Define the bookable inventory.
Start with the services a customer may book. For each service, record its duration, eligible staff, location or meeting channel, preparation time and the minimum notice required. Add the hours when it may be offered and any dates that must remain unavailable.
Buffers deserve their own rule. A 45-minute appointment may occupy 60 minutes when setup and reset time are included. If travel is involved, define a realistic travel window instead of treating every open space as usable capacity.
Collect the minimum details before offering a time.
List the information required to identify the right service and resource. This may include the customer's name, contact method, purpose, preferred location, time zone and any eligibility condition. Avoid collecting details that do not affect the booking.
Time zones should be explicit. The scheduling system should store one authoritative time zone and show the customer the time zone used in every offer and confirmation. Relative phrases such as “tomorrow morning” should be converted into a full date, time and time zone before confirmation.
Offer slots without promising too early.
An available slot is not a confirmed appointment. Decide how long an offered time may be held, whether several customers may see the same option and what happens if availability changes before acceptance. The final availability check should happen immediately before the booking is created.
Write a confirmation template that includes the service, date, time, time zone, location or joining method, preparation instructions and a clear way to change or cancel. The customer-facing confirmation and the calendar record should describe the same appointment.
Set cancellation, rescheduling and no-show rules.
Define the cancellation window, any fee or deposit policy, who may approve an exception and which times may be offered after a change. Never let automation move another customer's booking simply to make a requested slot available. If the rules do not cover a request, route it to a person instead of inventing an answer.
Reminders also need consent, timing and channel rules. Decide how many reminders are appropriate, what information they contain and how a customer can stop them. These choices should reflect the business's local obligations and customer expectations.
Prepare an exception list before launch.
- Urgent or same-day requests
- Group bookings or linked appointments
- Customers who need accessibility support
- Staff absence after confirmation
- Conflicting time zones or daylight-saving changes
- Requests outside the published service list
Give each exception a named human owner and a response target. Test normal booking, rescheduling, cancellation and at least three exceptions before allowing any automated scheduler to speak with real customers.
Check Kai's current status separately.
The current roles page shows what is available for onboarding now. The capabilities page explains live channels and controls, and Kai's operating profile distinguishes current work from planned roles. Until Reception is released, do not assume a Kai calendar connection or appointment workflow exists for your business.
For the control layer that any future scheduler should include, read why every AI agent needs a stop button and the guide to writing escalation rules.