This paper studies the potential impacts of autonomous vehicle (AV) sharing with mobility-on demand service on the public transit system. We analyze the current travel demand in the public transit system in Singapore with a special focus on the first-/last-mile problem during morning peak hours. The first-/last-mile in this paper is defined as the gap between origin/destination and the heavy rail stations. A feasible method to integrate AV sharing in current transit system is proposed, which is to use on-demand AV sharing service as the alternative to the low-demand buses to improve the first-/last-mile connectivity in the study area. An agent-based simulation model is built to evaluate the performance of the new integrated service. The simulation models the behaviors, movements, and interactions of the agents—passengers, AVs, and traditional buses. A bus-only scenario is firstly simulated for validation purpose based on the real-world statistics. Then a series of scenarios integrating AV sharing in public transit system with various fleet size and passenger’s sharing preference are simulated. The results under the AV sharing scenario show that, by letting everyone in the system share their last-mile rides, with careful selection of the size of AV fleet, the new service is able to (1) reduce the average out-of-vehicle time for the passengers, (2) occupy less road resources than the low-demand buses, and (3) have a higher possibility to be financially viable.