The integration of AI-driven collaborative robots (cobots) into industrial environments has led to significant advancements in adaptive automation, particularly in custom-oriented manufacturing processes where human-robot interaction is critical. This paper presents a novel approach to embedding artificial intelligence within the control architecture of a collaborative robot, enabling real-time assessment of a human worker's psychophysical state and adaptive support mechanisms. The system leverages a multi-modal perception interface consisting of depth cameras, motion sensors, and physiological indicators to continuously monitor worker fatigue, loss of concentration, stress due to lack of training, or potential health impairments. A fuzzy inference system (FIS) is employed to process these heterogeneous inputs, providing dynamic estimations of worker states on a scale from 0 (non-existent) to 1 (critical), facilitating an intelligent decision-making framework for robotic assistance. Based on the estimated worker condition, the cobot autonomously generates appropriate responses—ranging from verbal or gestural alerts to direct physical assistance, such as adjusting task execution speed, tool handling, or temporary workload redistribution. The collaborative robot is fully networked within an information-driven workspace, ensuring seamless access to production schedules, technical documentation, and real-time operational data. This AI-enhanced embodiment of collaborative awareness ensures a safer, more efficient human-robot partnership, reducing occupational risks while improving productivity in flexible, low-volume manufacturing environments. The paper details the system architecture, AI-driven perception and inference mechanisms, and practical implementation in a structured industrial setting.