Confirming and reminding appointments with AI

How the bot works

Nine steps. The scheduling arithmetic, the cadence and the opt-out handling are ordinary code and cost almost nothing to run; the model is only used to read what somebody wrote back.

1. It reads the book you already keep. The appointments, the people, the staff member, the location and the type — out of the system the booking already lives in. It creates nothing and it takes no new bookings; it works what is there.

2. It confirms while the booking is still fresh. The confirmation goes out when the appointment is made, in your words: the time, the place, who they are seeing, what to bring, and how to change it. If only one message in the whole workflow is right, it should be this one.

3. It reminds on the rhythm you set, and no more. How many, how far out, and which one is the last — your decision, per appointment type. There is a hard ceiling per appointment and a hard quiet-hours window, because the limit has to live somewhere a model cannot move it.

4. It asks something that takes one word to answer. Not "this is a reminder" but "still good for Thursday at two?" A reply is a small commitment, and it is also information: a no, days ahead, frees a slot while you can still fill it.

5. It reads the reply, not just the yes. Confirmed, can't make it, wants a different time, asked a question, opted out, or something that needs a person. Six outcomes, and only three of them belong to the bot.

6. It offers only the slots you opened. Two or three numbered choices, from the slots your rules allow for that type and that staff member, inside the window you allow a change in. Reply with a number and it is held. There is no free-text negotiation with a bot.

7. It writes the change back where the book lives. One source of truth. The calendar, the front desk's screen and the reminder schedule all move together, so nobody is reminded about an appointment that moved this morning.

8. A missed appointment goes to a person, quietly. It records the miss, stops messaging, and puts it on somebody's list. No automatic call ninety seconds later, no fee, no chase — a person who has just been caught out is not somebody a bot should be ringing.

9. It keeps the whole thread. Every message sent, on which channel, at what time, from which template and version; every reply, how it was read, and what happened next. When somebody says they were never told, you can look.

What you configure

It can be configured to run in any business. This is what that actually means — the list of things that are yours, not ours.

Your reminder rhythm

How many reminders an appointment gets, how far ahead each one goes, and which one is the last — set per appointment type, because a cleaning booked six months out and a follow-up booked on Monday are not the same problem.

Your ceiling and your quiet hours

The most messages any one appointment may ever generate, the hours nothing sends in, and what happens when the same person has three appointments in one week. Both are hard limits, not defaults.

Your words

The confirmation, each reminder, the rebooking offer and the opt-out acknowledgment — your sentences, and the versioned prompts behind reading a reply, yours to read and yours to tune.

Your channels

Text, email, voice or the app you already use; which is tried first, which is the fallback, and whether a voice reminder is used at all.

Your consent rule

Who may be contacted and on which channel, where that permission is recorded, what counts as an opt-out, and which number or address the messages come from.

Your rebooking rules

Which slots may be offered and to whom, how close to the appointment a change is still allowed, how many choices to show, and what happens to the slot somebody declines.

Your appointment types

Which types are reminded at all, which are never messaged, and which always go to a person instead — a first visit and a routine recall often want different answers.

Your hand-over

Who receives the replies the bot will not act on, in which inbox or queue, and how quickly somebody is expected to have looked.

Your systems

Which tool holds the book, whether we read and write it by API or by a scheduled export, and where the message history has to end up.

Transparent, and configurable by you

The dark band below is what the system will not do. This is what you can see and change — and it is how we build everything, not something invented for this page.

You see what it did, and why

Every message is on the record with its template and version, and every reply shows how it was read and why. When a reply went to a person instead of being acted on, the reason it was handed over is written down next to it.

Versioned prompts you can tune yourself

The prompts behind reading a reply are yours to read and yours to tune. A change is a reviewable diff, replayed against replies you have already handled by hand so you can see exactly what it would have done differently, and rolled back the same way.

A human in the loop wherever judgment actually matters

Confirming, reminding and offering a slot you opened are automatic. Cancelling, charging, chasing a miss and answering a question are a person's — that boundary is a property of the system, not a promise about how carefully we'll use it.

Every run logged, along with what it cost

One row per message and one per reply: the appointment, the channel, the template version, the model and prompt version where a model was used, the classification and what happened next — plus token usage and cost for that run.

It can be configured to run in any business. The cadence and the rules are yours: your types, your quiet hours, your ceiling, your slots, your words. The appointment wiring on this page is one configuration of it, not the limit of it.

Works with your tools

This runs on top of the booking system and the phone number you already use. We're not asking your customers to install anything.

• A practice management system
• An online booking tool
• Google or Outlook calendar
• A field service or dispatch tool
• Toast or Square
• A spreadsheet you control
• SMS from your own number
• Email from your own domain
• A VoIP phone system
• Your website's booking form

Works well for

• Dental practices — Recalls and treatment visits are booked months ahead, and the reminder is the only thing holding the slot.
• Veterinary clinics — Vaccination and follow-up visits are booked far ahead, and the reminder has to reach the owner, not the patient.
• Salons and nail studios — Short appointments run back to back, so one empty chair is the whole hour and the next person is already waiting.
• Physical therapy and chiropractic clinics — A course of visits is booked as a block, and one missed session pushes the rest of the plan out.
• Tax and accounting practices — Appointments cluster into a season, and a missed one is hard to refill while the deadline stays where it is.

