Designing A Human-Governed Agentic Project Management Model: A Closed-Loop Architecture for Autonomous Project Intelligence, Decision Control and Human–AI Teaming

Abstract

Purpose: This article develops a Human Governed Agentic Project Management (HG-APM) model for integrating autonomous AI capabilities with human project management authority. Design/methodology/approach: An integrative conceptual synthesis combines four literature streams: AI-enabled project management, autonomous Manager Agents, human oversight governance, and human–AI teaming. Source supported findings are separated from conceptual inferences, and the resulting architecture is expressed as a closed-loop model. Findings: The model comprises Sense, Understand, Diagnose, Predict, Plan, Evaluate, Authorize, Act, Monitor, Learn and Replan. It proposes a consequence sensitive, graduated autonomy rather than a single autonomy level. Routine and reversible actions may be automated within predefined boundaries, whereas consequential financial, contractual, scope, safety, stakeholder, reputational and strategic decisions require stronger authority from the designated human or organizational decisionmaker. Originality: The model integrates project-management intelligence, human–AI teaming and governance into one operational architecture. It contributes to a project specific decision boundary and testable propositions for evaluating whether governed agentic AI improves project outcomes without weakening human control.