Human factors in designing trustworthy agentic AI for industrial systems

Abstract

This thesis examines how human-factors principles support trustworthy agentic AI in industrial environments such as SCADA systems, manufacturing cells, and intralogistics. Through a literature synthesis (2015–2025) and a survey of 28 industrial professionals, the study identifies key barriers to appropriate reliance, including limited explainability, unclear handover, and weak governance. Findings reveal a trust–understanding gap where workers often trust AI output but lack insight into system reasoning. The thesis proposes a practical framework emphasizing intent visibility, uncertainty communication, controllable autonomy, and strong auditability. These measures help organizations deploy agentic AI that enhances rather than replaces human judgment, situational awareness, and safety.