Build a repeatable set of fictional intake tests

The operational decision

Check the core inquiry path consistently after configuration changes. A meaningful intake test starts with an expected staff outcome. Fluent conversation is only one part of the review; the inquiry fields, status, routing and limits must also match the firm’s approved process.

Prepare the workflow

Write normal cases for category, contact preference, consultation request and staff handoff; define the exact record and status expected.

Example wording for approval

“Use a fictional caller requesting an estate-planning consultation next week. The expected result is a brief inquiry awaiting firm review, not an accepted matter.”

Worked example

Create fictional fixtures for a normal estate-planning request, an unknown practice category, a preferred callback window and a failed staff route. Write the expected fields and states before running them. After a prompt edit, compare the new records with those expectations. A different conversational style is acceptable only if the approved facts and workflow outcomes remain correct. Keep the fixtures synthetic so a regression test can be repeated safely without exposing or altering real inquiries.

Handoff record

Record these task-specific fields: scenario identifier, expected fields, expected status, reviewer.

Acceptance check

Run the same fixture after a prompt update. Every required field and status should remain correct even if the spoken wording changes slightly.

Related guides