An underwriting multi-agent workflow must run credit, property, and compliance checks that do not depend on each other’s outputs, then merge results. Latency SLOs are tight. Which orchestration approach is most appropriate?
Select an answer to reveal the explanation.
Short Explanation
If the credit, property, and compliance checks don't feed each other, don't make them wait in line. Run the three specialist agents in parallel, then join their results at a merge step before the decision agent. That cuts wall-clock time and protects tight latency SLOs.
Full Explanation
The correct approach is parallel multi-agent orchestration: run independent specialists concurrently, then join outputs before downstream decisioning. This improves task duration when dependencies allow, matching checks with no data dependencies. The strictly sequential approach is wrong because unnecessary sequencing increases latency without a correctness benefit. The random single-specialist approach is wrong because omitting required checks fails underwriting completeness, even if it saves cost. The manual multi-chat approach is wrong because separate user-invoked chats are not a controlled multi-agent solution. Exam tip: no inter-check dependencies means fan-out/fan-in parallelism.