"AI Competition" in the Annual Reports of Listed Banks: By 2025, the six major banks will have invested over 130 billion yuan in financial technology, with scenario implementation coexisting with risk challenges

Ask AI · From Digital to Intelligent Data: What Are the Key Drivers for Upgrading Bank AI Strategies?

Everyday Economic News Reporter: Liu Jakuai Editor: Wei Wenyu

As the annual report season for A-share listed banks concludes in 2025, a set of figures sketches a new landscape of financial industry’s intelligent transformation—ICBC’s total fintech investment reached 28.59B yuan for the year; China Merchants Bank claims its AI (artificial intelligence) applications replaced over 15.56 million man-hours in one year; Ping An Bank’s large model application scenarios doubled to nearly 400 within a year…

(“Everyday Economic News” reporter, hereinafter “the reporter”) notes that in 2025, the combined investment of six major state-owned banks—ICBC, Agricultural Bank of China, Bank of China, China Construction Bank, Bank of Communications, and Postal Savings Bank—exceeded 130 billion yuan, further increasing from 125.46B yuan in 2024. Behind this massive investment, a deeper transformation is underway: AI has shifted from a future-looking technology chapter in annual reports to a key metric for measuring banks’ core competitiveness.

Meanwhile, across the ocean, JPMorgan Chase is painting another AI picture—CEO Jamie Dimon regards artificial intelligence as a “transformative technology comparable to the printing press and steam engine,” and announced an annual investment of over $2 billion to build a “fully AI-collaborative enterprise.” This Wall Street financial giant is not content with isolated applications but aims to deeply embed AI into every capillary of the organization.

On one side is the domestic banking system’s systematic, large-scale AI investments and scenario implementations; on the other side, international financial giants are advancing comprehensive intelligent restructuring through ecosystem thinking. This trans-Pacific wave of financial intelligence is quietly changing every core link—from credit approval and risk pricing to investment decision-making.

However, behind this hot AI investment and vision, challenges such as deep water in data governance, the “illusion” of models, and compliance risks from algorithm “black boxes” are testing the depth and sustainability of this transformation. The AI journey in finance, while demonstrating enormous potential, has entered a critical stage requiring more wisdom and prudence.

[Image suspected to be AI-generated source: Everyday Media Asset Library]

Strategic Upgrade: The Race from “Digital” to “Intelligent Data”

The reporter’s review of 2025 listed bank performance reports finds that “artificial intelligence” has risen from a technology outlook chapter to a key performance indicator for future core competitiveness. The focus of this race is shifting from “whether to apply AI” to “how deeply and how systematically,” showing clear features of systematic and scaled implementation.

State-owned giants, leveraging their abundant resources, are building “heavy infrastructure” for AI transformation. ICBC explicitly states in its annual report that it has fully upgraded its four-year “Digital ICBC” (D-ICBC) strategy to “Intelligent ICBC” (AI-ICBC), with its core “ICBC Wisdom Surge” large model deployed across more than 30 business areas, totaling over 500 application scenarios. China Construction Bank disclosed that AI technology has scaled to empower 398 scenarios within the group. Bank of China has built the BOCAI large model capability platform, deploying over 400 intelligent assistants.

Joint-stock banks and city commercial banks show even greater agility in scenario deployment speed and breadth. China Merchants Bank revealed at its earnings conference that it has 856 AI application scenarios, replacing over 15.56 million hours of manual work annually, equivalent to the full-time effort of over 8,000 people. More importantly, AI is shifting from a “cost center” to an “efficiency engine”: its intelligent assistant for relationship managers has increased effective customer visits per person by 14% and transaction scale per customer by 20%. Ping An Bank’s large model scenarios surged from “over 200” to “over 390” within a year, with AI-generated code accounting for over 30%. CITIC Bank has established a “large model + small model” collaborative mode, with over 120 scenarios deployed by the end of 2025.

From “AI priority” to “AI native,” leading banks are attempting to embed intelligence deeply into organizational fabric, building new competitive barriers.

A senior banking researcher told the reporter that the dense disclosures of AI achievements in 2025 annual reports mark that China’s banking digital transformation has entered a “deep water zone” centered on intelligent decision-making and process reengineering. This is driven by the industry’s continuous net interest margin compression, making efficiency and growth through technology an inevitable choice. AI investments are no longer just a budget item for tech departments but strategic investments directly linked to cost reduction, efficiency, risk control, and revenue enhancement.

Deep Application: The Implementation of Risk Control, Inclusive Finance, and Operational Efficiency Revolution

After years of exploration, AI’s application in banking has long surpassed early-stage intelligent customer service and facial recognition payments, penetrating core business areas and demonstrating disruptive potential in efficiency and risk management.

In the “heart” of risk management—credit and anti-fraud—AI is transforming from “rule-based judgment” to “intelligent perception.” Traditional risk control relies on historical data and static rules, struggling with complex, evolving risks. AI-driven intelligent risk control systems based on machine learning and graph computing can process massive heterogeneous data in real time. For example, Postal Savings Bank built a full-chain anti-fraud model system that protected over 100k potential victim accounts in the first half of 2025. China Merchants Bank’s online risk control platform approved nearly 600 billion yuan in corporate loans in 2025, up 44% year-on-year, with AI-assisted post-loan risk early warning times averaging 42 days earlier than traditional manual methods.

In inclusive finance, AI analyzes substitutable data to solve classic problems of “difficult, expensive financing” for small micro-enterprises. Many banks use AI models to integrate tax, invoice, supply chain, and utility data to create credit “profiles” for small micro-enterprises lacking traditional collateral, enabling rapid credit granting.

