The growing adoption of remote and hybrid work models has highlighted the need for intelligent systems that can accurately evaluate employee productivity without continuous manual supervision. Traditional methods, such as attendance records and manual time tracking, often lack efficiency, accuracy, and transparency. To address these limitations, this study presents an AI- powered Employee Attention Monitoring System that integrates computer vision and machine learning for real-time assessment of employee engagement and activity. The system utilizes webcam-based facial landmark detection to determine presence and attentiveness while simultaneously collecting behavioral metrics, including typing speed, idle time, session duration, and application usage. These inputs are analyzed using a trained machine learning classifier to categorize employee states as Working, Idle, or Distracted, with results transmitted to a Flask-based back-end for secure storage and productivity score computation. An interactive web dashboard provides real-time analytical summaries, and an integrated alert mechanism notifies administrators when prolonged distraction exceeds predefined thresholds. Overall, the framework delivers a scalable, cost-effective, and automated solution for productivity monitoring, with future scope for fatigue detection and predictive performance analytics.