From intelligent cockpits to AI-driven vehicles, Banma Intelligence is focused on the upcoming evolution of automotive software.

      As large AI models increasingly integrate into vehicles, the competitive landscape of intelligent cars is undergoing transformation. Historically, smart cockpits primarily emphasized voice assistants, in-car applications, and multimedia services. However, with the emergence of on-device omni-models, AI agents, and AI operating systems, cars are transitioning from simple smart terminals that follow commands to AI agents capable of understanding users and proactively offering services.

      Recently, TechNode had the chance to converse with the Banma Intelligence team regarding AI-enhanced cockpits, on-device omni-models, AI agents, and the international expansion of China's intelligent automotive software sector. The company positions itself not only as a provider of AI models for smart cockpits but also as a platform technology entity that integrates comprehensive AI technologies, service ecosystems, and engineering capabilities. Central to this strategy is its vision of AI in All.

      Founded in 2015, Banma Intelligence has progressed from perception-oriented AI to generative AI and now to agentic AI. The journey from its initial AI voice assistant to the introduction of its AI in All strategy and the Yan AI smart cockpit technology brand in 2024 reflects a gradual development of a complete AI technology system covering foundation models, AI operating systems, and AI agents.

      Transitioning from smart cockpits to AI cockpits, Banma Intelligence’s current AI cockpit strategy revolves around Yan AI, which encompasses foundation models, on-device models, AI agents, and AI-native operating systems. Yan AI consists of models tailored specifically for smart cockpit scenarios, highlighted by AutoOmni, an on-device, omni-modal foundation model that integrates visual, voice, and textual information for comprehensive user perception and understanding.

      This methodology is encapsulated in Banma Intelligence’s concept of No Touch, No App. Traditionally, users had to activate a voice assistant, issue a command, or navigate the in-car system to access specific functions. Banma Intelligence envisions vehicles that can recognize user needs proactively rather than waiting for user initiation. For example, a vehicle might evaluate a driver’s fatigue and adjust the seat and music accordingly or utilize the user’s schedule to anticipate routes and charging stops.

      AI agents are moving the focus from merely responding to queries to executing tasks. Last year, Banma Intelligence released SystemAgent, marking a shift from conversational AI to task performance through a SystemAgent + AI Agents framework. Subsequently, the introduction of SuperAgent expanded agent functionalities into various areas, including entertainment, mobility, lifestyle, and vehicle services.

      This year, the company introduced AutoClaw, designed to deconstruct complex user requests, plan tasks, and mobilize various agents and tools to fulfill them. This transition signifies a shift in the role of automotive AI—from simply answering questions to actively accomplishing tasks for users. For instance, in a parking situation, a parking assistant agent could utilize external cameras to gather surrounding information while an on-device foundation model manages recognition and decision-making, completing parking payments by accessing relevant services. Additionally, when users exit the vehicle, the system could detect items like smartphones and laptops, proactively reminding them of any overlooked belongings.

      As these capabilities advance, cars are poised to move beyond mere terminals offering information and entertainment to becoming AI agents that comprehend their environment, utilize tools, and carry out tasks.

      Banma Intelligence believes these developments signal a fundamental shift in the automotive industry’s competitive logic from Software-Defined Vehicles (SDV) to AI-Defined Vehicles (AIDV). In the SDV era, competition has extended beyond traditional hardware capabilities—such as engines and chassis—to include software features like smart cockpits and intelligent driving. AIDV further advances this by embedding AI within the vehicle’s fundamental architecture.

      According to Banma Intelligence, a genuinely AI-native vehicle should not simply be a conventional car with an added AI feature; rather, AI must be integrated within the foundational operating system and core decision-making structure, designing the vehicle around AI throughout its architecture, data, interaction, and driving systems. Consequently, AIOS is seen as a potentially critical component of future development infrastructure. Banma Intelligence aims to create a technology stack that includes chip adaptation, system infrastructure, on-device omni-models, and an AI agent ecosystem, aiming to embed AI deeper than just a surface application atop the automotive operating system.

      In terms of on-device AI and global expansion, Banma Intelligence sees significant long-term potential in on-device AI for vehicles. Cars require the ability to process vast amounts of real-time data from cameras, microphones, and vehicle sensors while addressing privacy and safety concerns. This necessity underscores the importance of low latency, offline capability, and data security in automotive AI.

      However, Banma Intelligence maintains that cloud and on-device AI should not be viewed as mutually exclusive. Instead, the company is pursuing a collaborative cloud-device strategy: complex tasks can utilize cloud computing, while real-time tasks requiring stronger privacy measures can primarily operate on the vehicle itself.

      Additionally, the evolution of automotive AI is creating new pathways for Chinese intelligent automotive software firms to expand globally.