An airline has been putting agents into production with no record of the actions each agent is allowed to take. What should governance require before the next agent goes live?
Select an answer to reveal the explanation.
Short Explanation
A crew does not leave the gate without a list of what that flight is cleared to do. You require a record of the agent's allowed actions before production. A brand-voice guide does not write that record.
Full Explanation
Agent governance includes knowing what the agent is allowed to do before it can act on customers. The airline is shipping agents with no such record, so the gate is to finish that record of allowed actions before production release. A brand-voice guide describes tone, and a checklist that accepts it in place of the action list approves the wrong artifact. A satisfaction score gathered after go-live tells you how a trip felt, and it arrives too late to be the release condition. A second production copy that people edit in place breaks the line between build and production and can still omit what the agent is permitted to do. The record should name each action, the system it calls, and the condition that stops it, and release should fail closed when that record is missing. Writing the record after go-live leaves a customer-facing agent with an unknown boundary. Exam caveat: the record is the permission boundary, and a chat score does not replace it even when the agent has a conversational channel. Operational check: block the release until the allowed-action record is present, and confirm a reviewer can name each action before the agent is published.