You implement orchestration for a due-diligence multi-agent workflow: financial analysis, legal review, and market research can run independently, then a partner agent synthesizes findings. Which orchestration pattern should you implement?

Multi-agent orchestration patterns include parallel, sequential, hub-and-spoke, peer-to-peer, and orchestrator-subagent. The correct pattern uses parallel execution for independent workstreams plus a join/synthesis stage. The strict sequential chain that forces finance to finish before legal starts adds unnecessary latency when there is no data dependency. The peer-to-peer free-for-all with no synthesis step and no completion barrier lacks aggregation and quality control. The hub-and-spoke pattern where every specialist must call every other specialist for each sentence over-connects agents and increases cost/noise. In Azure/Agent Framework or LangGraph, model this as concurrent nodes with a barrier before the synthesize node. Exam tip: Choose parallel when tasks are independent; sequential when outputs feed inputs.