agent-team-orchestration-patterns

Multi-agent orchestration uses narrow role design and idle-agent re-dispatch to prevent duplicative work. Agents share a living scratchpad to log decisions and cross-reference findings, ensuring coherent execution across all sub-teams.

Iterative critique loops are structured by assigning specific roles like Critic and Researcher to separate agents. The team uses a living scratchpad to log open questions and rules, allowing agents to cross-reference outputs and refine the plan iteratively.

You need multi-agent orchestration when a planning task requires three or more agents, cross-source data synthesis, and iterative critique. It is applicable across domains like calendar planning, Notion research, and web gathering to maintain alignment.

Design agent roles by assigning narrow responsibilities with explicit outputs, such as a Data Gatherer, Researcher, Time Validator, and Critic. This role design ensures agents cross-reference shared scratchpad logs and maintain alignment during cross-source synthesis.

Yes, multi-agent orchestration applies to both calendar planning and web research. It coordinates agents to gather data, validate times, and synthesize findings iteratively, maintaining cross-referencing and alignment across diverse task domains.

Teams drift without clear orchestration patterns, leading to missed insights and duplicated effort. Implementing role design, explicit agent communications guidance, and a shared scratchpad for cross-referencing prevents this drift and maintains alignment.