Less Interaction But More Explanation: A Communication Perspective on Agentic AI Interfaces

Agentic AI requires a different communication model than traditional chatbots: Rather than reducing all interaction, these systems need more explanation and oversight communication as they take autonomous actions on users' behalf, shifting from a responsive to a proactive model.

Users struggle to identify who is responsible for AI actions: The paper identifies a critical risk where agentic AI's multiple communicative roles can blur whether users perceive the AI system or themselves as the source of actions, potentially undermining accountability and trust.

Three types of explanations are essential for responsible agentic AI: The authors propose that systems must provide explanations about (1) what actions the AI will take and why, (2) uncertainty in the AI's decision-making, and (3) how the AI coordinates with other systems or agents—moving beyond simple transparency to actionable oversight.

Customizable explanations preserve human agency: Rather than overwhelming users with constant information or hiding it entirely, allowing users to control when and which explanations they receive helps maintain meaningful human oversight as AI systems become more autonomous.

Communication design is a governance tool: The paper frames how AI systems communicate as central to responsible deployment, suggesting that interface and explanation design choices directly impact whether humans can meaningfully oversee and trust increasingly autonomous AI systems.