Referrals are how many Indian hospitals grow: a family doctor sends a patient, a health camp finds someone who needs treatment, a nursing home refers a case it can't handle. But a referral only turns into a visit if someone at the hospital calls the patient, understands what they need, and tells the right department they're coming. A multi-specialty hospital group in Gujarat uses an AI voice agent in Gujarati for exactly this intake call. This post explains the flow so you can run it for your hospital.
Build an intake agent in your language: Start free on Edesy — Rs 50 free credit, no demo needed.
Why referral intake slips
- The front desk is busy with walk-ins. Calling referred patients is the task that waits.
- Details arrive incomplete. A referral slip may have a name and phone number, but not what the patient needs or when they'll come.
- Departments don't know who's coming. The specialist finds out when the patient is already at the counter.
What the AI agent does on each call
The hospital group's agent works as a polite medical assistant. On each call with a referred patient it collects four things, one at a time:
- The patient's full name, confirmed back to them.
- What treatment they need, in the patient's own words.
- Which department's doctor they should see, for example orthopaedics or cardiology.
- When they will arrive at the hospital.
It closes the call by confirming what it heard. The agent does not diagnose, does not suggest medicines, and does not promise a doctor's time it can't see. If the patient describes an emergency, it tells them to use the hospital's emergency number.
Where the details go
When the call ends, the captured details are written to a Google Sheet: one row per call with the time, patient name, treatment needed, department, expected arrival and a short call summary. The front desk and department coordinators work from that sheet. If you prefer your own system, the same data can go through a webhook or Zapier.
One practical lesson from this deployment: the sheet columns must match the field names the agent produces. When they didn't, the rows arrived with blank columns until the names were aligned. See blank Google Sheets columns after AI calls.
Designing your intake script
- Keep it to what the hospital needs to act. Name, need, department, arrival. Don't collect history the doctor will take anyway.
- Accept everyday words. Patients say "knee pain" or "heart check-up", not department names. Let the agent map the need to a department and repeat it back.
- Confirm the arrival time as a day and time, so coordinators can plan (see know who's arriving today).
- Tell patients the call is automated and give them a way to reach a person.
- Retry unanswered referrals within your calling window instead of dropping them.
Where else this works
The same intake flow fits health camp follow-ups, diagnostic centres receiving doctor referrals, and hospital groups that route patients between branches. For after-hours enquiries, pair it with an AI night receptionist.
See the real deployment
Read the anonymised case study: how a Gujarat hospital group runs referral intake calls in Gujarati, or see what an AI voice agent for healthcare can do.
Create a free account and test an intake call on your own phone, or talk to us about connecting your system.