A clinic AI agent answers appointment requests on WhatsApp at any hour, offers the slots the clinic has actually opened, books or reschedules them, and sends reminders so fewer patients fail to show. It does not give medical advice: any question about symptoms, medication or test results goes to staff.
Appointment requests arrive while the front desk is on a call, during a consultation, and after the shutter comes down. The ones that arrive after closing are the expensive ones: a patient who needs a slot tomorrow messages at nine at night, gets no reply, and books with the clinic that answers.
A missed call in a clinic is not a lost lead in the marketing sense. It is a patient who needed something and found a closed door, and it usually does not come back to check whether the door has opened.
Give the agent the appointment book, not the consulting room. In our clinic builds that means it handles requests after hours and during busy periods, offers only slots that are genuinely open, confirms them, and manages the flow of changes that eat front-desk time.
Everything is transactional: who, which doctor, when, how to get there, what to bring. That is a narrow job and a measurable one.
Symptoms, medication and dosage, test results, pregnancy, young children, and anything that reads like an emergency are not appointment requests. They are clinical questions, and the agent's only job is to route them to staff — quickly, and with the message intact.
Emergency routing deserves its own standing rule, written before go-live and tested: the number to call, the instruction to go to the nearest emergency department, and no attempt at triage in between. An agent that tries to be helpful here is a liability, not an assistant.
A clinic agent is only as good as the calendar behind it. If the practice management system has an API, bookings should write into it directly. If it does not, the agent takes the request, tells the patient when they will hear back, and a staff member confirms — slower, but honest, and far better than a double booking.
Decide the rules before configuration: how long each consultation type blocks, how much buffer sits between patients, which doctors have fixed theatre or clinic days, and what the agent does when two patients want the last slot at the same moment.
Two reminders do most of the work: one the day before, one a couple of hours ahead, each with the address, the doctor and a way to cancel. The second reminder is the one that catches the patient who forgot, and a one-tap cancel is what turns a wasted slot into one the waitlist can take.
Be careful with the promises here. No-show rates differ enormously between specialities and localities, and no honest vendor can tell you a percentage in advance. Measure your own rate for four weeks before and after, and use that.
Appointment messages contain personal data, and in a clinic that is not a small matter. Before switching anything on, get straight answers to four questions: where the conversation transcripts are stored, how long they are kept, who inside the clinic can read them, and what is written into the patient record versus kept only in the chat.
Also settle the number itself. A business number on the WhatsApp Business API is the right base for patient messaging, with a verified business profile so patients can see who is writing. Personal data handling brings obligations under applicable Indian law, including the DPDP Act, and the vendor should be able to describe their storage and deletion in plain sentences rather than a policy page.
Patients write in English, Hindi, Hinglish and the local language, often in one sentence and usually without punctuation. The agent has to read that, answer briefly, and never use clinical vocabulary it cannot explain. Short messages work better than complete ones: a time, a doctor, a place, a confirmation.
One number should decide whether this stays: bookings confirmed without a staff member touching the conversation. Around it, four supporting measures tell you where to fix things.
It will not create open slots, shorten a waiting list that is long because the clinic is full, or make staff keep availability current. Nor does it replace a receptionist: it takes the messages that arrive when nobody is free, and the routine changes that eat a front desk's morning.
Said plainly: the agent buys your staff time and stops after-hours requests from dying unanswered. It does not increase the number of patients a doctor can see.
It depends on the software. If your practice management system exposes an API, the agent can check availability and write the booking back. If it does not, the agent takes the request, confirms the patient will hear back shortly, and a staff member completes the booking — no invented slots.
It must not, and that should be a hard rule rather than a model setting. Anything about symptoms, medication, results, pregnancy or emergencies is routed to staff, with emergency messages handled by a fixed instruction to call or attend an emergency department.
Safety here means storage and access, not the model alone. Ask where transcripts live, how long they are retained, who in the clinic can read them, and what is copied into the patient record. A business number on the WhatsApp Business API with a verified profile is the right channel; handling personal data brings obligations the clinic and vendor should both be able to describe plainly.
Reminders and one-tap cancellation reliably help, but no honest number can be promised in advance — no-show rates vary by speciality and locality. Measure your own rate for four weeks before and after, and judge on that comparison.
On our plans, one agent on one channel starts at ₹999 per month on Standard, and ₹14,999 on Pro where several agents or channels are needed. Setup and the CRM or calendar connection are included; WhatsApp conversation charges from Meta are separate and usage-based.