Live prices
Prices and durations come from your services list on every message. Change a price at 3pm and it quotes the new one on the next message.
The assistant replies to patients on your WhatsApp number using your real prices, hours and doctors. It books into your calendar under the same rules as your staff, and hands anything tricky to a person.
Every message takes the same five steps. The cheap safety check runs before the expensive model, and the calendar has the final say.
A patient writes to your clinic's number. A burst of messages becomes one reply: it waits 8 seconds for them to finish typing.
Fixed patterns and a small model check for emergencies, abuse and requests for a person before anything else runs.
One model with tools for your services, doctors, availability, bookings and knowledge base. It picks the tools it needs.
Bookings go through the same conflict engine your staff use. It can't double-book a doctor.
The answer goes out on WhatsApp and lands in the shared inbox labelled 'Assistant', so staff see every word.
Six real situations. See what the guard decides, which tools run, and what the patient gets back.
Pick a message
Hi, how much is a cleaning & exam?
Safety guard
Safe. Passed to the assistant.
Assistant tools
list_services
Outcome
A cleaning & exam is $120 and takes 45 minutes. Would you like me to find you a time?
It starts switched off. Turn it on when your prices, hours and knowledge base are ready.
Give it a name, add instructions in your own words, list the words that should always reach a person, and write the message patients see when it hands over.
The assistant answers from your own data, read fresh on every message. When it doesn't know, it says so and hands over.
Prices and durations come from your services list on every message. Change a price at 3pm and it quotes the new one on the next message.
It offers times from each doctor's real roster, leave and bookings. Never a slot your staff couldn't book.
Parking, walk-ins, payment plans: write them as questions and answers. It searches them before answering.
It's told never to diagnose or advise on medicines, and it can't reach clinical notes at all. That part is structural.
Emergencies and requests for a person mute it on that thread and alert your desk with the reason.
An activity log shows each reply or handoff, the reason, the tools it used and how long it took.
Each turn is recorded as replied, escalated, skipped or failed, with the reason, the message that triggered it, the tools it called and how long it took.
Filter by outcome, conversation or date. When it hands over after failing to find an answer, you see the gap and can fill it in the knowledge base.
| Time | Patient | Outcome | Reason | Tools used | Took |
|---|---|---|---|---|---|
| 23:52 | Maria Lopez | Replied | Booked appointment | check_availability → book_appointment | 21.4s |
| 23:49 | Maria Lopez | Replied | Price and slots | list_services → check_availability | 24.8s |
| 22:41 | Priya Raman | Escalated | Voice note, no caption | none | 0.3s |
| 22:05 | Jordan Reed | Replied | Parking question | search_knowledge | 17.9s |
| 21:30 | Daniel Okafor | Escalated | Asked for a person | escalate_to_human | 2.1s |
| 21:12 | Hannah Wu | Skipped | Staff took over | none | 0.1s |
| 20:47 | Priya Raman | Failed | Provider timeout, handed to staff | list_services | 60.0s |
Two situations mute the assistant on that thread. The rest alert staff while it keeps answering anything else. The 'someone will reply' notice goes out at most once an hour, except for emergencies and requests for a person.
| Situation | What happens | Assistant |
|---|---|---|
| Possible emergency | Handoff message plus your country's emergency number. Staff alerted. | Muted |
| Patient asks for a person | Handoff message. Staff alerted. | Muted |
| One of your keywords | Staff alerted with the keyword that matched. | Stays on |
| Voice note or image, no text | Handoff message. It can't listen to audio or look at images. | Stays on |
| Clinical, billing or a complaint | The guard flags it for staff. | Stays on |
| Too many replies in an hour | It stops and alerts staff. | Stays on |
| A provider error or a stuck turn | The thread goes to staff. | Stays on |
We'd rather tell you now than have you find out in week two.
No. Prices and durations come from your services list every time. If a service isn't in your list, it can't quote it, and the activity log shows which tools it used for every answer.
Yes, when booking is on (the default once the assistant is enabled). It uses the same availability and conflict rules as your staff, so it can't double-book a doctor or book outside a roster. Cancelling stays off until you turn it on.
Emergency wording is caught by fixed patterns before any model runs, in English, Roman Urdu and Urdu script. The patient gets your handoff message plus the local emergency number, the assistant goes quiet on that thread, and your front desk is alerted.
No. The assistant has no tool that can reach clinical notes or prescriptions. It talks to whoever holds the phone, so it's kept away from anything clinical by design.
We assign a model to each clinic, from providers including Google and Anthropic. A model that hasn't been verified for reliable tool use is never given the booking tools: it answers questions and hands over instead.
No. It's included in Professional (3,000 replies a month) and Enterprise (25,000). Free and Starter clinics get the shared inbox without it.
Yes. Anyone at the front desk can press Take over on a thread. The assistant stays quiet there until someone hands it back, or until 24 hours after the last message.
By default, yes. That's when most clinics need it. You can turn off replies outside business hours in its settings.
One shared, real-time inbox for the front desk.
Rosters, leave and a calendar that can't double-book.
Confirmations and reminders on WhatsApp.
Set up your clinic in an afternoon. The free plan needs no card and no sales call.
Or book a demo