This study introduces an Artificial Intelligence-enhanced Consensus-Based Bundle Algorithm For Cooperative Robots Task Planning. The method combines an optimized Consensus-Based Bundle Algorithm (CBBA) with a solution to the Travelling Salesman Problem (TSP), enhancing the robots’ path planning efficiency. It uses graph convolutional neural networks (GCNNs) to better understand task requirements and the physical layout of areas, leading to more effective cooperative planning. The paper’s main contributions are two-fold: first, it introduces a GCNN-based architecture that improves how tasks are assigned to robots. Second, it integrates an improved TSP solution to optimize the paths that robots take. This approach is decentralized, allowing for effective distribution of tasks among multiple robots while considering each robot’s abilities and the need for communication between them. The effectiveness of this method was thoroughly tested in an indoor demonstration at the City, University of London.