Testing one successful GoHighLevel booking does not establish that multi-calendar routing works. The difficult cases are the calls that could fit several calendars, change direction, or reach a correctly selected calendar with no suitable availability.
Build a routing test matrix before collecting a set of pleasant demo recordings.
What should a multi-calendar test verify?
Check the intended appointment type, eligible calendar, offered time, saved record, and fallback. Those are separate assertions. A correct spoken department name does not prove the right calendar identifier was used.
HighLevel's Voice AI booking documentation currently supports multiple eligible calendars. Verify the specific configuration in use rather than transferring assumptions from a different product channel or an older setup.
Start with a small, labelled scenario set
For a fictional agency client, define three calendars: initial enquiry, technical consultation, and existing-customer support.
| Caller scenario | Expected result |
|---|---|
| Explicit new enquiry | Initial-enquiry calendar |
| Existing customer needs help | Support calendar |
| Unclear request | Clarification before calendar selection |
| Caller changes intent | Re-evaluate using the final request |
| Correct calendar has no slot | Approved fallback, not a different service |
Add examples using customer language. A caller might say “something isn't working” rather than use the exact support-calendar label.
Keep the expected result independent of the agent's explanation. Otherwise, the evaluator can accept an incorrect calendar because the generated reasoning sounds plausible.
Inspect configuration and output together
Capture descriptions, routing cues, durations, buffers, availability rules, and notification settings as part of the test version. When one changes, you need to know which earlier results may no longer apply.
Use fictional contact details. Keep the appointment reference and final calendar reference with the test record. For an unavailable-calendar case, verify that no unwanted appointment was created elsewhere.
Booking notifications and post-call summaries have different purposes. Confirm the customer receives the intended appointment message rather than assuming a post-call email proves the scheduling flow completed.
Add negative tests before expanding the client rollout
Test missing information, contradictory requests, a holiday, staff unavailability, and the absence of a suitable calendar. If the fallback creates a review task, verify its owner and callback details.
This matrix takes longer to prepare than a single demonstration. It gives the agency a useful regression check when client calendars change, which is often where an initially working configuration starts drifting.
For diagnosis of an existing routing error, use the wrong-calendar guide. This article is the acceptance and regression layer around that configuration.
Multi-calendar testing questions
Does one call per calendar provide enough coverage?
It checks the obvious paths. Add ambiguous, corrected, and unavailable cases.
Should the test use the agent's own explanation as its verdict?
No. Compare actual outputs with independently defined expectations.
What should the client receive?
A versioned matrix with expected and actual outcomes, including failures. Use it when reviewing readiness.