Decide: Choosing the Next Action

"Decide" is the middle step: choosing the next action based on the goal and whatever was just perceived. It's not a fixed script that runs the same way every time — the next action depends on what the agent actually found when it looked.

Say a search comes back empty. A fixed script with no decision step would just move on to the next line, whether or not that still makes sense. An agent's decide step can instead choose to try a different search term, because that's what the current state actually calls for.

This is what separates an agent from plain automation: automation runs the same steps regardless of what happens along the way, while an agent's decisions branch based on what it perceives. That doesn't make the decision magic — it's still the same kind of reasoning covered elsewhere in this app, just applied fresh at every cycle instead of once.

What makes the "decide" step different from a fixed automation script?

A fixed automation script runs the same steps regardless of what happens; an agent's decisions ____ based on what it perceives.

The "decide" step in an agent loop always produces the exact same next action no matter what was just perceived.