Agents Make a Plan BEFORE They Start

The best AI agents write a plan first. Then they do the steps one at a time — just like you with a project.

Some examples

• Step 1: search for facts. Step 2: write a draft. Step 3: check it.
• Plans help agents stay on track.
• If something breaks, the plan helps them recover.
• Plan-first agents are smarter than 'just do it' agents.

Try it!

Ask AI: 'Make a 3-step plan to bake cookies, then explain each step.' Notice how the plan comes first.

Why planning before acting leads to better results

Have you ever started building something without reading the instructions first, and then had to tear it apart halfway through? That's what happens when AI agents skip the planning step. A planning-first agent looks at the entire task before taking any action. It asks itself: 'What's the goal? What are the steps? What could go wrong? What order makes the most sense?' Only after answering those questions does it start acting. This process is called decomposition — breaking a big goal into smaller, ordered steps. An agent asked to 'plan a school fundraiser' might decompose it into: research similar events, pick a date, draft a budget, create a sign-up form, and write announcements. Each step feeds into the next. Skipping decomposition leads to agents that get halfway through a task, realize they've forgotten a key step, and have to backtrack — wasting time and sometimes causing errors that are hard to undo.

• Decomposition: breaking a big task into small ordered steps before starting
• Dependency order: figuring out which steps must happen before others can begin
• Risk preview: spotting which steps are most likely to go wrong before they do
• Plan approval: showing the plan to a human before executing any irreversible actions