AI appointment booking assistant

Why bookings fall through the gaps

A booking is rarely lost at the moment it is made. It is lost in the handling after. The request comes in on Instagram, the receptionist means to add it to the diary after the current patient, and by then two more messages have arrived. Or two people book the same 11 am slot from two different channels. Or the patient booked ten days ago, nobody reminded them, and the slot sits empty while someone else was turned away for it.

The assistant fixes the handling. Every request, whichever channel it came from, ends up in the same calendar within seconds, with the conversation attached, and the reminders go out without anyone remembering to send them.

What you decide, what it decides

You decide which services it may book on its own, which doctors and which hours, how far ahead, and what needs a person first (a new patient for a specialist, say, or anything outside hours). It handles the rest: offering the nearest free slots, confirming, rescheduling, cancelling, and writing each change to the calendar and the patient list.

AI output is a strong first guess, not a guarantee. That is why every booking shows your team the message it was taken from, and why anything the assistant is unsure of is held for a person instead of guessed.

What the patient sees, step by step

A patient who has never been to the clinic sends an Instagram message after seeing a post: "Do you do teeth whitening? When can I come?" The assistant replies with the service, the time it takes, the doctor who does it and three free slots this week. The patient picks Thursday at 5, gives a name and a number, and gets the confirmation with the address and a note on what to avoid before the session. The whole exchange takes under two minutes and happens in the app the patient was already using.

A returning patient is simpler still: "same as last time, next week?" is enough, because the number is recognised and the last visit is on record. The assistant offers the usual doctor's free slots next week and confirms the one the patient picks. Reminders, reschedules and the message after a missed visit all arrive on the same thread, so the patient never has to look for a different number or remember which channel they booked on.

How it works for a US practice

In the US most patients do not book on WhatsApp; they call, send a text or use the form on the practice's website. The assistant works the same way on those channels. Calls are answered by an AI voice agent on your number, or only on calls forwarded after hours or when the front desk does not pick up. Text messages (SMS) get the same replies and reminders a WhatsApp patient would. A chat box on your website books from the same calendar. Every booking lands in Google Calendar, a Microsoft calendar or practice software that offers an API, and the front desk sees all of it in one list.

After hours is where it earns its place: a patient who calls at 8 pm hears a real answer, picks a slot and gets a confirmation text, instead of reaching voicemail and calling another practice in the morning. No-shows are handled by text: a reminder the day before, another a few hours before, and a rebooking message if the patient does not come. A patient can reply STOP at any time and the messages end.

Two practical notes. Text messages to US numbers have to come from a business number registered with the carriers through the SMS provider (the A2P 10DLC registration), which we set up with you; it usually takes from a few days to a few weeks. And if you handle US patient data, ask us about your legal requirements before we build.

Every channel into one calendar

Patients book the way they already behave. Some message on WhatsApp, some send an Instagram DM after seeing a reel, some fill the form on the website, some phone, and some walk in. The assistant takes the first three itself. Phone calls can go to an AI voice agent that answers, checks the same calendar and books, or stay with the front desk. Walk-ins and phone bookings are added by staff from the desk. All of it lands in one calendar, so a slot taken on Instagram at 10 pm is not offered on WhatsApp at 10:01.

The channel is recorded with each booking. After a month the owner can see which channel brings bookings that actually turn up, which is often a surprise, and which one brings questions but no bookings.

The rules a clinic actually needs

Booking software fails at the edges, and clinics are all edges. The assistant works from rules you set once: the length of each service (a consultation is not a cleaning), the gap a doctor wants between patients, which doctor sees new patients, how far ahead patients may book, whether a deposit is needed for long procedures, and what happens with a request outside hours. It can keep a waitlist: when a patient cancels, the next person who asked for that doctor is offered the slot before it goes back to general availability.

Returning patients are recognised by their number, so a patient who has been before is not asked for their details again, and the doctor's screen shows that this is a follow-up and when the last visit was. None of this needs the patient to install anything or remember a login; it is a message to the number they already have.

What the doctor and the owner see

The doctor sees their day: who is coming, for what, whether they have confirmed, and whether it is a first visit or a follow-up. No more asking the front desk what is next. The front desk sees every doctor's day and the live state of each booking: requested, confirmed, reminded, arrived, done, no-show.

The owner gets a weekly summary on WhatsApp: bookings made, attended and missed, by doctor and by channel, and how many bookings came in outside opening hours. Those numbers are the clinic's own, built from its own bookings, not an industry statistic. They are usually the first clear view an owner has had of how much of the week's business arrives when nobody is at the desk.

When it should not book, and what it says instead

Some requests must not be turned into a calendar entry. A message that reads like an emergency (chest pain, a child with a high fever, bleeding that will not stop) gets an immediate reply telling the person to call the clinic's emergency number or go to the nearest hospital, and the chat is pushed to a person at once. A request for a prescription refill, a sick note or a referral letter is not a booking either; the assistant tells the patient the doctor will respond and holds it for the doctor's queue. A procedure that needs a consultation first is offered the consultation, not the procedure.

These rules are written with the clinic before anything goes live, and they are tested with real messages from the clinic's own history, with the names removed. The assistant's reading of a message is a strong first guess, so the rule for anything it cannot place with confidence is the same: say something safe, and hand it to a person. A missed booking costs a slot; a wrongly handled emergency message costs far more, so the system is built to be cautious at exactly those points.

After the booking

The reminder goes the day before and again a few hours before. A patient who replies that they cannot make it gets offered another time at once, so the slot is reused instead of wasted. A patient who simply does not come gets a message later that day to rebook, and your team sees the no-show in the list. Over a month, that quiet follow-up is usually where the assistant earns its place.