Author | Observation Team

Source | New Economy Observation Team

Since the emergence of ChatGPT at the end of 2022, the domestic and international markets have continued to heat up with debates, technological competitions, and practical applications surrounding large models, driving artificial intelligence large models to become the hottest track at present. As an industry that is intensive in scenarios, data, and knowledge, the natural endowments of the financial industry provide optimal conditions for the value creation of large models.

In the past two years, the development of financial large models has entered the fast lane, and the attitude of domestic financial institutions towards large models has quietly evolved: from initially heated discussions among hundreds of institutions about whether to enter the field, to the current situation where large models have taken root in various segmented scenarios within the financial sector, empowering intelligent marketing acquisition, wealth management, risk management, and other links to improve quality and efficiency, adding new momentum to the digitization and intellectualization processes across the financial industry and driving the accelerated development of new productive forces in finance.

01

Financial Large Models 'Flourishing'

At the recently held Bund Summit, Xiao Gang, former Chairman of the China Securities Regulatory Commission, candidly stated, 'A year ago, domestic financial institutions were still vigorously discussing whether to apply large models or were preparing for their application. This year, rapid changes have occurred, and many financial institutions have already begun to implement them…'.

As he said, large models in the financial sector have now entered a state of 'flourishing'.

The 'Insights into the Application Practices of AGI in the Financial Sector' report (hereinafter referred to as the 'Report') released by the China Academy of Information and Communications Technology's 'Foundation Building Program' in conjunction with the InfoQ Research Center shows that the participants in financial large models include licensed financial institutions such as banks, insurance companies, securities firms, and consumer finance companies, as well as financial technology giants like Ant Group, Tencent Financial Technology, JD Technology, Qifu Technology, and professional technical service providers such as Megvii and Zhipu AI, along with cloud service providers like Tencent Cloud, Volcano Engine, and Baidu Intelligent Cloud.

Taking banks as an example, the six state-owned banks and several leading commercial banks have ventured into this field. In their 2024 semi-annual reports, the six major banks all disclosed their latest progress in digital finance, AI large model research and development, and application. Liu Jun, President of ICBC, stated that the bank's AI large model technology has enabled the implementation of 56 new business scenarios.

The Report counts over 40 large models in the financial industry, including 12 vertical large models such as Ant Group's AntFinGLM, Qifu Technology's QifuGPT, and Max Consumer Finance's models.

Among them, Max Consumer Finance has launched the nation's first retail finance large model, 'Sky Mirror.' Currently, the 'Sky Mirror' large model has realized eight major application scenarios including intelligent marketing interaction, data decision support, and anti-counterfeiting security, comprehensively serving nearly 200 million users and expanding the coverage and benefit of financial services.

In 2023, Qifu Technology launched its self-developed large model for the financial industry, QifuGTP; and in the same year, Ant Group released its AntFinGLM large model for the financial industry.

Driven by participants in large models, the industry has unleashed unprecedented innovative vitality. According to the Report, in 2024, the market size of financial AGI reached 380 million yuan, accounting for 13.1% of the overall AGI market size (on the enterprise side), representing an increase of over sevenfold year-on-year compared to 2023.

02

Highlighting the Application Value in Segmented Scenarios

With the explosive development of financial large models, 2024 is known as the first year of large model application implementation. It is a consensus in the industry that large models should be close to scenarios and create value.

In the financial sector, the initial application of large models was primarily concentrated in simple operational and non-decision-making business segments such as customer service, knowledge-based questions, and databases.

However, in the past two years, with technological innovations and the continuous emergence of business needs, financial large models have released significant value in areas such as data integration and analysis, risk management, personalized product customization, and intelligent investment advisory, continuously integrating with specific businesses, thereby empowering financial services in terms of both depth and breadth.

For instance, at last year's Bund Conference, Ant Group released its AI business assistant, ZhiXiaoZhu 1.0, based on large model technology. This year, ZhiXiaoZhu 2.0 was iterated and successfully applied in multiple financial application scenarios within Ant Group's wealth management and insurance businesses, including sales, claims, financial writing, and marketing innovation. Widely validated, it has achieved remarkable results. In the sales service sector, ZhiXiaoZhu significantly improved the efficiency of experts' business development. Test data shows that the introduction of digital avatars in insurance agency organizations increased per capita productivity by 150%.

At Qifu Technology, since the launch of QifuGPT, it has gradually been implemented in four major application scenarios: small and micro-enterprise identification, intelligent marketing, customer service assistance, and software development, significantly expanding the coverage and benefit of financial services.

