Implement error handling in agent flows

Bundled into the existing Designing and Building Integrated AI Agent Solutions in Copilot Studio premium course — no separate purchase.

14-day money-back guarantee — no questions asked.

Included in this chapter:

Consideration	Run after branch	Try and Catch scopes	Retry policy	Terminate
What it changes	Which action executes next, based on the previous action's outcome	The unit that succeeds or fails, from one action to a whole block	How many times the action itself is attempted before it is called failed	Whether execution continues at all
Where you configure it	Run after settings on the action that follows	Run after settings on the Catch scope, pointing at the Try scope	Settings of the individual action being retried	An action you add on the failure path
The fault it is for	Any outcome you want a different next step for: failed, timed out, skipped, successful	A transaction whose steps are only meaningful together	Transient faults: momentary network loss, temporary unavailability, busy-service timeouts	A nontransient failure with no valid path forward
What it costs	One branch to maintain per outcome you handle	Granularity: a failed scope reports a single Failed status for the whole scope	Elapsed time inside the synchronous response window, plus more calls to a service that may be throttling	Everything after it, which is why cleanup goes first
What the agent gets	Whatever the branch you routed to maps into Respond to the agent	Whatever the Catch path maps, if the response action is reachable outside the scopes	Nothing different: retries are invisible to the agent unless they exhaust the window	Nothing, unless a response action ran before it

Cheat sheet

Unlock with Premium — includes all practice exams and the complete study guide.

Cheat sheet