How it works
Planning is implicit in modern agentic frameworks: the LLM decides what to do next given the goal and current state. Some agents make planning explicit by generating a step-by-step plan first, then executing each step. Plan-and-execute is more reliable on complex tasks; direct planning is faster on simple ones.
Example
An agent asked to 'organise my inbox and respond to urgent items' might first plan: (1) categorise emails by urgency, (2) draft responses to urgent ones, (3) archive non-actionable ones. Then execute each step, replanning if a step reveals new information.