Among them, data shows that in the small and micro-enterprise sector, QifuGPT enhanced the accuracy of tagging small and micro-enterprise users' industry attributes to 96.7% through in-depth mining and analysis of massive data, creating exclusive 'industry portraits' for credit scenarios for over 30 million small and micro-enterprises and part-time family workers. Additionally, it mined and supplemented industry information for 6.89 million broad small and micro-enterprise users, further expanding the coverage of inclusive finance. Meanwhile, based on a knowledge graph constructed from 72,000 event nodes, it better identifies risks associated with small and micro-enterprises and optimizes service timing for them.

Thanks to the efficient operation of large models, as of the end of Q2 2024, Qifu Technology has established partnerships with 160 financial institutions, helping them provide credit services to over 53 million small and micro-enterprises and individual consumers, injecting new vitality into inclusive finance through new technologies.

According to data from Baidu Finance, its 'Xuanyuan' financial large model has been applied across various business scenarios at Baidu Finance, from marketing, customer service, risk control, office work to research and development, with initial success. In terms of code assistants, the adoption rate of code generated with the assistance of large models can reach 42%, improving the company's overall R&D efficiency by 20%. In customer service, large models have boosted service efficiency by 25%.

It can be seen from the above-mentioned vendors' landing exploration practices that financial large models have continuously highlighted their value in the industrial chain, fully unleashing their high-quality data value, deeply optimizing financial services, and driving comprehensive upgrading and inclusive development in the financial industry.

The '2024 Report on Generative AI Applications in the Financial Industry' released by institutions such as Baidu Finance indicates that generative AI is expected to bring incremental business value of RMB 3 trillion to the financial industry and may fundamentally change the way transactions are conducted, investments are managed, and risks are assessed.

03

Opportunities and Challenges Coexist for Financial Large Models

Despite the unstoppable development of financial large models, the financial industry's unique nature poses challenges in risk prevention, regulatory policies, and data accuracy.

The Report candidly states that in the financial industry, AGI technology applications are primarily concentrated in general business scenarios such as office work, marketing, and operations. For core businesses such as investment research, trading, and risk management, most financial institutions remain relatively conservative in their actions. This is related to the strongly regulated nature of the financial industry. Industry characteristics also restrict the large-scale application process of AGI technology products in the financial industry to a certain extent. Financial institutions face four major challenges in applying AGI: technological challenges, data challenges, resource challenges, and compliance and regulatory challenges.

Nonetheless, the development value of financial large models for the financial industry is self-evident and is bound to become a new driving force for innovation and development in the financial industry. The Report recommends that financial institutions should balance technological advancement and business adaptability under compliance requirements.

Recently, Professor Ouyang Rihui of the China Internet Economy Research Institute at Central University of Finance and Economics wrote an article pointing out that the basic path for future AI large models to promote high-quality development in finance is to strengthen infrastructure construction; develop coordinated supervision and development of data, algorithm models, and computing power; promote the compliant landing of 'large models + finance' applications; facilitate cross-industry cooperation between technology enterprises and financial institutions; enhance multilingual and cross-lingual capabilities; and promote the sustainable and responsible development of multimodal integration of financial large models.

Faced with challenges, industry players are also making efforts. On September 6, the 'Large Model Industrial Application Technology Alliance' led by Ant Group was officially established, aimed at promoting collaboration among members, jointly building industrial application technology resources for large models, promoting the openness of credible capabilities and service systems for large model industrial applications, and facilitating the maturity and promotion of large-scale AI applications in the industry.

In terms of ensuring financial data security, Qifu Technology continues to explore and apply the latest security technologies to build a more comprehensive data security protection system. Recently, it was awarded the 'Data Interface Security Risk Monitoring and Assessment DiRM Thematic Work (Pilot) One-Star Certification' for its outstanding security risk monitoring capabilities, demonstrating its leading position in financial security.

From a market perspective, the financial large model market remains a blue ocean. The Report predicts that China's AGI application market size will reach 454.36 billion yuan by 2030. Starting in 2028, the market will enter a period of stable development, with annual market growth rates remaining around 50%, and is expected to exceed 100 billion yuan in market size by 2027.

Faced with opportunities and challenges, vendors will 'dance with shackles,' actively exploring new scenario implementation solutions for large models while striving to balance data security, risk control, and compliant development, cautiously promoting the integration of large models and businesses.

It is foreseeable that through a series of actions, vendors will push the application of financial large models towards more complex and in-depth areas, further integrating them with the entire business process, promoting cost reduction and efficiency enhancement in the financial industry across broader areas and at deeper levels, and contributing long-term strength to the vigorous development of the real economy across the industrial chain.

*Disclaimer: The New Economy Observation Team publishes this article for the purpose of conveying more information and does not constitute any recommendation. Original articles may not be reproduced without authorization.

End