By 2025, artificial intelligence (AI) in the financial sector will have moved beyond the "pilot testing" phase and firmly established itself in large-scale deployment. Large-scale AI models and Agent technologies are evolving from mere efficiency boosters into the cornerstone infrastructures that drive business innovation, refine risk pricing, and elevate customer experiences.

From millisecond-level risk management responses to cognitive augmentation in investment research, and on to personalized customer services, a sweeping transformation powered by "large models + agents" is reshaping the core operations of financial institutions across the board.

01. Intelligent Risk Management: From "Rule-Based" to "Predictive Insight"

Traditional risk management models, which rely heavily on static rules and historical data, act primarily as post-event responders, struggling to keep pace with the complex, dynamic, and often hidden nature of modern financial risks. In contrast, the new generation of intelligent risk management systems, built on large-scale models and integrating Graph Neural Networks (GNN), real-time stream computing, and behavioral modeling, is undergoing a qualitative shift from passive reaction to proactive prediction.

In the realm of bank credit and fraud prevention, AI is systematically redefining risk identification. For instance, China Construction Bank revealed in its 2025 mid-year report that its "Tianyan" Intelligent Risk Management System (an upgraded version of the former "Huiyan" platform) now covers 98% of the bank's retail credit operations. In the first half of 2025, the system helped reduce credit card fraud losses by 52% year-on-year and brought the non-performing loan ratio for inclusive micro and small enterprises down to 1.03%, a 1.7 percentage point drop from the end of 2024, significantly outperforming the industry average.

In high-frequency, fragmented payment and consumer finance scenarios, the real-time and context-aware capabilities of risk management have reached new heights. Ant Consumer Finance's "Scenario-Based Real-Time Risk Management System" can, in the split second a user clicks "confirm payment," conduct a millisecond-level risk assessment and dynamically adjust credit limits based on hundreds of data points, including merchant credentials, product categories, geographic location, and even weather conditions. Currently, the system provides "frictionless risk management" services to tens of millions of users, greatly enhancing user experience without compromising security.

In the insurance sector, AI is also making significant strides. Ping An Property & Casualty Insurance has leveraged AI technologies such as image reasoning and vehicle risk control models to build a digital risk management system that covers "pre-event warning, in-event response, and post-event review," substantially improving fraud detection accuracy and efficiency. In 2024, the system intercepted fraudulent claims totaling 11.9 billion yuan.

02. Intelligent Investment Research: From "Information Overload" to "Cognitive Enhancement"

Faced with tens of thousands of daily announcements, research reports, public opinions, and alternative data sources, traditional investment research models are trapped in an "information overload" dilemma. AI-powered investment research tools, through semantic understanding, logical reasoning, and causal modeling, are freeing analysts from repetitive tasks and enabling them to focus on higher-order strategic judgments.

In the securities sector, AI investment research has become a critical driver of Alpha (excess returns). CITIC Securities' proprietary "CITIC Securities AI Researcher" can automatically process vast amounts of public information, research reports, and alternative data, freeing researchers from basic information gathering and organization tasks. This has boosted their information processing efficiency by over 90%, allowing them to concentrate more on core value judgments.

Bank-owned asset management institutions are also ramping up their AI initiatives. China Postal Savings Bank has launched an investment research AI assistant that deeply integrates lightweight large models and Multi-Agent technology, covering core scenarios from market research to risk analysis. The assistant can compress traditional offline hours-long data searches into seconds and shorten the generation cycle of hundred-page research reports from tens of hours to minutes, achieving a seamless loop from "data to knowledge to decision-making."

Notably, the industry is shifting from "general capabilities" to "professional credibility." The "Huatai A-Share Observer Assistant," jointly developed by Huatai Securities and ByteDance's Kouzi Space, taps into professional financial sector APIs to access real-time and accurate market and financial report data. It also employs Python for complex calculations, enhancing the reliability and professionalism of its analysis results. This "AI + professional toolchain" integration model is becoming a key strategy for leading institutions to build their competitive edges.

03. Intelligent Services and Operations: From "Efficiency Gains" to "Value Creation"

If risk management and investment research represent AI's backend prowess, then on the frontend, AI is redefining customer experience.

In customer service, the "digital employee" ecosystem is maturing. For example, China Merchants Bank's "Xiaozhao AI" can now handle over 90% of common inquiries, including transaction failure troubleshooting, credit card limit adjustments, and wealth management redemption processes. Its speech recognition accuracy reaches 99.2%, and its emotional recognition module can detect customer emotions, automatically transferring calls to human agents when anger or anxiety is detected, achieving "empathetic intelligent services."

CITIC Securities has elevated digital employees to higher-value scenarios. As of November 2025, it has deployed 18 high-value digital employees covering critical nodes such as account opening reviews, compliance Q&A, product matching, and transaction monitoring. These employees have processed approximately 50 million requests, with a total of nearly 100 billion Tokens invoked and a daily average processing volume exceeding 1.3 billion times. The related technologies have secured 10 national invention patents and 4 software copyrights, with its comprehensive strength ranking among the industry's best.

In wealth management, AI's personalization capabilities are particularly prominent. Ant Group's "Zhixiaobao" wealth assistant provides dynamic portfolio adjustment recommendations based on users' risk preferences, life cycle stages, family structures, current asset allocations, and market dynamics. For instance, when the system detects that a user's child is about to enter university, it automatically reduces the proportion of equity assets and increases allocations to money market funds and short-term bond products, achieving "life stage-adapted" intelligent wealth management. This "companion-style wealth manager" model allows ordinary investors to enjoy private banking services previously limited to high-net-worth individuals.

More profoundly, AI is breaking down the "high-net-worth barrier" in wealth management. In the United States, intelligent advisory platforms like Wealthfront and Betterment offer automated investment services with annual management fees as low as 0.25% through algorithm-driven ETF portfolios for mass clients. In China, platforms such as Tencent Licai Tong, JD Finance, and Du Xiaoman have also launched "AI advisory" products with minimum investment amounts as low as 1 yuan, truly living up to the original aspiration of "inclusive finance."

04. Future Outlook: AI as a Collaborator, Not a Replacement

Looking back from the 2025 vantage point, AI has not "replaced financial practitioners" as early predictions suggested but has become a "super collaborator" for human experts. Top investment bank traders no longer stare at screens for market monitoring but collaborate with AI to formulate strategies. Risk control officers no longer bury themselves in reports but oversee the operational logic of AI models. Financial advisors no longer recite product manuals but leverage AI to provide clients with life-cycle financial planning.

It is foreseeable that the future core competitiveness of finance will no longer lie in capital scale or branch network count but in the integration capability of "AI + data + talent." Institutions that can build high-quality data loops, continuously iterate AI models, and cultivate a "human-machine collaboration" culture will stand out in the new round of competition.

As Nobel laureate in economics Paul Romer said, "What truly drives economic growth is not capital or labor but the generation and application of new ideas." In finance, AI is the catalyst for this ideological revolution. It will not end finance but will inevitably reshape it—making it more efficient, inclusive, and intelligent.

- End -