Gasgoo Munich-At Gasgoo' s 4th AI-Defined Vehicle Forum 2026 on September 18, Wu Ziping, SYNCORE' s CPO, outlined the roadmap for AI cockpits. His presentation detailed the direction, architecture, and implementation of AI in vehicles.
Wu said smart cockpit AI must evolve from a "you say, I do" model. It is becoming an exclusive partner that proactively fulfills user needs. To address fragmented entry points, Wu stressed the need for a unified interface. It handles multimodal inputs from inside and outside the cabin to provide proactive services. The strategy builds product clusters around high-frequency scenarios like travel and music. This breaks the boundaries of traditional pre-installed applications. Users can customize the AI' s personality to match different MBTI traits. Options range from a proactive caregiver to a professional butler. Virtual avatars or robots support these roles. They enhance the partnership through appearance, expressions, and tone. "We aim to turn AI and cars from tools into partners," Wu said. This encapsulates the product's core positioning.
Wu outlined SYNCORE' s approach to building full-domain perception capabilities. Perception scenarios extend from inside the cabin to the outside environment. This advances the integration of cabin and driving systems. It meets the coordination demands of larger vehicles. On the hardware front, the company plans for external microphones. It also plans for dashboard cameras (DVRs) during vehicle planning. This balances scene recognition clarity with safety. Wu detailed the core value of the AIOS system. Built on scenarios and architecture, it translates user needs into coordinated actions. The execution chain spans the entire process. This includes intent understanding and task breakdown, along with capability invocation and feedback. The system supports function calling for quick responses. It establishes a technical barrier through six underlying capabilities. This serves as the central hub for cockpit and vehicle intelligence.
End-cloud collaboration is a principle Wu returned to repeatedly. He explained that the vehicle side prioritizes operational safety and privacy. It also ensures real-time responsiveness. High-frequency needs that the edge cannot handle are moved to the cloud. The system adapts to data rules for global markets. It strengthens security governance for both edge and cloud. Regarding implementation, Wu outlined three principles. First, build a system for collecting high-value sample data. Second, monitor user feedback to adjust product definitions. Third, balance shared features with regional differences. "We ensure user privacy on the device side," Wu said. "The cloud continues to learn," he added.
Wu demonstrated the agent's capability to evolve from simple recommendations. It moves toward personalized orchestration across all scenarios. The system recognizes scenario-based needs without exact app boundaries. It infers intent and delivers optimal results. Users do not need to break down their commands. This covers content curation, audio adjustment, and sports searches. It also includes dining recommendations and travel planning. The system integrates third-party capabilities to complete the loop. Wu highlighted adaptations for family scenarios. This includes designing warm experiences for family and child travel. It simultaneously monitors for abnormal conditions inside the cabin. A demo video showed the system identifying driver fatigue. It also identified family outings and concurrent commands. The system then automatically matches the appropriate services. "Once the agent connects with services, it's not a question-and-answer session," Wu stated. "It delivers a final result."
To build an agent ecosystem, Wu proposed an in-cabin agent center. This unified platform allows users to configure activation and permissions. Automakers can leverage this to explore niche scenarios. They can continuously operate agent services and assess value. This assessment is based on penetration data. Concluding, Wu summarized three core principles for vehicle intelligence. First, a unified hub uses agents to connect all capabilities. Second, end-cloud collaboration prioritizes privacy on the device side. It enables continuous model iteration in the cloud. Third, closed-loop implementation achieves full-chain connectivity via AIoT. This runs from understanding needs to execution. SYNCORE will continue driving the evolution to full vehicle intelligence. Cars will transform from tools to understanding partners.
