The rising demand for premium agricultural produce underscores the need for efficient, accurate sorting technologies. This paper presents an AI-based smart segregation system designed to automate tomato sorting, integrating deep learning, image processing, and robotic automation. The system employs a conveyor belt, ultrasonic sensors, a high-resolution camera, a weigh scale, and robotic arms to categorize tomatoes into reject, ripe, or unripe classes based on visual and weight attributes. Utilizing the YOLOv8 object detection model trained on 731 tomato images, the system delivers high-precision, real-time classification validated through rigorous testing. Results reveal substantial improvements over manual sorting, reducing labor costs, error rates, and processing time while enhancing operational efficiency. Its scalable design suggests applicability to diverse agricultural contexts, heralding advancements in automated farming.