Human-in-the-Loop (HITL) architecture is a foundational control pattern for enterprise Artificial Intelligence where AI systems assist with analysis, recommendations, content generation, or task execution while designated people retain authority at critical decision points. As Generative AI and Agentic AI become more autonomous, enterprises require a systematic architecture for determining when human review is mandatory, how work is routed, what evidence reviewers receive, how approvals are recorded, and how overrides and escalations are governed. This paper proposes a Human-in-the-Loop Enterprise AI Architecture integrating AI models, RAG, enterprise data, policy engines, workflow orchestration, identity, approval services, audit trails, security, Responsible AI, and operational monitoring. It defines risk-based review patterns, confidence thresholds, exception routing, separation of duties, and measurable governance controls. The architecture enables organizations to balance automation and human judgment while improving safety, accountability, explainability, compliance, and enterprise trust.

Sanjeeve Kumar Gajadi, "Human-in-the-Loop Enterprise AI Architecture" International Journal of Scientific Research and Engineering Development, V9(5): Page(662-665) September - October 2026. ISSN: 2581-7175. www.ijsred.com. Published by Scientific and Academic Research Publishing.