Glossary

Human-in-the-Loop

A setup where an AI agent handles most of a task but hands off to a person for anything it shouldn’t decide alone.

Human-in-the-loop means an AI agent is designed to know its limits. For routine requests, it acts on its own. For anything sensitive, unusual, or high-stakes, it escalates to a person instead of guessing. AI Abstraction builds these escalation rules into every agent so nothing important gets handled by AI alone without a reason to trust it.

See every term in the glossary