Motivated by lower predicted operational costs, and opportunities for efficient real-time control, automated, centrally coordinated vehicles have in many studies shown great potential as a shared resource within public transit. One particular use case that has grown in popularity over recent years is the application of smaller, automated shuttles as an on-demand feeder to mass transit solution. To investigate differences in fixed versus on-demand operational policies, this paper discusses the operational design and analysis of an automated feeder solution. To this end, a simulation model of demand-responsive transit is developed and incorporated into the transit simulation model BusMezzo. An estimation of operational cost reductions with vehicle automation motivates the case study of two fleets that are deemed comparable with respect to service capacity and operational cost per hour. Results from simulation studies of varying levels of demand indicate that the on-demand policy reduces average total passenger travel times and, for the larger fleet, lowers average vehicle-kilometers traveled per passenger relative fixed service operations. Without achieving a competitive reduction in waiting times, however, on-demand coordination often underperforms with respect to level-of-service and reliability when compared with fixed service operations. When there is slack in fixed service capacity, the performance of the on-demand service outperforms the fixed service with respect to both level-of-service and vehicle utilization only for the lowest demand level tested and the smaller fleet. Average total system costs under on-demand operations improve, however, for the lowest demand levels and the larger fleet due to a reduction in vehicle-kilometers traveled relative a fixed service. When fixed service capacity is exceeded it is found that on-demand coordination outperforms fixed operations with respect to average level-of-service, vehicle-kilometers traveled, and total system costs. Furthermore, when planned service capacity is exceeded, it is found that total passenger waiting time is more equally distributed under on-demand operations relative to fixed.