How a rollout works

Nothing goes to a customer on day one. The first milestone is the messages the bot would have sent for a week you have already worked, side by side with what your front desk actually sent.

1. 01Discovery

   The quote call, then a working session. We go through where the book lives, your appointment types, the cadence you want for each, your quiet hours and ceiling, how consent is recorded today, which slots may be offered and who picks up a hand-over. You get the message set, the reply classes, the exception list and a written scope.

2. 02Dry run

   The bot reads a stretch of your book and drafts every message it would have sent, with the timing. Nothing is delivered to anybody. You read them, we cut the ones you would not have sent, and the ceiling is easiest to cut at this step. This is where your own numbers come from.

3. 03Assisted mode

   Live on one appointment type or one location first. Messages go out on your cadence; every reply the bot will not act on lands in a person's queue with its reason. Your team stops typing confirmations and starts reading exceptions. The control doesn't change.

4. 04Optional wider rollout

   Optional. Once a type has run clean for a while, more types and locations come on, with every hard stop in force, the ceiling unchanged and the kill switch. Plenty of teams run one type forever, and that's a good place to stop.

What we don't claim

• No numbers we haven't earned. No no-show reduction, no confirmation rate, no revenue recovered. We haven't messaged your customers. Every figure you see on pages like this one belongs to somebody else's marketing; yours come out of the dry run and the first type you run live.
• No customer claims on this page. No names, no case studies, no logos.
• More reminders is not the product. Whether a reminder is welcome or irritating is a real argument, and it is not one a vendor should settle for you. The design gives you a ceiling and an opt-out that cannot be overridden, and the dry run writes out every message before any of them is sent, which is the easiest moment to cut one.
• It does not make your consent somebody else's problem. Who you may message, on which channel, and what you had to tell them first are yours to decide and yours to be able to show. We build the opt-out, the quiet hours and the record — we will not tell you that compliance is handled.
• Fully autonomous messaging isn't on offer. A reply that is not a plain answer, a question, a complaint and anything that reads as an emergency all go to a person. That's deliberate, and it's why the boundary is written down.

Questions people ask

Can AI send our appointment confirmations and reminders?

Yes - and the part that matters is what it does with the answer. An agent can read the appointments already on your book, send the confirmation when the booking is made, remind on the cadence you set for that appointment type, and read what comes back. A plain yes is logged, a plain no frees the slot under your rule, and a request for a different time is answered with numbered slots you have opened. Everything else - a question, a complaint, anything that reads as an emergency - goes to a person with the message attached.

How many reminders will it send?

As many as you set, and never more. The number, the timing and which one is the last are configuration per appointment type, there is a hard ceiling on how many messages one appointment may ever generate, and quiet hours are enforced in code. The dry run writes out every message the bot would have sent, with its timing, before any of them goes anywhere - which is the easiest moment to decide that a cadence is one message too long.

What happens when somebody replies with something other than yes or no?

It goes to a person. The bot classifies a reply as confirmed, declined, wanting a different time, a question, an opt-out, or something needing a person, and it only acts on the first three. A question about what a visit will cost, a complaint, a request to change something other than the time, or anything that reads as distress is handed over with the thread, not answered on a guess.

Can it reschedule somebody by itself?

Inside your rules, yes. It offers two or three numbered slots drawn from the openings your rules allow for that appointment type and that staff member, within the window you allow a change in, and if they reply with a number it holds that slot and writes the change back to the book. What it will not do is negotiate in free text, offer a slot you have not opened, or move an appointment nobody asked to move.

Does it call people, or only text?

Whichever channels you turn on - text, email, the app you already use, or a voice reminder - with one tried first and another as a fallback. Many teams never switch voice on, and that is a reasonable choice. What it never does is answer your incoming phone: a call nobody picked up is a different job with a different design.

How do opt-outs and consent work?

Who may be contacted, on which channel, and where that permission is recorded are yours to define, and the bot only ever messages people your rule says it may. An opt-out cancels every scheduled message to that number and address and blocks any future one, in code, with no model judgment involved. We build the opt-out, the quiet hours and the record of what was sent; what you are required to tell people is your decision and your responsibility, and we will not pretend otherwise.

Will it chase a no-show or charge a fee?

No. A missed appointment is recorded, messaging stops, and it goes on a person's list. The bot holds no card, takes no deposit and applies no fee - and it does not ring somebody the minute they miss an appointment, because that is the moment they feel worst about it and a robot is the wrong thing to arrive. Whether you charge, and how you ask somebody back, is a human decision about a relationship.

Does it work with the booking system we already have?

Yes, for most of them. Appointments are read from wherever the book lives - a practice management system, an online booking tool, a shared calendar, a field service or dispatch tool, a point-of-sale system like Toast or Square, or a spreadsheet you control - and an agreed change is written back to the same place. Messages go from your own number and domain so replies land where your team already looks. Naming a system means we can read or write it, not that we are partnered with it.

Is there a record of what was sent to whom?

Yes. Every message records the appointment, the channel, the time, the template and its version, and every reply records how it was classified, which model and prompt version read it, and what happened next - including the reason anything was handed to a person. Retention is yours to set, the reason for a visit can be excluded from the messages entirely, and nothing is used to train anything.