This Recommendation describes a unified framework and requirements for AI-enabled vehicle voice assistant to address limitations in current systems, such as poor contextual understanding, inability to autonomously plan complex tasks, and lack of personalized memory. The framework integrates large language models, multimodal perception, and dynamic task generation, enabling a shift from command-based control to adaptive, human-like interaction. This Recommendation presents an approach that elevate vehicle voice interaction systems into context-aware, personalized assistants capable of human-like conversational fluency and adaptive control decisions.