Question 11 — AB-620 Practice Exam 3
A Copilot Studio agent flow starts a document-processing operation that can take several hours because later stages may depend on human action. The conversation must continue promptly, preserve a reference that can be used to obtain the final result, and comply with agent-flow execution limits. How should the flow be designed? (Select one answer)
Keep all processing before Respond to the agent and allow the Action node to wait for up to 30 days
Return a tracking reference through Respond to the agent within 100 seconds and continue later actions asynchronously for up to 30 days
Enable Asynchronous response and delay Respond to the agent until the human-dependent processing has finished
Remove Respond to the agent and let the topic continue while the same Action node waits for the final result
Correct answers
Return a tracking reference through Respond to the agent within 100 seconds and continue later actions asynchronously for up to 30 days
Explanations
Correct:
Return a tracking reference through Respond to the agent within 100 seconds and continue later actions asynchronously for up to 30 days
An agent flow must respond to the topic within 100 seconds. For a long-running process, the flow can start the background work and return a job ID or tracking reference through Respond to the agent before that limit expires. Actions after the response boundary can continue asynchronously for up to Power Automate's 30-day flow-duration limit.
Wrong:
Keep all processing before Respond to the agent and allow the Action node to wait for up to 30 days
The topic pauses at the Action node only while waiting for the synchronous flow response, and that response must arrive within 100 seconds. The 30-day duration applies to asynchronous processing that can continue after the response boundary.
Enable Asynchronous response and delay Respond to the agent until the human-dependent processing has finished
The required callable configuration uses Asynchronous response set to Off so the topic can receive the declared outputs synchronously. Delaying the response for hours would exceed the 100-second response limit.
Remove Respond to the agent and let the topic continue while the same Action node waits for the final result
Respond to the agent is required for returning values to the topic and establishing the synchronous response boundary. Removing it does not provide a valid mechanism for the topic to receive a tracking reference and continue.