The objective of this study is to provide analytical guidelines for the design of shared-vehicle Autonomous Mobility-on-Demand (AMoD) systems. Specifically, we consider the fundamental issue of determining the appropriate fleet size from operational perspectives. In this study, we model and analyze the AMoD system, whereby all modes of personal transportation in a city are replaced by one centralized controlled fleet of automated vehicles. A framework which integrates traffic assignment, vehicles routing and automated vehicles rebalancing is provided to estimate fleet size. Experimental results, based on simulations, are provided using actual demand data obtained from NYC Taxi and Limousine Commission. Results reveal that in midtown Manhattan during weekday morning peak hours, an AMoD fleet whose size is 63% of that currently in operation can satisfy all travel demands with the passenger waiting time less than 6 minutes.