How to measure your AI receptionist's real bookings

Take every appointment your AI receptionist counted as booked, find it in your own booking system, and check that the AI actually created it, that staff did not have to fix it, that it was not cancelled, and that the customer came. Then read a sample of the calls to confirm the appointment matches what the caller asked for. The share that passes is your audited share of bookings.

These are the checks of the Attify Audited Booking Standard (v0.6), all of them counted, as for a vendor that claims kept appointments. A vendor that claims appointments booked is audited only on who booked it, a new booking, a staff fix and the conversation read, with its cancellations and no-shows shown beside the claim (the claimed unit).

1. Make sure you can tell the AI's bookings apart#

Your booking system must record who created each appointment. That works when the AI books with its own credential: its own integration, user, API key or scheduling link. If the AI only takes a request and your staff enter the booking, the AI did not book it. Set this up before the period you measure.

2. Take a settled period#

30 days of first scheduled times, checked at least 48 hours after the last one, so every appointment has had time to be changed, kept or missed. Leave appointments more than 90 days ahead for a later check.

3. Check each appointment#

CheckFails whenWhere to look
Not booked by the AIIt is not in your system, is only a request or a hold, or staff created itCreated by, source (Calendly scheduling method api, Cal.com's AI event type, a PMS's creating user)
Not a new bookingThe AI moved or confirmed an existing appointment, or duplicated one the same dayAppointment history
Staff fixed itStaff changed the service, time, provider, location, party size or price within 48 hours, or cancelled and rebooked itWho changed it (Calendly canceler type, Cal.com who rescheduled)
CancelledCancelled by anyone and not rebooked for the same service within 14 daysStatus
No-showMarked no-show, or no recorded visit where your system records visitsCheck-in, completed or paid status; no-show marks

Then read a sample of the conversations behind the appointments that passed, and fail any that do not match what the caller agreed to.

4. Compare with your own staff#

Put your staff's bookings from the same period through the cancellation and no-show checks. No-show rates differ widely between sectors (US outpatient visits ran 6.2% to 7.9% in 2024, Epic Research; salons and spas 1% to 4%, Zenoti, 2026), so your own staff are the only fair benchmark. Split both by new and returning customers and by lead time.

Worked example#

Synthetic example. A dental practice; the AI receptionist counted 400 appointments as booked in 30 days.

Reason an appointment failed (first check)Appointments
Entered by staff from the AI's request, or not in the system20
Moved or confirmed an existing appointment, or a duplicate16
Staff changed it within 48 hours24
Cancelled and not rebooked within 14 days22
No-show18
Wrong booking, from a sample read of the remaining 300 (4.0%)12
Passed every check288

Audited share 288 ÷ 400 = 72.0%. At $400 a month flat, each audited booking cost $1.39 against $1.00 per claimed booking.

Booking rate#

If the vendor also reports a booking rate ("70% of calls booked"), ask for its denominator. A fair one is calls in which the caller asked for a new appointment, with calls you could not serve (no slot, a service you do not offer) kept in and shown as their own row. See booking rate.

Do it with Attify#

Connect Calendly, Cal.com, Google Calendar or Microsoft 365 on the buyer dashboard for the monthly audited share, your staff baseline, and the cost per audited booking. Practice-management systems that record visits (Open Dental, NexHealth and others) are not read yet.

Questions

Do AI receptionist bookings have more no-shows than staff bookings?

Attify knows of no published independent data on it. Measure it in your own practice: put the AI's appointments and your staff's appointments for the same period through the same checks, split by new and returning customers.

My vendor says it booked 400 appointments. How many were completed?

Find each of the 400 in your booking system by its id and read its status: booked by the AI or entered by staff from a request, changed by staff, cancelled, no-show, or a recorded visit.

Does a call marked 'successful' mean an appointment was booked?

No. On voice platforms such as Retell and Vapi, that flag is an LLM grading the call against a prompt the vendor or you wrote. Only your booking system shows the appointment.

Who can see my patients' call data if I connect it?

See Attify's security and privacy pages. Text is redacted on arrival and deleted after 30 days by default; scores and counts are kept.

Sources

  1. Attify. Attify Audited Booking Standard v0.6. Published 9 Nov 2026. Read 9 Nov 2026.
  2. Epic Research. Patient portal use associated with 21 million fewer visit no-shows in 2024. Published 1 Jul 2025. Read 8 Nov 2026.
  3. Zenoti. No-show revenue calculator (2026 Beauty and Wellness Benchmark Report). Read 8 Nov 2026.
  4. Retell AI. Post call analysis. Read 31 Oct 2026.
  5. Vapi. Call analysis. Read 31 Oct 2026.