An agent that answers WhatsApp, web chat and missed calls, books and reschedules inside your rules, sends reminders and intake forms, and hands anything symptom-shaped to a human immediately. Admin only. Every clinical decision stays with a clinician.
Thirty minutes, free. You leave knowing which workflow to start with, what a pilot covers and what it costs.
PilotAI Agent Pilot: one workflow, 2–3 weeks, $4,500 (₹3,50,000) fixed. What the pilot includes.
Who this is for: Outpatient clinics, specialty practices and small multi-doctor centres with one to five receptionists and a WhatsApp inbox nobody can keep up with.
Monday 8:00: forty unread weekend messages, a ringing phone and two patients at the counter. Whoever is standing in front of the desk wins. Everyone else waits, and some of them book elsewhere.
"Can I move Wednesday?" is a two-minute job that happens thirty times a day. A slot that opens at 10:40 stays empty because nobody had time to call the waitlist.
Demographics, history and consent are filled in by hand, typed in later, and asked again at the next visit because the form never made it into the record.
Trigger: A patient messages "Can I see Dr Mehta this week?" at 9:40 pm.
| Measure | Today | With the agent |
|---|---|---|
| Staff time on bookings, per day | ~4 hours | ~1.2 hours |
| Staff time on bookings, per month | ~100 hours | ~30 hours |
| First reply to a patient message | Next morning, or whenever the desk is free | Under a minute, 24 hours a day |
| Intake typed in by hand | Every new patient | Exceptions only |
Worked example. The assumptions:
This is the visible saving. No-show rate, waitlist backfill and patients who stop booking elsewhere are the larger numbers, and they are the ones a pilot measures against your baseline rather than estimates.
Put your own numbers in the automation savings calculator, or get a price range from the AI agent cost estimator.
Each agent has a small set of tools it is allowed to call and nothing else. That list is the guardrail, and it is written down here so you can see it.
Lives on your WhatsApp Business number and web chat. Books, reschedules, cancels and answers the fixed-knowledge questions (hours, fees, which doctor does what, parking) from a knowledge base the clinic owns. Escalates everything else.
Runs on a schedule. Confirms tomorrow’s appointments, backfills a cancelled slot from the waitlist, sends the follow-up message a clinician approved after a visit, and nudges overdue reviews.
Collects demographics, history questionnaire and consent before the visit, checks the answers for gaps, and writes them to the patient record as a draft a clinician reviews. It never interprets what the patient wrote.
We build clinic software for a living: PHMS is our offline-first clinic management system, running in Indian clinics with patient records, appointment queues, prescriptions and a doctor–patient chat built in. The booking agent is a reference build you can try below. No clinic has published results with it yet, and we say so rather than quote a number we have not measured.
No. It is designed so it cannot: it has no clinical tools, and any message that looks like a symptom, a result or a medication question is answered with your approved urgent-care text and escalated to a named person. The safety is in what it is allowed to touch, not in hoping the model behaves.
Anything with an API or a calendar it can write to. PHMS, Practo Ray and Cliniko are straightforward; a clinic running on Google Calendar and a notebook is also fine, and often the quickest to start. If your system has no API we put a shared sheet between the agent and the desk for the pilot and decide on integration after.
The AI Agent Pilot is a fixed fee for one workflow in two to three weeks, usually the booking and reminder flow on WhatsApp. A full front desk with intake and follow-ups is a scoped build from USD 8,000. The cost estimator on this site gives you a range in two minutes.
The agent sees the conversation it is having and the slots it is offered, not the patient’s clinical history. Records stay in your system; the model provider does not train on your data under the API terms we use; and we write the data-flow diagram into the handover so you can show it to anyone who asks. The DPDP Act obligations stay yours and we design to them.
It will, occasionally, which is why it only books and reschedules inside rules you wrote, and why anything outside them goes to a person. Every action is logged with the message that caused it. In the pilot we review the log with you weekly and tighten the rules; after that the practice manager can do it alone.
How many doctors, which system you book in, and where the messages come from. We reply with a first take on what a pilot would cover.
Or start with a number: the AI agent readiness quiz takes three minutes and gives you a score and a 30-day plan.