Smart operations and customer service are the most direct manifestations of AI cost reduction and efficiency. China Merchants Bank’s intelligent assistant for over 10,000 “Jin Kuihua” relationship managers has become a daily work partner. Ping An Bank uses generative AI (AIGC) to assist in marketing content creation, saving about 60 million yuan in 2025 alone. In operational back-end, AI “digital employees” are taking over many repetitive tasks. CITIC Bank has improved operational efficiency over twofold by using AI to streamline corporate account opening, information changes, and other processes.

“The success of AI in these areas hinges on solving the problems of handling massive data, complex patterns beyond rules, and real-time response under high concurrency,” said the aforementioned banking researcher. These mature applications form the “basic plate” of bank AI capabilities, with value reflected in cost savings, risk reduction, and improved experience.

He believes current applications are mainly “optimizing existing processes,” but the next phase of competition will focus on how to use AI to “create new processes” and even “generate new business,” transitioning from “internal efficiency” to “external revenue.”

Overseas Status: Breakthroughs from Process Optimization to Value Creation

While domestic banks focus on optimizing internal processes and customer service with AI, international giants like JPMorgan Chase are extending AI into more disruptive areas: investment decision-making itself.

In venture capital (VC) and private equity (PE), AI is reshaping the underlying logic of project discovery and due diligence. Traditional reliance on networks and industry research (e.g., Wind, Bloomberg) is being replaced. For example, Sequoia Capital has developed internal AI tools to automatically scan global startup data, academic papers, patents, and news, providing daily preliminary analysis briefs to investment teams, enhancing project screening breadth and efficiency.

In client-facing wealth management and investment banking, AI is moving from back-office support to front-end services. JPMorgan applied for the trademark of “IndexGPT” in 2023, a generative AI-based investment advisory tool that automatically analyzes and selects securities based on client input themes or focus areas. This model is trained on JPMorgan’s proprietary macroeconomic and research data, aiming to provide personalized investment portfolio advice.

In lending, AI-driven refined risk grading and pricing are already mature overseas practices.

The same banking researcher explained that overseas financial institutions’ AI practices reveal two key trends: first, AI is shifting from “internal process optimization” to “external value creation,” directly involved in investment advice and product design; second, top institutions leverage their unique, high-quality data barriers (e.g., transaction data, in-depth research) to train specialized large models, building new, hard-to-copy competitive moats. Compared to this, domestic financial institutions still have room to develop in using AI for direct investment decision-making and providing deep intelligent wealth advisory services, which may be future high ground to conquer.

Precarious Challenges: Data Governance, AI Illusions, and Talent Shortages

Beyond mature applications like anti-fraud and intelligent customer service, the financial industry is cautiously pushing AI into more advanced, core areas, trying to unlock new value and let AI act as “analysts” or even “junior decision-makers” in complex financial activities.

The reporter learned that in real-time market sentiment analysis and early warning, some institutions are training AI to capture and analyze massive unstructured data—news, research reports, social media, even satellite images—to detect potential market or company risk signals. For example, Orient Securities’ “Orient Brain” AI platform processes nearly 70k market sentiment pieces daily, automatically identifying corporate entities and classifying negative sentiment.

In post-loan management and asset preservation, AI is used for continuous, automated risk monitoring of existing loans. By analyzing corporate operational data, judicial information, and public opinion changes, models can pre-warn potential risks proactively. Some banks have begun using large models to assist in generating post-loan review reports, significantly reducing writing time.

More disruptive explorations are happening in core trading and investment areas. In quantitative trading, beyond optimizing existing strategies, cutting-edge efforts include developing “virtual traders” capable of autonomous learning of market microstructure and executing some trades independently. Reports indicate JPMorgan has launched its AI quantitative trading platform supporting high-frequency trading and multi-factor strategies. In client trading (e.g., forex, interest rate derivatives), AI is also studied for providing real-time optimal quotes and hedging suggestions.

However, despite promising prospects, deep AI applications in core finance face constraints: data governance, model “hallucinations,” and a shortage of interdisciplinary talent are major hurdles.

First, data governance. High-quality, standardized data is the “fuel” for AI. But financial data involves sensitive personal privacy and commercial secrets, often scattered across different departments, forming “data silos.” KPMG experts point out that financial institutions generally face difficulties in multi-source heterogeneous data coordination and internal data sharing.

Second, the “hallucination” problem of large models and reliability risks. The inherent “hallucination” issue of large language models is fatal in high-stakes financial decision-making. Postdoctoral researcher Lou Feipeng from Postal Savings Bank of China noted that if hallucinations occur in risk management, banks may lose understanding of risk logic, making effective responses impossible.

Third, the shortage of interdisciplinary talent and organizational transformation pains. Talents who deeply understand complex financial logic and AI algorithms are extremely scarce. Meanwhile, traditional hierarchical banking cultures emphasizing rigor conflict with the agile, trial-and-error development modes AI requires.

The aforementioned banking researcher summarized that future financial competition will be a comprehensive ecosystem of “technology—data—governance—talent.” Institutions that can first build high-quality data assets, establish trustworthy AI governance frameworks, and successfully transform organizational culture will gain long-term advantages in this profound “data intelligence” revolution.

Disclaimer: The content and data of this article are for reference only and do not constitute investment advice. Please verify before use. Risks are borne by the user.

Everyday Economic News