Analogies and Metaphors for Agentic AI: How Industry Practitioners Explain Emerging AI Systems

Abstract

Agentic Artificial Intelligence (AI) systems transition organizational technology from reactive applications to distributed, goal-oriented architectures. While earlier research has focused on their technical capabilities, little is known about how these systems are intellectually framed by those who develop and deploy them. We conducted 18 semi-structured interviews with industry practitioners working with agentic AI systems, including developers, consultants, engineers, and founders. Our results indicate that agentic AI systems are mainly explained through organizational and anthropomorphic source domains, such as employees, teams, or assistants, which embed abstract system qualities within familiar social structures. We contribute to HCI research by demonstrating that such conceptualizations serve as pre-structuring mechanisms that influence mental models, expectations, and interactions with agentic AI systems before direct engagement. Based on this, we outline implications for designing explanation strategies in human-AI interaction.