PUBLISHER: ResearchInChina | PRODUCT CODE: 2109326
PUBLISHER: ResearchInChina | PRODUCT CODE: 2109326
TSP Research: Leading providers collectively turn to AI agents to provide all-scenario active services
Telematics Service Providers (TSPs) are the core hub of the telematics industry chain, connecting vehicles, users and services. At the upstream end, they connect with OEMs, automotive equipment suppliers and network operators. At the downstream end, they connect content and application providers. Therefore, they assume full-link responsibilities for data collection, processing, distribution and closed-loop service delivery.
The current TSP industry has gone through the infancy stage centered on call centers, and the growth stage with the entry of Internet companies, and enters into the in-depth development stage driven by intelligent connectivity and AI. In 2025, China's passenger car infotainment system installation rate hit 90.7%, the OEM installation rate of satellite navigation 82.7%, the mobile phone remote control installation rate 90.4%, and the phone-IVI interconnection installation rate 66.1%. Telematics services have evolved from "optional" to "standard".
Foundation models and AI agents are reconstructing the underlying logic of the TSP operating system. In 2026, cockpit foundation models enter a critical stage for mass production. The installation rate of AI foundation models soared from 10.4% in Q1 2025 to 35.3% in Q1 2026. Doubao has been installed in over 7 million vehicles, over 50 brands, and 145 vehicle models; Baidu Duthink, Tencent's five major agents, Banma Zhixing, etc. have all completed agent upgrades.
In the field of TSP, TAI focuses on software capabilities and cloud platform layout. The core evolution route is: content ecosystem empowerment -> AI foundation model-based agent cockpits -> people-car-life digital ecosystem.
As of 2026, TAI's TSP solution has been launched in over 60 brands and over 16 million vehicles; in the field of automotive cloud, it has cooperated with over 100 OEMs/mobility technology companies; over 30 OEMs and technology companies have used Tencent's overseas cloud services. TAI's core advantage in the TSP field lies in: its ability to introduce the WeChat ecosystem (1.15 billion monthly active users) and the applet ecosystem (8,000+ partners, 300+ joint solutions) into vehicles, and the AI agent architecture driven by Hunyuan Large Model + DeepSeek.
TAI 6.0 realizes AI upgrades with Tencent Hunyuan Large Model and DeepSeek. It uses five major agents, namely "AI local life agent", "AI all-round companion agent", "AI mobility exploration agent", "AI mobility connection agent" and "AI map navigation agent" to improve AI active services around all mobility scenarios. In April 2026, Tencent released the Open All-Scenario Mobility Agent Platform to promote automotive AI from "dialogue" to "active execution".
Baidu CarLife's evolution roadmap in the TSP field is: search + map empowerment -> foundation models + agent reconstruction -> AI mobility platform. As of 2026, Baidu CarLife has cooperated with more than 70 OEMs; CarLife+ has covered over 95% of smartphones, over 1,300 vehicle models, and over 70 brands; Baidu Maps has covered more than 200 countries and regions around the world.
The core advantages of Baidu CarLife include: the DuerOS X driven by ERNIE Large Model, the global coverage capabilities of Baidu Maps, and the mass production and application of the AI navigation agent "Duthink".
On March 27, 2026, Baidu officially launched its next-generation mass-production AI cockpit agent, with key highlights including:
The core architecture is unified by a "central agent" that schedules navigation, entertainment, vehicle control, and chat sub-agents, achieving a closed loop from conversational understanding to complex task execution.
Based on end-to-end voice language foundation models, users do not need to memorize fixed operating skills and can complete navigation and cockpit control through natural dialogue.
Relying on the deep integration of Baidu Maps and ERNIE, the agent can accurately identify dialects and unpopular place names, has emotional perception, scenario adaptation and long-term memory capabilities, and makes speech responds in milliseconds, achieving in-car interaction closer to real-person conversations.
In March 2026, this technology was mass-produced for Buick Electra OS; during the 2026 Beijing International Automotive Exhibition, it was unveiled at the booths of SAIC-GM and Dongfeng Motor.
Banma Intelligence's core evolution roadmap in the TSP field: SAIC's exclusive OS -> driven by AI foundation models+OS -> industry-wide open platform. As of 2026, Banma Intelligence has served 69 OEMs and over 10 million vehicles, and covered over 16 international markets. As for the TSP ecosystem, Banma Intelligence has cooperated with more than 400 partners, and the automotive applet service platform has boasted over 4 million registered monthly active users.
Released in September 2025, Yan AI is a high-performance AI software stack with agent-specific features that has been specially developed and built exclusively for automotive scenarios. On January 18, 2026, Banma Intelligence's "Yan AI" passed the ASPICE 4.0 CL2, becoming the first domestic company to obtain this certification in the field of mass-produced AI foundation models.
Yan AI has agent characteristics and can actively make decisions and take actions to achieve specific goals without the need for continuous manual input. "System Agent" is part of the software stack of Yan AI, integrating agents from multiple automotive systems to meet user needs efficiently and safely. Auto Omni is available offline in 90% of scenarios as an on-device foundation model. The multi-agent collaboration solution "AutoClaw" enables a multi-agent collaboration system - complex tasks can be broken down into simple statements (such as simultaneously booking restaurants and adjusting itineraries) and run 24/7 in the background.
To date, Banma Intelligence's Yan AI has achieved mass production/designated cooperation with over 10 brands, including IM, BMW, Hongqi, Roewe, Volkswagen, MG, and Hyundai. In 2026, BMW will upgrade its entire lineup to Yan AI, marking Banma's transition from a dedicated OS for SAIC to an open platform for the entire industry.