Every clinic knows the pattern. The front desk is busy with the patient in front of them, the phone rings twice, then stops. Someone calls at 9:40 pm to ask whether the doctor sits on Sunday. Nobody answers, and that patient books somewhere else.
An AI receptionist can answer those calls and WhatsApp messages around the clock. Done carelessly in a healthcare setting, it can also say the wrong thing to a worried patient. The difference is almost entirely in the preparation.
This checklist is for clinics, diagnostic centres, dental practices and physiotherapy centres in India. It covers what to decide before go-live, whichever vendor you use.
1. Write down exactly what it may do
Start with a short list of jobs, in plain words. A sensible first scope:
- Answer questions about timings, doctors’ days, location, parking and consultation fees
- Book, move and cancel appointments within rules you set
- Send directions and a booking confirmation on WhatsApp
- Take a message and a callback number for anything else
And a list of what it must never do:
- Give medical advice, interpret symptoms, reports or prescriptions
- Promise a diagnosis, outcome or a doctor’s opinion
- Quote prices for procedures that depend on examination
- Discuss one patient’s details with another caller
If a caller asks something outside the first list, the agent’s job is to take a clear message and get a person involved. Write that down too.
2. Decide what counts as an emergency
This is the most important part, and it can’t be left to the AI’s judgement. Agree on a short list of phrases and situations that trigger an immediate response: chest pain, difficulty breathing, heavy bleeding, unconsciousness, a child having a seizure, thoughts of self-harm.
For those, the agent should stop the booking conversation, tell the caller to call 112 or go to the nearest emergency department, and alert the on-call staff member. Test this more than anything else, in English, Hindi and the regional language your patients use.
3. Get the booking rules right
Most of the agent’s value is in booking. Write the rules as you’d explain them to a new receptionist:
- Slot lengths per doctor and type of visit (new consultation, follow-up, procedure)
- Which slots can be booked by the agent and which need staff approval
- How far ahead people can book, and the cut-off for same-day bookings
- What happens when a doctor is on leave
- How many reschedules are allowed before staff step in
- Whether advance payment is needed, and how the link is sent
Connect the agent to your actual calendar or practice software. A receptionist that books from a separate spreadsheet will double-book within a week.
4. Load only approved information
Make one document with everything the agent is allowed to say: doctors’ names and qualifications as you want them stated, timings, fees, address, landmark, parking, insurance or TPA tie-ups, documents to bring, preparation instructions for tests. Have the doctor or practice manager sign it off.
Then decide who updates it and how often. Out-of-date timings frustrate patients quickly, whether a person or an AI gives them.
5. Choose the languages and the voice
Patients won’t switch to English to suit your software. Decide which languages to support and test the agent with people who actually talk the way your patients do, including Hinglish and code-switching mid-sentence. Make sure it says medicine names, doctor names and local landmarks correctly.
Pick a calm, unhurried voice. Callers to a clinic are often anxious or unwell.
6. Be honest that it’s an AI
Start every call with who is speaking: the clinic’s name and that this is an automated assistant. Offer a person. “If you’d like to speak with our staff, I can arrange a callback” is enough. Patients who feel tricked don’t come back, and a clear introduction also keeps you on the right side of the rules on automated calls. Our guide to TRAI’s 2026 rules for automated calls covers outbound calling, such as reminder calls.
7. Treat patient data as sensitive
Health information is among the most sensitive personal data you hold. The DPDP rules apply in full from May 2027, and patients expect discretion now. Before go-live:
- Collect the minimum: name, phone number, preferred slot, and the reason for the visit only if the doctor needs it in advance.
- Check where call recordings and transcripts are stored, who can access them and how long they are kept.
- Get the vendor’s data processing terms in writing.
- Publish a short notice explaining that calls may be handled by an automated assistant and recorded.
Our DPDP checklist for enquiry data applies here as well.
8. Test with real scenarios, then watch the first fortnight
Before any patient talks to it, run at least 30 test calls covering: a simple booking, a reschedule, a fee question, a doctor on leave, an angry caller, a caller who only speaks the regional language, a bad mobile line, an emergency phrase, someone asking about another patient, and a request for medical advice.
For the first two weeks after launch, have someone read every transcript each day. Fix the information and rules quickly. After that, a weekly review is usually enough.
9. Measure what matters
Track the numbers that tell you whether it’s working:
- Calls answered that would previously have been missed, especially after hours
- Appointments booked by the agent, and how many of those patients turn up
- Calls handed to staff, and why
- Complaints or corrections
If no-shows rise, add a WhatsApp reminder the day before. If handoffs are mostly one type of question, add that answer to the approved information.
Where we fit
Our AI receptionist is built for this kind of job: calls and WhatsApp answered from your approved information, bookings within your rules, emergencies and anything unusual routed to your team, and every conversation logged for review. It discloses it is an AI at the start of the call. It costs ₹30,000 once for setup, then ₹10,000 a month including 1,000 minutes, and ₹9 a minute after that, with no GST charged. The monthly fee covers monitoring and updating as your clinic changes. If you’re comparing vendors, our guide to what an AI phone agent really costs shows how to compare quotes.
Whoever builds it, the checklist above is the part that keeps patients safe. Insist on doing it.