Interview with Zhang Lei, Chief AI Scientist and Senior Vice President of Yixin: How AI is Reshaping the Automotive Financial Services Ecosystem

As large models transition from the technological singularity to the industrial foundation, and as intelligent agents move from laboratories to production lines and clinics, the third wave of artificial intelligence is reshaping the global economic texture with unprecedented sharpness.

China demonstrates dual advantages in this transformation: it is not only a testing ground with ultra-large-scale application scenarios but also launches assaults in deep waters such as chip breakthroughs and algorithm open-sourcing. From breaking through single-point technologies to ecosystem-level innovation, from efficiency tools to new engines of productive forces, an AI development path with Oriental characteristics is emerging at an accelerated pace.

On July 26th, the World Artificial Intelligence Conference (WAIC 2025), themed "The Intelligent Era: Global Solidarity," brought together tech giants, academic pioneers, and policymakers from the AI field. This super feast covering technology, ethics, and art indicates that AI has evolved from an "industrial variable" to a "civilizational constant."

At this grand event regarding the artificial intelligence industry, 36Kr not only acts as an industry observer but also participates deeply as an industry connector. It set up the "Kr Star Live Studio" in the exhibition hall to uncover the underlying logic of the advancement of the artificial intelligence industry through dialogues.

During the dialogue, Zhang Lei, the Chief AI Scientist and Senior Vice President of Yixin Group (02858.HK), said that the auto finance industry will gradually transition from a labor-intensive to an AI-intensive one. Meanwhile, AI will gradually transition from being a so-called tool to making core decisions, including in terms of automation and autonomous decision-making. The entire industry still has a long way to go.

The following is the transcript of the dialogue, edited by 36Kr:

36Kr: First, please introduce the basic situation of the company and your overall impression of attending WAIC this time.

A: Yixin is an AI-driven fintech platform established in 2014 and listed in Hong Kong in 2017. It is committed to providing inclusive and convenient auto financing and value-added services for consumers. At the same time, it uses AI to drive technology and provides convenient, complete, and efficient fintech solutions for partners in the auto industry chain through technology empowerment. Yixin's services cover more than 340 cities across the country, serving over 100 automakers and over 100 financial institutions. It has also established cooperative relationships with 42,000 dealers nationwide, serving over ten million users in total.

This is our first time participating in this conference. I think it is very helpful for improving our cognition and broadening our horizons. Especially in the forum section, through the sharing of world-class scientists like Jeffrey Hinton, we have learned a lot. At the same time, this conference also provides an opportunity to communicate and exchange ideas with peers in the professional field of AI.

36Kr: I see. The theme of this WAIC is artificial intelligence. What are the biggest technical bottlenecks or challenges when Yixin implements artificial intelligence or AI technologies? And are there any common challenges in the entire industry?

A: OK, I think there are two aspects of challenges. On the one hand, we need to make a large number of high-quality decisions in the process of doing fintech, and these decisions should also be relatively transparent to users, enabling them to clearly understand why such decisions are made.

Second, data should not leave its domain. The financial field actually has very high requirements for data privacy and compliance. It is very challenging to achieve good results while ensuring compliance.

For the entire auto finance industry, there are actually two core difficulties: decision-making in complex scenarios and long-chain decision-making processes. Currently, most enterprises deal with these two problems by deploying a large amount of manpower.

36Kr: Both artificial intelligence and AI are changing the ecosystem of serving customers. Can you share a specific case to highlight the value of Yixin's services or solutions for customers?

A: If we break down the entire auto finance service chain, there are roughly several links. The first link is channel management, the second is customer intake, the third is financial risk control, the fourth is capital link matching, the fifth is customer service, and the sixth is post-financing asset management.

For example, in the financial risk control link, in the past, in our industry, when customers applied for auto financing, they needed to provide a lot of paper materials, and the risk control approval was basically done on a daily basis, or at the fastest, on an hourly basis. At Yixin, our intelligent online approval can achieve the fastest "second-level" approval while ensuring good risk control, which is very good in terms of efficiency and user experience.

36Kr: After AI or large model technologies are incorporated into the industrial ecosystem, has there been any new change in the overall competitive landscape? What trends will there be in the future?

A: From the overall perspective of the industry, the degree of AI application is not high enough at present. Regarding the competitive landscape, I think the industry will gradually transition from a labor-intensive to an AI-intensive one in the future. Meanwhile, AI will gradually transition from being a so-called tool to making core decisions, including in terms of automation and autonomous decision-making. I think the entire industry still has a long way to go.

Our model can basically be regarded as a general model for the auto finance industry. In 2024, our independently developed "ZhiXin Multi-Dimensional" multi-modal general artificial intelligence model successfully passed the filing under the Interim Measures for the Administration of Generative Artificial Intelligence Services, making us the first in the industry. In March this year, we became the first enterprise in the industry to officially release and open-source a high-performance inference model. This year, we will also release an agentic AI model to solve the long-term pain points of the industry by deeply integrating the needs of the auto finance scenario with autonomous decision-making intelligent agents. We believe that overall, AI application will be the major trend in the future of this industry.

36Kr: After the AI in the industry becomes very mature, what kind of positioning does Yixin want to have to survive in this industry?

A: In the future, our own positioning is to be a technology enabler in the industry ecosystem. We hope to serve the global auto finance industry more precisely and efficiently driven by AI. This is also the reason why we open-source the inference model. We hope that the entire auto finance ecosystem can develop more intelligently, and we can build it together.

36Kr: Do other vertical industries rely more on the generalization ability of general large models or the in-depth development ability of vertical large models in a specific field?

A: In my personal opinion, it is vertical large models. Take a typical example. For instance, the popular general AI intelligent agent products this year are actually general assistants. What problems can they solve? They can assist users in booking flights and hotels. But if you ask them to solve problems in a vertical field, they may not even know what auto finance is. The same goes for other industries. So, from a business model perspective, general models cannot be directly generalized.

36Kr: Then in which actual application scenarios of enterprises can the generalization ability of general large models be better utilized?

A: Take the clothing industry as an example. What kind of clothing is currently popular globally? This is something that general large models are more suitable for. You can see that major enterprises like Google and OpenAI are actually doing what is called deep research to produce in-depth research reports. They can capture data across the entire network, and then the large models can conduct corresponding analyses based on these data.

36Kr: Looking ahead to the next 3 - 5 years, after artificial intelligence technology becomes more and more mature, what different changes do you think will occur in your industry? How will the company make advance arrangements?

A: I think the biggest change in this industry in the next three to five years will be the deep integration of AI and the industry. AI may change from a tool level to a real strategic level to achieve strategic empowerment.

All links will be AI-enabled, which can fundamentally improve the operational efficiency of the industry and provide users with a better and more convenient service experience. In terms of positioning, I think we may truly be rooted in serving automakers and dealer groups. In the future, we can also follow the national strategy to serve Chinese automakers going global. There are actually very few companies like ours overseas, especially those that can help Chinese enterprises.

During this exhibition, some people from domestic large enterprises and financial institutions came to our booth for exchanges. One of the conversations was quite interesting. When I explained some technical details to him, he asked if we developed them ourselves or adjusted third-party ones. I said that all our models, products, and technologies were independently developed because we have had an AI team since 2015 and have been working on it all the time. He could hardly imagine that we were so far ahead. This also shows the importance of AI at the strategic level of a company. If I have considered this matter clearly strategically, then we can think more clearly about things in the next three to five years or even five to ten years. This is actually the current stage our company is in, and we have come this way step by step since 2015.