Skip to main content
GET
Read a test on past tickets
Requires agent.read. Every text field of a ticket came from the ticket and is text, never markup. confidence.score is the retrieval score of the strongest source the reply cites, and 0 when the answer engine declined; it is not the model grading itself.
AI employees are in beta. Every route answers 404 unless your deployment has switched the beta on. See Test on past tickets. This has not yet run on a real Zendesk account.

Path

string
required
The test.

Response

object[]
Each with human_reply, proposed_reply, outcome, sources, confidence and verdict.
Last modified on November 7, 2026