# Human in the loop

A person takes part in an automated system's work, supervision or decisions.

On this site, a human in the loop is usually an approval step. [How do you stop an AI agent?](/research/stopping-an-ai-agent-approval-and-limits) found that n8n and LangGraph both let an agent pause before using a tool and wait for a person to approve or deny it. [CrewAI](/research/crewai-human-input-and-pricing) offers three ways to add the same check. [Guardrails](/research/ai-agent-security-and-guardrails) quotes OWASP's advice to put a person in front of privileged actions. Mackenzie's [startup story](/stories/gptboss-240k-ai-business) names the goal: an AI employee that doesn't need a person at every step.

IBM adds that in AI, it "means that humans are involved at some point in the AI workflow to ensure accuracy, safety, accountability or ethical decision-making" ([IBM Think](https://www.ibm.com/think/topics/human-in-the-loop)).

---

Published 2026-11-05. Web page: https://www.mackenziebowes.com/reference/glossary/human-in-the-loop
