Planner-Executor Pattern
The planner-executor pattern splits an agent into two roles: a planner that decomposes a goal into steps, and an executor that runs each step — improving reliability on complex tasks.
Models often plan well in one pass and execute well step-by-step, but combining both in the same loop produces drift. The planner-executor split externalizes the plan as a checklist; the executor focuses narrowly on the current step. Failures isolate to either planning (wrong steps) or execution (right steps, wrong action), making debugging tractable.
Example
Goal: 'send weekly digest'. Planner outputs: 1) gather metrics, 2) identify highlights, 3) draft, 4) send for approval. Executor handles each step in isolation.
How OpenLabor uses it
OpenLabor missions use planner-executor decomposition under the hood. You see both the plan and the execution log.
Is this the same as 'chain of thought'?
Related but different. CoT is reasoning within one call; planner-executor splits planning and acting across separate calls (often separate agents).
Related: multi-agent, agentic-workflow.
AI Labor Glossary