On September 9, the AI ride-hailing service jointly launched by Caocao Mobility and Doubao went live in Beijing, Hangzhou and Suzhou. Users can search for local life information within Doubao to generate travel plans and place orders, with support for smart vehicle matching, accessible premium vehicle identification and smart vehicle control preference settings. Caocao Mobility has previously applied AI in areas such as smart dispatching, operations management and safety risk control, and plans to deepen the partnership, expand to more cities and advance Robotaxi service development.

1. Three Cities Launch on the Same Day, AI Ride-Hailing Connects the "Search—Plan—Book" Chain

On September 9, the AI ride-hailing service jointly launched by Caocao Mobility and Doubao went live in three cities: Beijing, Hangzhou and Suzhou. Unlike earlier AI applications that mostly remained in back-end functions such as customer service Q&A and dispatch forecasting, this partnership connects mobility capabilities to an AI Agent, extending user-AI interaction from information retrieval to offline fulfillment.

On the Doubao side, when users search for local life information—including routes, leisure and entertainment venues, restaurants and more—the system not only replies with information, but can also proactively generate travel plans based on location information for users to choose from, forming a continuous flow of "information search—trip planning—ride booking." Users no longer need to switch back and forth between an information tool and a ride-hailing app; the demand entry point and the fulfillment step are placed in the same interface. For the platform, the AI Agent takes on the role of demand identification and entry-point distribution, while Caocao Mobility delivers the service through its existing capacity network.

2. Keyword-Driven Vehicle Matching, Accessible Premium Vehicles Enter Recognition Scope

Among the features launched this time, smart vehicle matching is one of the most tangible in terms of implementation. The system can recommend vehicle models based on specific user needs: when a user mentions keywords such as "large space" or "pre-ventilation," the system recommends a premium vehicle or the smart white vehicle. This mechanism turns steps that previously relied on users manually selecting a vehicle type and filling in notes item by item into parsing of natural language expressions.

In the accessible service being trialed in parts of Suzhou and Hangzhou, the system can recognize expressions such as "wheelchair" and "mobility difficulties" and recommend an accessible premium vehicle, simplifying a process that previously required users to filter on their own into a one-tap operation. Accessible travel has long suffered from low supply-demand matching efficiency; those in need often have to book by phone or leave notes in advance. If AI recognition can run reliably, it could reduce communication costs in this step. However, its actual coverage and recognition accuracy still depend on Caocao Mobility's capacity reserves in the relevant cities.

3. Smart Vehicle Control: Moving Preference Settings Ahead of Pickup

Beyond vehicle matching, the partnership also leverages the smart vehicle control capabilities of Caocao Mobility's custom vehicles. Users can set preferences such as route, air-conditioning temperature, ventilation mode and music style in advance in the Doubao conversation interface, and the vehicle completes adjustments before pickup; the system also records users' frequently used preferences and automatically applies them on subsequent rides.

The value of such features lies in moving adjustment actions that previously took place after boarding to before the vehicle arrives. For the operator, preference data also provides finer-grained input for vehicle dispatching and in-cabin environment management. However, smart vehicle control depends on the hardware support of custom vehicle models, and the proportion of vehicles it can cover will affect the feature's actual penetration in overall orders.

4. From Back End to Front End: Caocao Mobility's AI Implementation Path

Caocao Mobility is an important vehicle for the commercialization of RoboX under Zhejiang Geely Holding Group. It has previously applied AI technologies in areas such as smart dispatching, operations management and safety risk control, most of which are in the back end invisible to users. This partnership with Doubao, on the one hand, connects AI agents with mobility capacity resources, and on the other, extends service capabilities to user terminals, providing scenarios and fulfillment support for AI implementation.

From an industry perspective, the core proposition of combining AI Agents with mobility platforms is not the conversational experience itself, but whether demand can be accurately translated into executable capacity instructions. Vehicle matching, accessible identification and vehicle control preferences are essentially processes of converting unstructured natural language into structured order fields; the accuracy of this step directly determines whether the loop can truly run through.

5. City Expansion and Robotaxi: Pace Depends on Scenario Replication Costs

The company said it will continue to deepen cooperation with Doubao, explore more features such as smart vehicle control, and plans to extend this experience to more cities while advancing R&D of Robotaxi service capabilities.

The selection of the three pilot cities shares certain commonalities: all are cities where Caocao Mobility has relatively high capacity density and relatively concentrated deployment of custom vehicles. When the features are extended to more cities, the dependence of capabilities such as accessible premium vehicles and smart vehicle control on vehicles and regional resources may cause differences in replication pace. Regarding Robotaxi, the company currently describes its efforts as advancing service capability R&D and has not announced operating regions or a timetable. Whether the demand aggregation capability brought by the AI entry point and autonomous driving capacity can form synergies still needs to be verified by subsequent operational data.