Automated appointment reminder calls that cut no-shows

Every business that books time has the same leak. A client forgets, a patient double-books, a customer means to cancel and never does. The slot sits empty, the person who could have filled it went elsewhere, and the cost is invisible because it never appears on an invoice.

Reminder calls are one of the few fixes with solid evidence behind them. What the evidence also shows, and what most reminder software glosses over, is that the type of call matters.

What the research says about reminder calls

The randomised trial. Parikh and colleagues randomly assigned patients at an academic outpatient practice in spring 2007 to one of three groups: a reminder call from clinic staff three days before (3,266 patients), an automated reminder three days before (3,219), or no reminder (3,350). Published in the American Journal of Medicine in 2010, the results were:

All the differences were statistically significant. Both reminder groups also had higher cancellation rates than the no-reminder group, which sounds bad and is actually good: a cancellation three days out is a slot you can refill, while a no-show is a slot you cannot.

The systematic review. Hasvold and Wootton reviewed 29 studies of telephone and SMS reminders for hospital appointments (Journal of Telemedicine and Telecare, 2011). Across them, reminders reduced non-attendance by 34% of the baseline rate on average. Manual phone calls did better than automated ones (39% against 29%). And they found no difference between reminders sent the day before and the week before. The average cost per reminder across SMS, automated and manual calls was about €0.41.

Both studies come from healthcare, because that is where no-shows are measured most carefully. The mechanism, people forgetting or failing to cancel, is the same at a salon, a garage or a law firm.

Why staff calls beat recorded ones

The research does not say exactly why, but the design of the two calls suggests the answer. A recorded reminder is one-way: it plays the appointment details and perhaps offers "press 1 to confirm". A staff call is a conversation. The receptionist hears "oh, Thursday... actually, I'm away that week" and moves the booking there and then. The patient who was going to no-show becomes a rescheduled patient, and the slot goes back on sale.

That is the gap a two-way AI reminder call is designed to close. It is not a recording; it listens. When the person says they cannot make it, it can offer another time. When they say "I'll call back to rebook", it records the cancellation so the slot can be offered to someone else. Whether it fully matches a good receptionist is something you should measure on your own bookings, which is easy to do (see the test plan below). What is clear from the research is which direction to push: more conversation, not more recordings.

What a good reminder call sounds like

Short, clear and finished within a minute or two. Here is an example for a fictional physiotherapy clinic:

Milo: Hi, is that Sarah? This is Milo, an AI assistant calling from Harbour Physio. I'm calling about your appointment on Thursday at 2:30 with Tom. Can you still make it?

Sarah: Oh, Thursday. I've got a work thing. Is there anything on Friday?

Milo: Let me check. I can do Friday at 10:00 or 4:15. Would either of those work?

Sarah: 4:15 is perfect.

Milo: Done, you're booked for Friday at 4:15 with Tom, and Thursday is cancelled. Anything else I can help with?

Four rules make the difference:

• Say it is an AI, straight away. Dial Milo does this by default. People mind discovering it later far more than hearing it up front.
• Confirm it is the right person before giving details. Especially for medical or legal appointments, where the appointment itself is private information.
• Ask one question: can you still make it? Not "press 1 to confirm". Then deal with the answer.
• Keep sensitive details out. Day, time, place and who the appointment is with. Not what it is for.

Setting up automated appointment reminder calls

1. 1

   Export tomorrow's, or the next three days', bookings

   A CSV with phone number in international format, first name, appointment date, time, and anything else the call should mention, such as the practitioner or location. Extra columns become placeholders such as {appointment_time} in the script.

2. 2

   Create a manual campaign

   Write a short script: the opening line, the reason for the call, what to do if they want to reschedule or cancel, and what to say if they ask a question you have not covered. See writing a script.

3. 3

   Connect your calendar

   With Google Calendar connected, Milo can check real free slots and book a new time during the call, instead of just taking a message. See connecting Google Calendar.

4. 4

   Set calling rules

   Calling hours in each person's local time (10:00–20:00 by default), two attempts a few hours apart, and an end date before the appointment so nobody is reminded after the fact. See calling hours and contact limits.

