Zhang Zhan, COO of Caiyue Xingchen: By 2025, AI will break boundaries for universal access, and C-end intelligent applications will be key to financial transformation.

"On May 24, the Science and Technology Innovation Board Daily reported (Reporter Chen Mei) that 'After this year’s Spring Festival, I observed in third- and fourth-tier cities that even small business managers have started using AI tools to generate contracts, with efficiency far exceeding manual work — this indicates that AI applications have broken through industry boundaries.'"

At the recent 14th Dual-Mode IT User Conference 2025, Zhang Zhan, COO of Caiyue Xingchen, stated that with the evolution of technologies such as DeepSeek, the "AI+" wave is accelerating, and AI is penetrating traditional industries deeply from the tech circle.

Regarding the application of AI technology in the financial industry, Zhang Zhan stated that customer service is the core proposition. As users become accustomed to using AI for daily tasks, financial services must actively integrate into this trend. For example, in the elderly financial service scenario, many older users find it difficult to use apps due to their complex functions. However, if interactions can be achieved through AI in dialects such as Shanghai dialect, it can significantly lower the usage threshold. This type of technology already has mature feasibility.

In this context, financial institutions should seize the opportunity of the integration of "AI + APP" and take the optimization of C-end experience as the starting point to seamlessly embed professional services into users' daily lives.

The outbreak of C-end agents has also driven the ecological upgrade of domestic computing power. In Zhang's view, massive user scenarios provide rich training data for large models, which can accelerate model iteration. At the same time, the demand for large-scale computing power will open up market space for domestic chip substitution. "Taking Shanghai as an example, relying on the layout of computing power integration at the policy level, Caiyue Xingchen has built a computing power base based on Huawei's Ascend platform since last year, and has strategically cooperated with the head brokerage Guotai Haitong to build a large model of Junhong Lingxi's 100 billion parameters, providing intelligent online services for a large number of users, which is at the leading level in China."

However, in the development process, the application of large models needs to break through the "concept-driven" approach and shift to a "results-driven" approach, with the core goal of effectively solving practical problems. Zhang Zhan further analyzed that the application of large models must be deeply integrated into business scenarios, adhering to the principle of "customer-friendly," and enhancing user application capabilities through targeted optimization to maximize technological value.

At the same time, in the field of model training, challenges and skill demands coexist, highlighting the increasing importance of integrated solutions. "Financial institutions face numerous difficulties in the application of large models and the transformation of skills, such as complex technology integration and gaps in personnel skills. If vendors can provide integrated solutions that are simple to use and cost-effective, it will significantly enhance the efficiency of model application and development, effectively lower the technical implementation threshold, and promote the rapid development and innovation of large model applications in the financial industry."

In order to improve efficiency, Zhang Zhan proposed the concept of "AI digital employee". That is, through intelligent tools, shorten the personnel training cycle and reduce labor costs. At the same time, due to the separation of technical architecture and business needs caused by the traditional division of labor model, the key challenge for the implementation of large models lies in the deep integration of business logic and technical capabilities, and it is necessary to give full play to the synergy of cross-domain talents and experts to break the barrier of "two skins of technology and business".

Zhang Zhan believes that the ideal model is a deep collaboration among computing power, models, and customer demands, shifting from "selling computing power and models" to "selling scenario-based solutions."

With the construction of a universal ecology for "AI infrastructure and scene implementation," the year 2025 will be a key year for the domestic AI industry to "practice internal skills and strengthen output." "Only when the computing power foundation, model optimization, and business scenario implementation are seamlessly connected, will the procurement and management logic of financial institutions undergo fundamental changes - no longer scattered technology procurement, but rather custom solutions tailored around specific business goals. This ecological collaboration can not only enhance the efficiency of technology conversion but also promote AI from being 'exclusive to the tech circle' to 'universal in the industry,' ultimately making financial services smarter and more inclusive," Zhang Zhan concluded.

Source: Science and Technology Innovation Board Daily

Author: Science and Technology Innovation Board Daily