The ORoLoC3D project proposes an AI-based platform to optimize loading and delivery in logistics, solving the combined 3D-Loading Capacitated Vehicle Routing Problem. This challenge involves transporting packages of various sizes and weights to multiple destinations, minimizing the total distance while meeting several constraints including vehicle capacity, axle loading, and stability constraints. OptiMathR, in collaboration with the Artificial Intelligence Research and Innovation Center of the University of Modena and Reggio Emilia, has developed a prototype software, equipped with a dedicated user interface and validated in a real-world setting. The current solution uses evolutionary algorithms for routing and beam search for 3D loading, but is limited to small scenarios and requires overnight execution times. To scale, the ORoLoC3D project aims to consistently reduce the execution time and improve solution quality, by optimizing algorithms and exploring advanced techniques like mathematical programming and deep reinforcement learning. Future goals include large-scale testing, partnering with an additional industrial customer, and disseminating results through publications and technical expositions.