Planning.

Planning in AI refers to the cognitive process where an agent breaks down a complex goal into a sequence of manageable steps or subgoals before executing actions. This capability distinguishes sophisticated AI systems from simple reactive models, enabling them to handle multi-step problems that require foresight and strategic thinking. Planning is fundamental to achieving complex objectives that cannot be solved through single, direct responses.

AI planning can occur through various mechanisms, from explicit step-by-step decomposition to implicit hierarchical reasoning within neural networks. Modern language models demonstrate planning through techniques like chain-of-thought reasoning, where they explicitly outline their approach before providing solutions. Some systems use tree search algorithms to explore different action sequences, while others employ learned planning representations that emerge from training on sequential decision-making tasks.

Effective planning significantly improves AI performance on complex tasks like coding, mathematical problem-solving, and multi-step reasoning challenges. However, planning introduces computational overhead and can sometimes lead to over-engineering simple problems that would benefit from direct approaches. A common misconception is that all AI reasoning involves explicit planning, when in fact many successful systems blend reactive and deliberative approaches depending on task complexity and time constraints.