32dots HEIDELBERG AI
Session 4 medium

Keep a human in the loop

LESSONLesson 4 · ~20 min

🎯Goal. Pause an agent before a consequential step so a person can approve, reject, or correct it.

▶ Try this prompt

Before sending the drafted message, ask the user to Proceed or Reject, and let them give feedback on the last assistant message.

Flowise has built-in human-in-the-loop review — add it to a flow so the run stops for sign-off instead of acting unsupervised.

  1. 1In your Agentflow, add a human-in-the-loop review step before the action you don't want fully automated (e.g. sending an email or finalising an answer).
  2. 2Run the flow: it pauses and shows the proposed output with Proceed and Reject controls, plus a box to give feedback on the last assistant message.
  3. 3Approve to continue, or reject with a note so the agent revises. This is how you ship agents you can trust on real tasks.
A Flowise human-in-the-loop review card asking 'Would you like to proceed...?' with green Proceed and red Reject buttons
Flowise's human-in-the-loop review: the agent pauses and asks a person to Proceed or Reject — and to give feedback on the last message — before acting. Source: https://flowiseai.com

You'll see. The run pausing at a review card with Proceed / Reject buttons and a feedback box — the agent waits for your decision before continuing.

💳Cost. Review adds no model cost while it waits; you only spend tokens when the agent resumes or revises after your feedback.

💡Takeaway. Human-in-the-loop turns an autonomous agent into a supervised one — a person signs off before anything irreversible happens.