China's listed banks are ramping up the adoption of artificial intelligence, accelerating deployments in the first half as they seek to boost efficiency, enhance customer experience and unlock new growth, with experts saying large AI models are shifting from tools of productivity to engines of value creation.

During its 2025 interim results briefing on Aug 29, Han Jing, executive vice-president of China Construction Bank, said: "We use technology and data-driven approaches, especially AI, to empower precise customer identification, targeted marketing and accurate profiling. This is reflected in both our corporate and retail business segments. By gaining insights into customer characteristics, we can precisely tailor our products and services."

In corporate banking, CCB managed long-tail clients through robotic tools, adding over 130 billion yuan ($18.3 billion) in new corporate deposits in the first half. In retail banking, its intelligent management helped achieve a nearly 95 percent renewal rate for maturing deposits, said Han.

Industrial and Commercial Bank of China introduced over 100 new AI-powered application scenarios across key business areas such as personal finance, financial markets and corporate lending, including an AI wealth assistant and an intelligent investment research assistant.

Postal Savings Bank of China has developed more than 230 large-model application scenarios. Its interbank ecosystem platform has deployed bill-transaction robots, achieving full-process intelligent management across all bill types. Its investment banking transaction robots enable intelligent inquiry and response for bond underwriting, boosting inquiry efficiency by more than 95 percent. The lender's intelligent loan review assistants support over 30,000 loan approval cases daily by effectively identifying, extracting and classifying more than 10 types of images, thereby further improving loan review efficiency.

While large State-owned banks maintain leadership in applying digital technologies, joint-stock commercial lenders are also striving to catch up.

In the first half, China Everbright Bank launched a large-model intelligent policy assistant, building a knowledge base of over 1,700 policy documents and an intelligent policy search and analysis agent. By linking policy documents with business scenarios, it has improved compliance execution and built end-to-end policy service capabilities.

Bank of Beijing, embracing an "All in AI" strategy, is accelerating its transformation into an AI-driven commercial bank. It has established an AI system built on an integrated computing power infrastructure and developed two major model development and operating platforms.

Huo Xuewen, chairman at Bank of Beijing, said the integrated computing power base allows unified management of enterprise-level AI resources, significantly improving operational efficiency and enhancing data linkage and analytical capabilities. The two platforms, driven by both large and small models, empower product innovation, customer service and risk management. The bank has developed a core AI capability system with over 100 AI functions, rolled out a suite of practical AI tools and built more than 300 application scenarios.

Ping An Bank's interim report showed that in the first half, the lender built a large-model capability system, refining models through supervised fine-tuning, autonomous large-model planning and intelligent agents to enhance application capabilities in key scenarios. As of the end of June, it had deployed over 330 large-model application scenarios.

Beyond business applications, digital finance is also playing an increasingly important role in risk management and compliance. For example, in corporate risk management, Ping An Bank has deepened its exploration of intelligent risk control protocols powered by large models, building a knowledge base of risk experts and leveraging intelligent systems to improve the efficiency of risk-control processes and broaden the applicability of large models.

In retail loan approvals and collections, the bank relies on large-model capabilities to extract and identify risk characteristics from unstructured data and other elements.

Industry experts point out that the use of large AI models in risk management remains at an exploratory stage. Looking ahead, in order to better ensure security, AI large models are expected to gradually unlock greater value in forward-looking risk identification, accuracy of risk assessment, speed of risk monitoring and breadth of risk management.