01 Live deployment · since 2024

A WhatsApp receptionist for a senior pulmonologist.

A 24/7 AI assistant that answers patients across 5 Mumbai clinics — in Hindi, Marathi, English, and any other language a patient happens to write in. It hasn't slept a night since it went live.

Live · serving patients today Channel — WhatsApp Business Cloud API Built & operated by SahAI
A The brief

Five clinics. One number. No staff to answer it.

A senior pulmonologist with 37+ years of practice across five clinics in Mumbai. Patients across Ghatkopar, Kurla, Chembur, and the surrounding areas were messaging the practice WhatsApp at all hours — asking the same questions about timings, addresses, fees, services, and appointments.

A receptionist was answering them during the day. At night, on holidays, and during consultation hours, messages piled up — and lost patients along with them. Hiring an after-hours receptionist costs ₹15,000–25,000 a month. That receptionist still sleeps, still speaks one language at a time, still can't be in five places.

We didn't want a chatbot. We wanted something patients could trust at 2am, in their own language, without anyone on staff lifting a finger. — The clinic's request, paraphrased
B What we shipped

A receptionist that never sleeps — and never oversteps.

Eight specific behaviours, all derived from real patient conversations during the first weeks of testing.

Always on Answers 24/7. Patients get an instant reply at 2am, during consultation, during holidays — whenever. No "please wait for the doctor."
Multilingual Mirrors the patient's language automatically. Marathi gets Marathi. Hinglish gets Hinglish. English gets English. No setup, no templates, no menu.
Location-aware Suggests the nearest of the 5 clinics based on the patient's area — with address, timings, and a Google Maps link.
Appointments Captures bookings with the patient's name, issue, and preferred time — then politely confirms.
Staff alerts The clinic team gets a formatted WhatsApp the instant a booking is collected — no thread reading required.
Doctor's voice Replies in first person, as the doctor. Warm, professional, calm. Tuned over weeks until it sounded right.
Safe by design Never diagnoses on WhatsApp. Will not promise availability it can't guarantee. Always defers to an in-person visit or a phone call when in doubt.
Private dashboard Every conversation searchable by the doctor and the clinic team. Live stats, parsed appointments, full thread history.
C See it working

A real conversation. A real dashboard.

Taken from the live deployment. Same system runs for any clinic — only the names, languages, and locations change.

Private monitoring dashboard — conversations, live stats, and a parsed appointment from a Marathi-speaking patient
The private dashboard — every conversation searchable, stats live, appointments flagged the instant they land.
Marathi — nearest-clinic suggestion & appointment capture.
English — dignified refusal to diagnose, booking captured instead.
D Under the hood

The architecture, in plain English.

Nothing exotic. Just careful choices, each one aimed at making the conversation feel human.

/ 01

WhatsApp Business Cloud API

The clinic's existing WhatsApp number is registered as a verified business account on Meta's official Cloud API. Free service conversations cover normal volume; per-conversation costs are folded into the monthly engagement.

/ 02

A prompt tuned to the doctor's voice

The system's behaviour is governed by a long, carefully written prompt that captures the doctor's tone, the practice's services, every clinic's address and timings, and the explicit safety rules. It was rewritten more than a dozen times during testing.

/ 03

Top-tier language models

Patient messages are interpreted and replied to by frontier multilingual models — fluent in every major Indian language. The reply is in the same language the patient wrote in, automatically.

/ 04

A private database, owned by the clinic

Every conversation is stored — searchable in the dashboard, exportable on request. Nothing shared with any third party. If the clinic ever leaves, the data leaves with them.

/ 05

Staff notifications, end-to-end tested

When a booking is collected, the clinic team gets a formatted WhatsApp message with the patient's name, issue, preferred time, and the source clinic — within a couple of seconds.

E Outcome

Quietly running. Loudly useful.

The assistant has been in continuous service since launch. It handles the bulk of repetitive WhatsApp traffic — addresses, timings, services, fees — and routes every real booking to the clinic team with the patient's name, issue, and preferred time already parsed.

The doctor's staff still handle the conversations that matter. The assistant handles the ones that used to get lost.

Note On naming

Why the doctor isn't named here.

Out of professional courtesy and patient confidentiality, this case study refers to the doctor only by specialty and experience. A direct introduction is available on request — message the studio.

Want one of these?

A version of this could be answering your patients by next week.

Two sentences about your clinic is enough to start. An honest answer comes back within a working day on whether it's a fit.

WhatsApp the studio Other ways to reach us
Up next

Continue reading ·