Glossary · Agents and autonomy
Planning (AI agents)
Planning in AI agents is the preparation of steps an agent intends to follow toward a goal. It involves breaking down a request, identifying dependencies and choosing an order of action, while allowing the plan to change as new results become available.
Planning (AI agents): the explanation
Starting from your request, the agent proposes smaller tasks and identifies what must be known or completed before another step can begin. When organizing a trip, it might first check constraints, then compare routes and finally present options. The plan can be a simple description or a structured record tracked by software. Either way, its purpose is to guide execution and help establish whether any necessary part of the requested work has been missed.
Planning is a longstanding area of artificial intelligence, including methods that describe goals, possible actions and the conditions those actions require. Agents using language models can also generate plans in ordinary language. These approaches do not offer identical guarantees. A plausible description of steps does not establish that every action is possible, that the required tools are available or that all your constraints have been satisfied. Those questions still need appropriate checks during preparation and execution.
A useful plan can change when new information appears. An unavailable booking, a missing file or a newly stated preference may require a different route. The agent should revise its approach while retaining the original objective. Planning also needs to remain separate from authorization: including a purchase in a plan does not grant permission to make it. An excessively detailed plan can add unnecessary overhead when a task is straightforward and its steps are already clear.
A concrete example
You want to organize a training day for a community group. The agent proposes confirming attendance, checking room availability, preparing a program and drafting invitations. It waits for essential information before setting times. When you explain that one speaker is only available in the afternoon, it rearranges the sessions and presents the revised schedule before any invitations are sent to participants.
Why it matters
Planning helps you understand what an agent intends to do and catch a mistaken interpretation before execution. It is especially useful when steps depend on one another. When evaluating a tool, check whether its plan reflects your constraints, distinguishes intended work from completed work and explains significant changes instead of continuing with an approach that no longer fits the actual situation.
Common misconceptions
- A well-written plan does not prove feasibility: an agent can propose apparently sensible steps without having the tools, information or permissions required to carry them out in your particular situation and connected working environment.
- Changing a plan is not necessarily a failure: adapting to new information can be appropriate, provided the agent preserves your objective and explains changes that significantly affect the task, its scope or its consequences.
Related terms: agents and autonomy
- AI agent
- An AI agent is a system that uses a model to decide which steps to take toward a task.
- autonomous agent
- An autonomous agent is an AI agent that can pursue a goal and choose several actions without involving you at every step.
- multi-agent system
- A multi-agent system is a collection of agents that divide tasks or exchange information to pursue an objective.
- agent memory
- Agent memory is the set of mechanisms that lets an AI agent retain and retrieve information useful to its work.
- human in the loop
- Human in the loop is an arrangement in which a person participates at selected stages of an AI system’s work.
- tool calling
- Tool calling is a mechanism through which a model asks software to run a specific function, such as searching documents or checking a calendar.
- computer use
- Computer use is a capability that lets an AI agent interact with a graphical interface by clicking, typing or scrolling.
- coding agent
- A coding agent is software that combines an AI model with tools to perform programming tasks.
Where this site covers it
The term appears on 45 pages of this site. A few of them:
- Deploying AI agents in business: use cases, selection, pilots, governance, and measurement Guide
- Cursor reference guide: the AI code editor, modes, rules, pricing, and limits Guide
- AI Act: a compliance checklist for businesses using AI agents and tools Guide
- When will AGI arrive? Published forecasts, their methods, and why they differ Guide
- AGI risks: what researchers say, without alarmism Guide
- The limits of deep learning: what models struggle with, according to François Chollet and other researchers Guide
- Writing your company’s AI use policy: what to include, sample rules, and rollout Guide
- Claude Code reference guide: installation, subscriptions and API access, commands, CLAUDE.md, MCP, subagents, and best practices Guide
Same family
Go further
Frequently asked questions
Should every task show a plan?
Not every small request needs one. A visible plan becomes useful when work involves dependencies, takes substantial time or can change data. It gives you a chance to correct the approach before consequential actions begin.
Can the plan change?
Yes. Search results or new constraints may require revision. The system should distinguish ordinary adjustments from changes that exceed your original request and require consultation before it continues with a different scope or outcome.
Does a plan grant permission?
No. A plan describes proposed actions. Execution still depends on granted permissions and required approvals. Approving research or a draft does not automatically authorize a purchase, publication or message send that appears later in the plan.
Neutrality. No affiliate links. No vendor pays to be reviewed, rated or ranked. Reviews are directed by François Chollet, neutral and factual, and every data point is dated and sourced. How we score