The construction sector contributes substantially to urban waste, yet conventional waste management practices remain labor-intensive, inconsistent, and lack traceability. This study presents a field-validated AI-Blockchain framework designed to improve real-time construction waste classification and accountability. A Convolutional Neural Network (CNN) was trained on 2,800 labeled images captured over a 30-day period at a G + 12 commercial construction site. The model achieved a validation accuracy of 88.6% and classified waste into five categories with an average inference time of 1.6 s per image. Simultaneously, 960 waste-related events were recorded using a private Ethereum blockchain network integrated with smart contracts to ensure tamper-proof, verifiable logging of waste disposal activities. The system achieved a 21.2% improvement in sorting accuracy (from 69% to 90.2%), a 70.5% reduction in average processing time, and an estimated operational cost saving of ₹48,000 within the observation period. Unlike prior theoretical studies, this work demonstrates a live deployment of an AI-Blockchain system in a construction environment, offering a scalable, secure, and transparent solution for real-time waste management aligned with circular economy goals. This field-based study highlights the transformative potential of digital technologies in driving operational efficiency and regulatory accountability in construction waste practices.