Your n8n agent doesn't know it's wrong

The exact clicks

1. Open the n8n workflow with your AI Agent in it, and find where its output currently connects straight into the next step (the one that sends or saves the result).
2. In the node panel on the left, search "If" and drag an IF node — n8n's branching block, one input and two outputs ("true" / "false") based on a condition you set — onto the canvas between the Agent and that next step.
3. Delete the direct wire from the Agent to the next step, then wire the Agent's output into the IF node instead.
4. Set the IF node's condition to check whatever tells you the answer might be shaky — a confidence field the Agent already outputs, or a plain rule like "is this field empty."
5. Wire the IF node's "true" output (the low-confidence branch) to a Slack message, an email node, or wherever a person will actually see it — not back into the original flow.
6. Wire the "false" output (the confident branch) into the same next step that was connected before.
7. Run the workflow once with a case you expect to be low-confidence, and confirm it lands with a person instead of going straight out.

What changes

• Without the check, the workflow ships the wrong result exactly the way it ships a right one — nothing distinguishes them downstream.
• By the time a human notices, the output has already gone out.
• One IF node before the output is enough — this is not a rebuild of the workflow.

What it will not do

• This only catches what the agent itself can flag as uncertain — it does not catch a confidently wrong answer.
• "Route to a human" only helps if someone actually reviews that queue — an unread inbox is the same as no check.