The AI appointment booking guide: the slot has to be real.

What happens when an agent books an appointment

1. Understand the request

   What service, for whom, roughly when. “Sometime next week, preferably a morning” has to become a query against a diary.

2. Check live availability

   The agent reads the connected calendar — Google Calendar or Microsoft Outlook — applying your service durations, buffers and working hours to find slots that are genuinely bookable.

3. Offer a short list

   Two or three options, not a recital of the whole week. On a phone call especially, the number of choices offered determines whether the caller decides or defers.

4. Collect what you need

   Name, contact number, service, and any question your team needs answered before the appointment.

5. Write and confirm

   The event is created in the calendar and a confirmation goes to the customer — spoken on a call, and in writing over WhatsApp or email where you have the details.

The double-booking problem

Two customers can ask for the same slot within seconds of each other on two different channels. Any system that reads availability, chats for two minutes and then writes without re-checking will eventually book them both into the same chair.

The protections worth asking about are re-validating the slot at the moment of writing, respecting the calendar’s own conflict rules rather than a cached copy, and keeping buffers between appointments so a small overrun does not cascade through the day.

• Ask how availability is read: live from the calendar, or from a synchronised copy that can drift?
• Ask what happens if the slot disappears between the offer and the confirmation.
• Check that entries your team adds directly to the calendar are respected immediately.

The rules that make a diary work

Most booking disappointments are configuration, not intelligence. The agent will follow whatever rules exist — so the rules have to exist.

• Service durations, including the ones that vary by customer type.
• Buffers before and after, plus any cleaning or setup time.
• Working hours per person and per location, plus holidays and one-off closures.
• Notice periods: how soon is too soon to book, and how far ahead you will take bookings.
• Who can be booked with whom — services that require a particular person or room.
• What requires a human: first consultations, complex cases, anything you would want to price by phone.

Rescheduling, cancellations and no-shows

Bookings are not the hard part; changes are. A customer who can cancel at 9pm by replying to a WhatsApp message is a customer who tells you rather than not turning up, which is worth more than the booking automation itself.

Reminders do most of the work here. A confirmation at the time of booking and a reminder before the appointment, both on the channel the customer used, reliably reduce the number of people who simply forget. Make sure the reminder can be replied to — a reminder the customer cannot respond to only tells you about the no-show after it has happened.

One diary, three channels

The point of booking inside a conversational platform rather than sending a scheduling link is that the customer never leaves the conversation. On the phone they never hear “I’ll text you a link”. On WhatsApp they never open a browser. On your website they book in the chat window they were already using.

That only works if the phone agent, the chat widget and the WhatsApp agent read and write the same calendar with the same rules — otherwise you have three booking systems and a reconciliation problem.

How to tell whether it is working

• Appointments created by the agent versus enquiries that ended in a callback promise.
• Out-of-hours bookings — the ones that would previously have been voicemail.
• Reschedules handled without a staff member touching them.
• No-show rate before and after reminders were switched on.
• Transcripts where booking was attempted and failed — the fastest source of rule fixes.