This research investigates the interaction between manufacturing system constructs and the operation strategies in a multiple-load Automated Guided Vehicle System (AGVS) when AGVs in a system can carry two or more loads. The load pick-up problem arises when an AGV stops at a pick-up queue and has to decide which part(s) in the queue should be picked up. Since an AGV can carry multiple loads, a drop-off rule is then needed to determine the next stop for the AGV to deliver one or more loads. Several real-time composite heuristic rules for selecting load and determining the next stop are proposed and evaluated in two manufacturing system constructs: the jobshop and the flexible manufacturing system (FMS). A number of simulation models are developed to obtain statistics on various performance measures of the two system constructs under different experimental conditions. The simulation results reveal that the pick-up rules affect the system more than the drop-off rules. In general, rules to avoid starving and blocking in workstations perform better than the rules for shortest distance in throughput. However, the rules perform differently in jobshop and FMS based on other performance measures, indicating an interaction between system constructs and load selection strategies. The difference in rule performance within the same construct is also affected by several AGVS design parameters. Overall our study suggests that no load pick-up rule is always a champion, and the design of an efficient multiple-load AGVS must consider all issues in a global fashion.