Summary of Key Points
Automobile companies' use of agents is not about a "night-and-day disruption of the industry" through some revolutionary technology. Instead, it's about identifying high-frequency, real-world scenarios that can form a closed loop and implementing them through the "multiplicative effect of three key capabilities: scenario connectivity, engineering expertise, and model-driven functionality*. SAIC Motor has chosen to improve efficiency internally (through sales training, research and development, and knowledge management), while ZeroRun is focusing on enhancing the user experience within the vehicle cabin. The ultimate goal is for these three capabilities to work together to achieve initial successes in specific scenarios and gradually build a competitive advantage.
Detailed Analysis
1. The Core Logic of Automobile Agents: Start with Small Scenarios
Past changes in the automotive industry (such as the adoption of new energy vehicles for urban commuting or intelligent cockpits for navigation and entertainment) have shown that it's not grand narratives that drive significant change, but rather the successful implementation of value loops within specific scenarios. The same principle applies to agents: they won't revolutionize the entire industry overnight; instead, they need to find practical use cases that users or employees encounter daily and can solve real problems (for example, using agents for sales training or ordering coffee from the cabin). Only when agents become an integral part of these daily activities can they truly take hold.
For instance, while voice assistants were initially just functional interfaces (like turning on the air conditioning), today's in-vehicle agents can automatically place orders for coffee based on past preferences, navigation routes, parking information, and promotional offers. It's only when such scenarios are successfully implemented that users will start to rely on them.
2. Three Steps to Implementing Agents
For automobile companies to adopt agents effectively, they must overcome three key barriers:
- Step 1: From Tools to Scenario Validation: Many companies start with standalone tools (such as AI customer service or code completion), but these often fail to solve practical problems and merely appear impressive. To overcome this hurdle, tools must be integrated into specific use cases (for example, using agents for sales training within the store training system) and clearly demonstrate their value and impact.
- Step 2: From Single Use Cases to Widespread Application: After a use case is proven successful, it shouldn't be immediately rolled out across the entire company. For example, AI programming should not just complement code; it must be integrated into the entire development process (including requirements, development, testing, and delivery), and knowledge bases should be used in business decision-making. The focus here is on stabilizing the value of these tools and identifying replicable methods.
- Step 3: From Partial Implementation to Systematic Operation: Agents should evolve from pilot projects into a regular part of business operations, with capabilities shared across departments (for example, research and development models being used by sales teams) and a unified management framework in place (including security permissions and auditing). Only then will agents become a permanent part of the company's infrastructure.
3. The Multiplicative Effect of Three Key Abilities
The effectiveness of agents depends on the combined impact of these three abilities:
- Scenario Connectivity: Agents must be integrated into daily business processes. For example, research and development teams should integrate them into the code development process, sales training teams into store operations, and cabin systems into navigation and lifestyle services. Agents should not exist as isolated tools but should be seamlessly integrated into user experiences.
- Engineering Expertise: Companies need to ensure that agents are reliable and secure. The automotive industry demands high levels of stability and security; agents must provide accurate information, run smoothly, and have clear access controls (for example, using existing knowledge to answer questions without mishaps, with data protection mechanisms in place).
- Model-driven Functionality: Agents should be both affordable and user-friendly. It's not about using the most complex models; rather, the right model should be selected for each use case (e.g., lightweight models for sales training and robust models for complex research and development tasks), balancing functionality with cost.
4. Two Approaches to Implementation
There are two common approaches:
- SAIC Motor: Focusing on internal efficiency improvements. SAIC is enhancing sales training, integrating agents into the R&D process, and consolidating knowledge across departments to improve efficiency.
- ZeroRun: Enhancing the user experience directly within the vehicle cabin. ZeroRun has placed agents in the cockpit, allowing users to perform complex tasks with a single command (e.g., ordering coffee). This approach relies on cross-system task coordination (maps, apps, payment systems), creating a seamless user experience that leverages the capabilities of agents.
Both approaches follow the same principle of starting with small, practical use cases and gradually building scale.
5. The Key to Success: Continuous Iteration
Achieving initial success with agents requires continuous iteration and accumulation of advantages:
- Integration into Real Scenarios: Agents must solve real problems from day one.
- Value Verification: Evidence of improved efficiency, increased training effectiveness, and higher user task completion rates is necessary.
- Scalable Replication: Success should be replicated across multiple scenarios.
The company that ultimately succeeds in this competition will not be the one that first introduced the concept of agents but rather the one that effectively manages, utilizes them, and measures their impact (e.g., by determining how much time they save for sales teams and how many problems they solve for users). By continuously iterating, these companies can turn small successes into a lasting competitive advantage.
Conclusion
The competition in the field of automobile agents boils down to the ability to identify and leverage real-world scenarios effectively. By combining the right capabilities and implementing them systematically, companies can transform agents from experimental tools into a core part of their operations. Whether the focus is on internal efficiency improvements or enhancing the user experience, the ultimate goal is to ensure that agents become an integral part of the business process, which is the key to achieving initial success.