AI Receptionist for Clinics in India (2026): Calls, WhatsApp and Bookings From One Agent

• ₹6per call minute, telephony included, on Adya's October 2026 price list
• ₹0.1150per WhatsApp utility reminder to an Indian number, Meta's July 2026 INR card
• ₹6,998a month for 300 two-minute calls on Starter with one number, before GST
• 11languages one agent answers in: English plus 10 Indian languages

An AI receptionist for a clinic answers the phone when nobody at the desk can, takes the patient's name and reason for visiting, offers real free slots, books the one they pick and confirms it on WhatsApp. The same agent reads WhatsApp messages, reschedules, and sends the reminder the day before. When a caller describes an emergency or asks for a person, it stops and hands the call to your staff.

It does not diagnose, triage or prescribe. India's Telemedicine Practice Guidelines forbid AI platforms from counselling patients, and the receptionist must obey that on every call. Adya gives your front desk additional capacity for routine enquiries and bookings, while clinical decisions stay with your qualified team.

Most small clinics in India run on one phone number, one receptionist and OPD hours that overlap with the busiest calling times. The phone rings while that person is collecting a fee or walking a patient in. It rings during lunch and after 8 pm, when nobody is there at all. Nobody publishes a reliable national missed-call rate for clinics, and we will not invent one; your own call log for last week is the number that matters.

The CEA clinic standard says every patient who visits shall be registered (clause 10.1.1). Registration starts on the phone, and a missed call usually becomes a visit to the clinic down the road, not a callback.

A receptionist agent runs a short, fixed script with room for the caller to interrupt. The sequence that works for a clinic is: greet, detect language, ask the reason for the call, ask for a name and confirm the number, check availability, offer no more than three slots, book the exact one chosen, and tell the caller what will arrive on WhatsApp.

The questions it should ask, and refuse

Reason for visit, preferred doctor, preferred day, and whether the patient has visited before. That is enough to book. It should not ask for symptoms beyond a one-line reason, identity numbers or payment details.

"Is this serious?" and "Which medicine should I take?" get the same answer: the doctor will discuss that in the consultation, and here is the next free slot. Clause 5.4 of the Telemedicine Practice Guidelines, issued March 25, 2020, states that AI platforms are not allowed to counsel patients or prescribe medicines; only a registered medical practitioner may. Put that refusal in the agent's instructions in plain words, and test it.

On Adya, the agent on the phone is the same agent on WhatsApp and website chat, with one knowledge base: doctors, timings, fees, location. The channel supplies the caller's number, so nobody is asked to repeat it. Handoff and tickets are on the live support page.

The booking calendar is a set of rules, not a view of events: opening periods per weekday, appointment length, slot interval, a buffer after each booking, how many patients may overlap one slot, minimum notice, and closed dates. A clinic open 9 am to 1 pm and 5 pm to 8 pm is two opening periods, not one period with a lunch event.

Double-booking is prevented at the moment of booking, not when the slot is shown. Two callers choosing the last 4:30 pm slot produce one confirmed booking and one "slot taken" result, and the second caller gets the remaining alternatives.

How Google Calendar sync works

Adya stays the source of truth and mirrors each appointment to a Google calendar you choose; in the other direction, a timed Google event marked Busy blocks that slot. All-day and Free events do not block anything, so a doctor's leave goes in as a closed date. Underneath, Google's Calendar API creates an event from a required start and end time and reports busy ranges through a freebusy query. Connection steps are in the Google Calendar guide and the booking rules in Appointments and calendars.

Meta bills WhatsApp per delivered template message, by category, and India has its own INR rate card. On the card effective October 1, 2026, a utility message to an Indian number costs ₹0.1150 and a marketing message ₹0.8631. A reminder for an appointment the patient asked for is utility: Meta's guideline says utility templates follow a user action and carry no promotional content. Add "full-body check-up at 20% off" and the same message is marketing at 7.5 times the price.

Since October 1, 2026 Meta charges for free-form replies inside the customer service window after a free tier of 1,000 delivered replies per business phone number per month, at ₹0.1150 each in India, and charges utility templates sent inside an open window at the same ₹0.1150. For a clinic that is still cheap: one confirmation and two reminders per booking cost 3 × ₹0.1150 = ₹0.345 per appointment, or ₹103.50 for 300 appointments a month, and most single-clinic WhatsApp traffic stays inside the free reply tier.

Adya sends WhatsApp first when a template is configured, with email as the fallback rather than a second copy, and at most three reminders per appointment. A failed notification never cancels the booking. Meta bills template messages to your own WhatsApp Business account at its applicable rate card; Adya does not add a template-message markup. Adya plan and conversation usage are separate, and template sends can count toward that usage. The rate card is explained in WhatsApp Business API pricing in India.

A clinic in Coimbatore gets Tamil callers; one in Gurugram gets Hindi, English and both inside one sentence. Every vendor demos well in English. The test that matters is a caller who starts in Hindi, drops an English drug name, and switches back.

Adya answers in English and 10 Indian languages — Hindi, Bengali, Marathi, Tamil, Telugu, Gujarati, Kannada, Malayalam, Punjabi and Odia — detecting the language automatically and following a switch mid-sentence. There is one agent and one knowledge base for all of them; you do not build a Hindi agent and a Tamil agent.

