Fully autonomous vehicles (FAVs) operated by AI voice agents create a unique multi-user environment without a traditional driver, allowing users to engage in diverse activities such as entertainment and rest. In addition to the convenience of vehicle control, these agents face the challenge of resolving conflicts arising from multiple simultaneous user commands. This study investigates whether AI voice agents can effectively manage these command conflicts and enhance the multi-user experience (MUX). The research was conducted in two phases: In the preliminary study, online focus group interviews (FGIs) were conducted with 10 participants to explore their perceptions of using AI voice agents in FAVs. Participants shared their experiences and watched a concept video to discuss their expectations of FAV agents. Based on the FGI results, explainable AI voice agents were designed to prevent user conflicts between user commands, focusing on user tendencies and conflict contexts. In the main study, multi-user interactions with these agents were evaluated through online experiments with 89 participants. Two specific experiments were conducted based on the sources of conflict identified in the preliminary study: the first focused on control authority and the second on speech overlap. Participants watched scenario videos and assessed five variables, including sense of agency, trust, problem-solving ability, disappointment, and safety. The findings suggest that explanations provided by AI voice agents can effectively mitigate conflicts and improve MUX in FAVs. However, the effectiveness of these explanations varied across different contexts, indicating the need for alternative approaches. This research provides valuable insights for designing MUXs that prevent multi-user conflicts and meet user expectations in FAV.