Human-AI Teaming on the Factory Floor: Cognitive Load, Trust Calibration, and Productivity Outcomes

The deployment of AI decision-support systems on manufacturing shop floors introduces a new class of human factors challenges that existing implementation frameworks do not adequately address. While technical performance metrics for industrial AI - model accuracy, fault detection rates, prediction latency - have improved substantially, the operational value of these systems depends not on what the AI can predict but on whether human operators act appropriately on its predictions. This paper investigates three critical dynamics of human-AI interaction in real-time production environments: cognitive load (how AI information overlays affect worker attention, situational awareness, and decision quality), trust calibration (the conditions under which operators appropriately rely on versus override AI recommendations), and productivity outcomes (whether AI augmentation delivers measurable throughput and quality improvements net of disruption costs). Drawing on human factors research, field evidence from automotive and electronics manufacturing deployments, and the author’s practitioner experience across industrial technology environments, the paper finds that trust miscalibration - both over-reliance (automation complacency) and under-reliance (algorithm aversion) - is the primary failure mode of shopfloor AI systems, and that this miscalibration correlates more strongly with implementation design (how recommendations are presented) than with model accuracy. The paper introduces the Trust-Calibrated Interface Design (TCID) framework, proposing seven design principles for AI interfaces in high-stakes operational environments that promote appropriate trust calibration. Implications are developed for manufacturing operations leaders deploying AI decision-support systems, UX/HCI researchers designing industrial AI interfaces, and AI product managers building operational tools.