For clinic owners, practice managers and medical directors

AI agents for clinics: the front desk that never puts a patient on hold.

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.

The week this fixes

If this is your Monday, keep reading

The phone and the WhatsApp inbox compete for one receptionist

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.

Reschedules and no-shows eat the schedule

"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.

Intake happens on paper, in the waiting room, twice

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.

One workflow, before and after

Booking, rescheduling and reminders over WhatsApp

Trigger: A patient messages "Can I see Dr Mehta this week?" at 9:40 pm.

Today

  1. PatientMessages at night. No reply until the desk opens.
  2. Front deskReads it at 8:10 am between phone calls, opens the practice system, finds a slot.
  3. Front deskReplies with two options, waits. Patient answers at lunch. Slot has gone.
  4. Front deskRepeats. Books. Types the name and number into the system by hand.
  5. Front deskReminder the day before, if there was time to send reminders that day.

With the agent

  1. AgentReplies within a minute: confirms the doctor, asks new or returning, offers three live slots from the practice system.
  2. AgentBooks the chosen slot, writes the appointment to the system, sends a confirmation and the pre-visit instructions the clinic approved.
  3. AgentSends the intake form link to a new patient; answers are written to the record before the visit.
  4. AgentReminder the day before. "Can I move it?" is handled the same way, inside the rebooking rules.
  5. Front deskSees the morning summary: bookings made, exceptions flagged, one message escalated as urgent.

What stays with a person

  • Anything symptom-shaped: the agent recognises it, replies with the clinic’s approved urgent-care guidance and flags a nurse. It never triages.
  • Fee disputes, complaints and anything the rules do not cover: handed to a person with the full transcript.
  • The rules themselves: which doctors, which slots, how far ahead, how many reschedules. Written once by the practice manager, changed without a developer.
What it saves

What the front desk gets back: a worked example

MeasureTodayWith 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 messageNext morning, or whenever the desk is freeUnder a minute, 24 hours a day
Intake typed in by handEvery new patientExceptions only

Worked example. The assumptions:

  • A three-doctor clinic handling about 60 booking-related messages and calls a day
  • Roughly 4 minutes of staff time per booking, reschedule or reminder, including the system entry
  • The agent completes about 70% of those within the rules; the rest still reach a person
  • Six working days a week

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.

The agents

Usually one to start, three when it is working

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.

Front-desk agent

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.

  • check_availability
  • book_appointment
  • reschedule_appointment
  • send_intake_form
  • escalate_to_staff

Reminder and follow-up agent

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.

  • list_tomorrows_appointments
  • send_reminder
  • offer_slot_to_waitlist
  • log_outcome

Intake and record agent

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.

  • create_patient_draft
  • attach_intake_answers
  • flag_incomplete

Guardrails

  • No medical advice, ever. Red-flag words route to a person with no attempt to triage in between.
  • Writes only to the appointment book and to draft records. Nothing clinical is created or changed by the agent.
  • Every action logged with the message that caused it, so the practice manager can audit a week in ten minutes.
  • Patient data stays in your systems and in India where you need it to; the model sees the conversation, not the record history.
  • Consent language for WhatsApp and for recording intake is yours to approve before anything goes live.

Systems it talks to

  • WhatsApp Business API (Meta Cloud API) and your website chat
  • Your practice management or clinic software: PHMS, Practo Ray, Cliniko, or a Google Calendar if that is what you run on today
  • Google Sheets or Airtable for the rules, the knowledge base and the morning summary
  • SMS or email for reminders where WhatsApp is not the patient’s channel
Proof, and its limits

What we have built, and what we have not measured yet

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.

FAQ

Questions clinics ask us

Will the agent give patients medical advice?

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.

Which clinic software does it work with?

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.

What does it cost and how long does it take?

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.

Is patient data safe with an AI?

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.

What happens when the agent gets something wrong?

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.

Next step

Tell us about your clinic

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.

Or send the details

Name and email are all we need. The rest helps us reply with something specific.