Appointment requests answered. Bookings saved in your calendar.

An appointment request can arrive by phone, by email or through a web form. AI scheduling clarifies what the customer needs, offers only permitted slots, books the chosen one in your calendar and confirms it. The confirmation follows only after the calendar has saved the appointment. Everything the rules do not cover goes to a person.

The rule that matters most: a free slot in the calendar is not yet a reservation. Availability can change between offering a slot and saving it, because staff and other applications also write to the calendar. A booking flow is reliable when access, booking rules and the saved calendar entry fit together.

"Book a suitable appointment" is too vague. Each appointment type needs a duration, buffers, notice, a working window, a location and the people or rooms it requires. Holidays, local closures and time zones are written down too. The model does not guess these rules from general knowledge. Relative dates such as "next Friday" are resolved into a specific date and time and confirmed with the customer.

If a booking application such as Microsoft Bookings already controls capacity, it stays authoritative. A second process must not bypass its rules by writing directly to calendars.

With Outlook, the getSchedule operation in Microsoft Graph returns availability for users and resources, and the documentation lists personal Microsoft accounts as unsupported for that operation. The event creation reference documents the separate write step. Our guide to AI appointment booking with Outlook covers permissions, room resources and retries.

With Google Calendar, a FreeBusy query returns occupied intervals and does not reserve the proposed slot. The event is created with events.insert. Our guide to phone booking with Google Calendar shows the booking rules and a test worksheet, and n8n with Google Workspace connects forms, mail and calendars.

Two callers can choose the same slot while both conversations are still open. A recheck alone does not stop two requests that pass it at the same moment. A central booking layer therefore coordinates the writes for each slot, or a booking system with its own reservation control decides. Manual edits in the same calendar must be considered too. The second caller is offered other options and must accept one. The assistant never books a different time silently to make the call look successful.

A missing response after a write is neither success nor failure. The request goes into reconciliation, and the calendar is checked before another attempt, so that a retry does not create a second appointment. Rescheduling and cancellation need the right existing event and the agreed verification: a name or a phone number alone can match several appointments. Our article on identity checks and business actions explains that verification.

Some requests do not fit any rule: an unknown service, a pending room approval, a caller who wants a person. They go to an owned queue with contact details, the preferred appointment and the technical state. Our guides on human handover and callbacks and on turning calls into CRM tasks describe that queue.

1. List the appointment types you book most often.
2. Write down duration, buffers, notice, location and resources for each type.
3. Decide which calendar or booking system is authoritative.
4. Test two callers for one slot, an interrupted connection and a corrected email address.
5. Go live with an owned queue for uncertain requests.

Our AI appointment scheduling page describes how a first project with us runs.