5. 5

   Test it on yourself

   Put your own number in the list, or use a test call, and play the awkward patient: wrong day, wants to cancel, asks about parking.

6. 6

   Upload and let it run

   Every call ends with an outcome, transcript and summary. Anyone who asked for a callback is listed with the time they asked for, and voicemails and no-answers are retried.

If someone misses the call and rings the number back, Milo recognises them, thanks them for calling back and picks up the same conversation.

Timing: when to call

The review found no difference between a reminder the day before and one the week before, so choose timing for operational reasons rather than psychological ones:

• Two to three days ahead gives you time to refill a cancelled slot. The randomised trial called three days ahead.
• The day before is too late to rebook for many businesses, but fine if you run a waiting list.
• One call, not three. Stacking reminders annoys people into opting out. One call, plus a text or email on the day if you already send them, is plenty.
• Inside sensible hours. Dial Milo calls each person in their own time zone, inside the window you set.

What it costs

On Dial Milo, an outbound call is the phone line (priced for the exact number dialled, billed per started minute) plus 9.57¢ per minute of AI talk time.

Worked example (illustrative figures). Say a dental practice has 400 appointments a month, with US patients, and today 12% are no-shows (48 empty slots).

• Assume 260 patients answer and the call lasts 90 seconds. Line: 260 × 2 started minutes × 4¢ = $20.80. AI: 260 × 1.5 × 9.57¢ = $37.32.
• Assume 200 voicemail attempts across the retries at about 5¢ each: $10.
• Total: about $68 a month, or 17¢ per booked appointment.

If the reminders turn just three no-shows into kept or refilled appointments, they have paid for themselves several times over at almost any appointment value. Price your own numbers on the rates page and model the month with the spend calculator.

Turning cancellations into filled slots

A reminder call that produces a cancellation three days ahead has done its job, but only if someone then fills the slot. Plan for it before you start:

• Keep a short waiting list of people who asked for an earlier appointment. A cancelled Thursday slot is worth more offered to them than left empty.
• Let the AI reschedule on the call. With a connected calendar, many cancellations become moves rather than losses.
• Review cancellations daily. The call outcomes and summaries show who cancelled and why, so the front desk can act the same morning.

Common mistakes with reminder calls

• Calling too late to act. A reminder the evening before tells you about a no-show you can no longer prevent or refill.
• Selling on a reminder call. "While I have you, we have a whitening offer" turns a service into a marketing call, legally and in the customer's mind.
• Saying too much. The reason for a medical, legal or financial appointment does not belong in a call that might reach the wrong person.
• Calling numbers that changed. Ask patients and clients to confirm their mobile when they book; a stale number makes every other part of the system pointless.
• Never listening to the calls. Read twenty transcripts in the first week. You will find at least one question people keep asking that the script does not answer.

Consent and the rules

Reminder calls are informational, not marketing, and that matters legally. But automated calls still have rules. The FCC's consumer guide says FCC rules require oral or written consent before an autodialled or prerecorded call to a mobile number, and that AI-generated voice calls are illegal unless the consumer has agreed or the caller is exempt. Since the FCC's February 2024 ruling, AI-generated voices count as artificial voices under the TCPA.

The practical answer is simple: when someone books, ask whether you can call or text to remind them, and record the answer. Keep marketing out of the reminder call entirely; the moment it starts selling, it is a different kind of call. And honour "please don't call me" immediately. Dial Milo records it as an opt-out and keeps that number off future campaigns.

How to test whether it works for you

Do not take the research, or us, on faith. For one month, split your bookings in half by appointment date or by alternating patients. Call one half with AI reminders and leave the other half on whatever you do today. Compare three numbers: no-show rate, late cancellation rate, and how many cancelled slots you refilled. A month of a few hundred appointments is usually enough to see the direction.

For the inbound side of booking, see answering service appointment scheduling; for the full picture of outbound campaigns, see our guide to AI outbound calling. When you are ready, create an account and hear Milo call you in the browser first.