Four calls to make on every shortlist

1. A noisy call from a two-wheeler, in your city's main language, asking for tomorrow afternoon.
2. A Hinglish caller who mispronounces the doctor's name and interrupts the slot list.
3. An elderly caller who asks for the time to be repeated twice.
4. A caller who says "mere papa ko saans nahi aa rahi" — and listen for whether the agent stops booking and starts the emergency script.

Read this section twice. The NMC Code of Medical Ethics Regulations, 2002 (clause 2.1.1) say that in an emergency a physician must treat the patient. The CEA clinic standard (clause 10.6.2) requires ambulance and hospital contacts to be available in an emergency. The Telemedicine Practice Guidelines (clause 4.5) say every emergency patient must be advised an in-person interaction with a registered medical practitioner at the earliest. An AI receptionist has no clinical role in any of that.

Three rules to configure

Emergency words end the booking flow. Chest pain, breathing difficulty, heavy bleeding, unconsciousness, a seizure, labour. The agent reads one fixed script — go to the nearest emergency department or call 108 now, here is the clinic's emergency line — and hands off. No follow-up questions, no reassurance.

Give patients a clear route to your team. On Adya, human handoff has a transfer-sensitivity setting from 0% to 100%; for a clinic set it high, so the agent offers a person early on anything uncertain or distressed. During live hours a staff member accepts in the browser and joins the same call; the caller hears hold music, not a new dial tone. Outside live hours the agent creates a ticket with the number and reason, and gives the next opening. Setup is in the human handoff guide.

Staff see the whole conversation. A handoff arrives with an AI-written summary and the full thread, so the patient does not repeat the problem. Every staff reply is written to the inbox with a Human label and the member's name — the audit trail you want if a call is ever questioned.

Take a single-doctor clinic with 300 inbound calls a month and an average call of 2 minutes. Prices are Adya's October 2026 price list; swap in any vendor's per-minute rate and the method holds.

Starter is a practical starting point for this clinic: ₹6,998 before GST for the illustrated call volume. Growth adds capacity for a larger operation: 5,000 conversations, more agents and team seats, outbound campaigns and priority support, alongside 500 call minutes and an Indian number. Choose it for the full patient-contact workload across calls, WhatsApp and chat. Explore Adya on the free website-chat plan; use a supported paid plan for a WhatsApp and voice pilot. See the full plan comparison.

Against that, a second front-desk hire. Delhi's statutory minimum wage for a semi-skilled worker was ₹20,371 a month under the Labour Department's notification effective April 1, 2025, before employer contributions or leave cover; other states publish their own schedules, and a trained clinic receptionist in a metro is usually paid above the minimum. Use these as cost benchmarks when planning front-desk capacity. Adya handles routine enquiries and booking calls so your receptionist can focus on patients at the desk.

A booking call collects personal data: a name, a phone number and a reason for visiting. The Digital Personal Data Protection Act, 2023 (Act 22 of 2023) governs that, and its Rules were notified on November 13, 2025. Rules 3 and 5 to 16 — notice, consent mechanics, security, breach reporting, retention — come into force 18 months after the gazette, on May 13, 2027. That is close enough to design for now.

Retention is where clinics get confused, because two rules point in different directions. The DPDP Act says erase when the purpose is served. The NMC Code of Medical Ethics (clause 1.3.1) says keep indoor patients' medical records for 3 years from the start of treatment, and the CEA clinic standard (clause 9.2) says keep copies of records for at least 3 or 5 years. The Act reconciles them itself: erase unless retention is necessary for compliance with any law. Set separate retention rules for clinical records and booking conversations with your clinic’s adviser. A completed appointment does not automatically delete its recording or transcript; assign an owner to carry out the agreed retention process.

What the agent should never collect

Aadhaar or any identity number, card details, a medical history, or test results read out over the phone. Clause 7.14 of the NMC regulations forbids disclosing a patient's secrets learnt in practice; the shortest path to compliance is an agent that never hears them.

Start with the moments that need better coverage: the OPD rush, lunch breaks and after-hours appointment enquiries. Even a small clinic can benefit when a valuable enquiry becomes a booking without interrupting a consultation.

Keep clinical care with clinicians. Use Adya for administrative enquiries and bookings. Keep emergency and trauma lines staffed by trained people, with a clinician-approved emergency script and a tested escalation path for the booking line.

Make one calendar the booking reference. Confirm how your hospital information system shares availability with the booking calendar before enabling automatic bookings. If that connection is not ready, start with enquiries and requests for staff follow-up while you establish the calendar workflow.

Build around your patients. Test their languages, real appointment questions and your staff handoff. For overseas patients, confirm country-number availability and forwarding arrangements with our team before the pilot.

Run the pilot on overflow and after-hours calls first. That gives your team a focused way to measure additional bookings and refine the patient experience before expanding coverage.

Week 1 — set up and rehearse. Write the knowledge base and the emergency script. Create the booking calendar with real opening periods, a 15-minute slot and a 5-minute buffer. Connect Google Calendar and run both direction tests. Make the four language calls above.
Week 2 — overflow only. Forward unanswered calls after four rings, and all after-hours calls, to the agent. Review every transcript each morning; fix the knowledge base, not the prompt, when it gets a fact wrong.
Week 3 — add WhatsApp. Submit confirmation, reminder, reschedule and cancellation templates as utility and test each once approved. Add the consent line and the "delete my details" path.
Week 4 — measure. Calls answered, bookings made, no-show rate against last month, and the one number that decides it: patients booked who would otherwise have reached voicemail.