Who Leads When AI Acts? Authority and Accountability in Agentic Organizations

Abstract

As artificial intelligence systems evolve from passive tools into autonomous agents capable of initiating actions, coordinating tasks, and making consequential decisions, traditional models of organizational authority are being fundamentally reshaped. This article examines how leadership, authority, and accountability function in “agentic organizations” where AI systems act with delegated discretion. We argue that authority in such settings becomes hybrid and distributed, emerging from interactions between human actors, algorithmic systems, and institutional governance structures rather than residing in a single identifiable leader. Drawing on organizational theory, governance scholarship, and AI ethics frameworks, we develop a conceptual model distinguishing operational authority (who executes decisions), epistemic authority (whose judgments are trusted), and moral-legal accountability (who bears responsibility). Through illustrative cases across corporate, public sector, and platform-based environments, we show how agentic AI complicates chains of command, diffuses responsibility, and challenges existing liability regimes. The article proposes governance design principles—including transparent delegation architectures, traceable decision pathways, and layered accountability mechanisms—to clarify leadership roles and preserve human oversight without undermining the performance advantages of AI autonomy. By reconceptualizing authority as socio-technical and dynamic, this work contributes to emerging scholarship on AI governance and offers practical guidance for organizations navigating the transition to agentic systems.

Keywords: Agentic AI; organizational authority; accountability; algorithmic governance; distributed leadership; AI ethics; socio-technical systems; human–AI collaboration; responsibility attribution; corporate governance.