What an AI receptionist does in a hospital, and what it must never do

Most calls to an OPD are about a handful of things: a time with a doctor, a change to that time, the fee, the timings, the way to the hospital. An AI receptionist is software that picks up those calls and deals with them the way a well-trained receptionist would, so the people at your front desk can look after the patients standing in front of them.

That much is true of any business. A hospital is different because of what the receptionist must not do, and that is where most of this guide goes.

What an AI receptionist is, and what it isn’t

An AI receptionist listens to what the caller says in their own words and answers in conversation. There is no menu. The caller says “कल डॉक्टर शर्मा मिलेंगे?” and gets an answer, not “press 1 for appointments”.

• It is not an IVR. An IVR makes callers fit its menu. An AI receptionist fits the caller.
• It is not a call centre. An agent takes one call at a time and works a shift. An AI receptionist can talk to several callers at once, at 11 at night and on Sundays, and hands over only the calls that need a person.
• It is not a doctor. It knows your hospital’s timings and fees, not medicine, and it is built to stay that way.

What it does on an OPD call

A good AI receptionist handles each of these from start to finish:

• Booking a time that is really free with the right doctor, and reading it back to the caller. It should say “booked” only once the booking is saved, and if someone took the slot a moment earlier, offer the next one.
• Moving or cancelling an appointment, after a clear yes from the patient.
• Answering questions about timings, fees, the address, parking and the documents to bring, using only what the hospital has approved.
• Finding the right department from what the caller describes. “घुटने में दर्द है” (my knee hurts) should lead to orthopaedics at a multispeciality hospital, without the caller knowing the word.
• Booking for the family: a parent, a child, or a second patient in the same call.
• Taking the details when it can’t help, so your team calls the patient back.

Here is what a booking can sound like. It’s an example written for this guide, and the names are made up:

Two details in that exchange matter more than they look. The second patient is handled in the same call, as a person at your desk would handle it. And the first booking is confirmed only once it is saved, not when the caller says yes.

What it must never do

Every business wants a receptionist who is polite and quick. A hospital also needs one with hard limits, enforced by how the system is built, not by hoping a model behaves:

1. It never diagnoses. “क्या ये दर्द serious है?” (is this pain serious?) is a question for a doctor. The receptionist’s job is to get the caller to that doctor, or to emergency care.
2. It never suggests medicines or doses, not even for a fever, and not even when the caller insists.
3. It never reveals the sex of a baby. Communicating the sex of a foetus is prohibited under the PC-PNDT Act, 1994, and a receptionist, human or AI, has no business near that question.
4. It never makes up a fee, a timing or a doctor’s hours. If the hospital hasn’t approved an answer, the right reply is “मैं पता करके बताती हूँ” (I’ll find out and let you know), followed by a call-back.
5. It never shares details of admitted patients with whoever happens to call.
6. It never claims to be a person. Asked “आप इंसान हैं?” (are you a person?), it should say it is the hospital’s AI receptionist.
7. It never books a routine slot for an emergency. More on that below.

When you speak to vendors, ask how each of these is enforced. “Our model is trained not to” is a weaker answer than “the system checks for it before anything is said”.

How it should handle an emergency

Emergencies don’t announce themselves at the start of a call. A caller booking a routine cardiology review may mention chest pain halfway through, almost in passing. So every sentence the caller says should be checked for warning signs before anything else happens.

When something sounds like an emergency, a good AI receptionist:

1. tells the caller what to do now, such as calling 108 or 112, or coming straight to the hospital’s emergency;
2. alerts your staff at once, so a person can follow up;
3. drops the routine booking.

Problems that are urgent but not an emergency should get the earliest time that day. Talk of self-harm should get a kind reply and the Tele-MANAS helpline, 14416, along with 112.

None of this makes it a triage system. It is a safety net. Be wary of any vendor who claims theirs catches every case.

Why the language it speaks matters most

Hindi is the mother tongue of 52.83 crore people, 43.63% of India, according to the Census of India 2011. In a study of 500 patients in the medicine OPD of a New Delhi public teaching hospital, 75.2% could not read English and Hindi was the main language for 99.6%.

On the phone, patients don’t speak textbook Hindi either. They mix in English words like appointment, report and fee. They say “15 के बाद” (after the 15th), “lunch के बाद” (after lunch) and “अगले बुधवार” (next Wednesday, which can mean two different dates). They read their number out in pieces and change their mind halfway.

So test any AI receptionist with calls that sound like your callers: an elderly patient, a noisy line, a son booking for his father, a caller who switches between Hindi and English mid-sentence. A polished English demo tells you very little.

How it fits with your phone line

Patients keep calling the number they already know, and they should. Your hospital’s line forwards calls to the AI receptionist. A sensible start is the calls your team misses and the ones after hours, then all calls once the team is comfortable. Never give patients a separate number to remember for the AI.

When a caller wants a person, the AI receptionist should be able to pass the call to your staff during the hours you choose, and add the caller to your call-back list if nobody picks up. Nobody should be dropped without a trace.

What happens to the patient’s data

Under India’s Digital Personal Data Protection Act, your hospital is the data fiduciary. It decides why and how patient data is used, and an AI vendor only processes that data for you. The core duties on notice, consent and security apply from 13 May 2027, under the DPDP Rules notified in November 2025.

Before you sign, get clear answers to these:

• Are calls recorded? If so, who can listen to them, and for how long?
• How long is the text of each call kept, and when is it deleted?
• Where is the data hosted?
• Who at the vendor can see your patients’ details, and is that access logged?

What it doesn’t replace

An AI receptionist takes the repetitive calls off your team’s hands. It doesn’t replace your reception desk. The calls it can’t finish should land with your team with the reason attached, and the questions it couldn’t answer should be answered once and then used on every later call. Done well, your receptionists stop reciting OPD timings over and over every morning and spend that time on the patients in front of them.

Eight questions to ask before you choose one

1. Does it talk the way our patients talk: everyday Hindi and Hinglish, with English words mixed in?
2. Does it work on our existing number?
3. What exactly happens, step by step, when a caller describes an emergency?
4. How does the system stop it from giving medical advice or inventing a fee?
5. Does it say it is an AI when asked?
6. Can callers reach a person, and what happens when nobody picks up?
7. Are calls recorded, how long is the text kept, and where is the data hosted?
8. Can we call it ourselves and try to trip it up before we decide?

Where DrMate fits

DrMate’s AI receptionist answers your hospital’s own number in everyday Hindi, books into your doctors’ real free times and puts everything it can’t handle on your team’s call-back list. No call audio is recorded, the text of each call is deleted after 30 days, and the app and its data are hosted in India, on Microsoft Azure (Central India). It is one part of a wider AI front desk, which also covers the queue and waiting times and the doctor’s room.

The last question on the list is the one we’d ask first. You can try ours on the demo line, 080 6553 5931, where it answers as Sunrise Hospital, a sample hospital we made up. Ask it something it shouldn’t answer, and see what it does.

Read this guide in Hindi (हिंदी में पढ़ें).