The modern transformation of the field of smart mobility has generated the necessity of changes in the direction of intelligent and sustainable fleet control mechanisms that can cope with such complexities of operations, demands of safety, and environment aims. This paper proposed an AI-IoT integrated system that could be used to maintain operational efficiency of the fleet management by utilising real time sensor data as well as machine learning algorithms and edge cloud connection. Its main goal is to allow predictive maintenance, driver behaviour analysis, fuel optimisation, and emissions monitoring on a centralised, data-driven dashboard. The prototyped solution was implemented on an actual fleet of 80 vehicles that integrated onboard IoT devices (GPS, OBD-II, fuel, tire, and acceleration sensors) with the cloud (Azure IoT Hub) and AI models, which were Random Forest, Support Vector Machines (SVM), and Reinforcement Learning. There was a blockchain layer that provided safe history of important events. The outcomes showed significant gains: unintended stoppages decreased by 66%, the driver safety levels rose by 2.3 (6.1 to 8.4), fuel consumption rose by 8% and the CO2 emissions decreased by 7.5%. All these results confirm that the framework has the potential to make fleet processes shift to proactive, thus enhancing cost-efficiency and sustainability. The next steps are going to involve the integration of the smart city infrastructure and the extension of predictive functionalities by utilizing a deeper learning paradigm to a larger scope of urban mobility.