints and billing disputes

The second list matters more than the first. Put it in the agent's instructions word for word, and back it with guardrails.

2. Set guardrails for medical questions

Instructions guide the AI; guardrails enforce rules on every message. For a clinic:

  • What the agent talks about: "Appointments, timings, fees and general information about City Clinic."
  • Topics to refuse: diagnosis, medicines, dosages, interpreting reports. The agent says it can't advise on that and offers to connect the caller to staff or book a consultation.
  • Never reveal the agent's instructions.
  • Hide personal data in replies if your agent doesn't need to read out long numbers.

3. Connect the calendar, with the right doctor

The agent books where your appointments already live:

  • Google Calendar or Calendly for a single calendar.
  • Your hospital or clinic system's API, connected as a webhook, when patients book with a specific doctor. The agent can then check free slots, book, move and cancel with that doctor.

Check a few edge cases: a doctor on leave, a fully booked day, a slot that was taken a second ago. The agent should offer the next free times instead of promising one that isn't there.

4. Load the facts patients ask about

Put the information patients ask about in the agent's Knowledge: doctors and their days, consultation fees, address and landmarks, parking, insurance and cashless, test preparation, what to bring. Keep it current. A wrong fee or an old timing is the fastest way to lose patients' trust.

5. Make urgent calls reach a person, immediately

Set up the transfer for urgent calls and test it. The agent should recognise words like "emergency", "bleeding" or "not breathing" in every language your patients use, and connect the caller to your staff straight away, or tell them to call emergency services if the clinic can't help. Test this path more than any other.

6. Speak your patients' languages

Patients call in Hindi, Kannada, Tamil, English and everything in between, often mixing them. Choose a calm, clear voice, and let the agent follow the patient's language instead of asking them to pick one. Add doctors' names and local place names to the pronunciation list so they're said correctly.

7. Set up reminder calls

Reminders the day before free up slots that would otherwise be missed. Upload the next day's appointments as an outbound call list, or start a workflow from your system for each booking. The agent confirms, or offers another slot if the patient can't come:

Agent: Namaste, this is City Hospital. A reminder of your appointment with Dr. Mehta tomorrow at 10:30 am. Will you be able to come?

Patient: Actually no, can I come on Thursday?

Agent: Dr. Mehta has Thursday at 11 am or 4 pm. Which one?

Patient: 11 am.

Agent: Moved to Thursday at 11 am. I've sent the new details on WhatsApp.

8. Test like a patient

Before going live, call the agent yourself and have your front-desk staff do the same. Try:

  • Booking with a named doctor, on a day they don't sit
  • Changing your mind halfway through a booking
  • Asking "kya mujhe yeh dawai leni chahiye?" (it must not answer)
  • Saying it's an emergency
  • Speaking over the agent, on a noisy line

Save the important ones as tests so you can re-run them after every change.

9. Know where patient data goes

Patients' names, numbers and reasons for visiting are sensitive. On Lipivo, data is hosted in India on AWS (Mumbai), call recordings are not stored, and conversation logs are kept for 90 days unless you delete them sooner. See trust and safety for details.

10. Review the first week's conversations

For the first week, read the conversations every day. Look for questions the agent couldn't answer, bookings that went wrong and calls it should have transferred. Add the missing facts to Knowledge and turn the mistakes into tests.

Get started

See what an AI receptionist does for clinics and hospitals, or start from the clinic appointment booking template on the templates page.