Systems and methods relating to management of automated delivery fleets are disclosed. Such systems and methods include using functionally-aligned machine learning engines to manage aspects of fleet dispatch and routing, inventory level adjustments, and order optimization. The techniques may be applied to manage an automated fleet delivery services system controlling a fleet of autonomous delivery devices (e.g., UAVs) to deliver products from a plurality of distribution hubs to fulfil customer orders. The interconnected machine learning engines enable efficient management of various aspects of such automated fleet delivery services in real time, including optimization of fleet distribution, optimization of inventory distribution, and optimization of order distribution. Optimization of order distribution may be obtained by dynamically adjusting delivery charges for orders based upon current fleet capacity and inventory levels.