Guotai Haitong: The tipping point for the landing of financial AI applications has arrived, driven by industry and policy.

Haitong releases a research report, indicating that both internal industry driving forces and external policy driving forces have led to a turning point in the implementation of AI in finance. The financial industry, as a typical data, information, and decision-intensive industry, has a high demand for digital transformation and is highly compatible with large-scale model technologies. By 2025, the release of DeepSeek R1 will promote a leap in general model reasoning capabilities and a sharp reduction in costs, as well as the open sourcing of models, becoming a turning point for the local deployment of AI in financial institutions. Currently, AI applications have accelerated penetration in various core business areas and back-office scenarios of financial institutions, and AI is expected to reshape financial business processes and organizational structure, opening a new era for financial digitalization. Both internal industry driving forces and external policy driving forces have led to a turning point in the implementation of AI in finance. Since OpenAI introduced the first version of the GPT model GPT-1 in 2018, after years of development, general large-scale model technology has crossed the key turning point from "technology verification" to "industrial adaptation." The large-scale application of AI in vertical domains is on the brink of realization and will gradually enter an accelerating period. The financial industry, as a typical data, information, and decision-intensive industry, serves as an excellent "testing ground" for the implementation of AI. In May 2025, Li Wei, Director of the Technology Department of the People's Bank of China, published a signed article in "China Finance Magazine," proposing "to steadily and orderly promote the application of artificial intelligence large models and accelerate the digitalization and intelligent transformation of finance." In recent years, artificial intelligence-related policies and top-level designs have provided external stimuli for financial institutions to further transform from "digitalization" to "artificial intelligence." From an industrial perspective, the continued expansion of IT spending in financial institutions has provided strong internal driving forces for the implementation of AI, with data showing a significant acceleration in large model-related bidding and procurement in the financial industry since 2024. From a technological perspective, the combination of AI and finance mainly involves two technological paths. One is training generic models with financial corpora, relying on the iterative capabilities of generic models. With the widespread use of reasoning big models, the ability of AI to solve complex financial problems has been enhanced. The release of the DeepSeek R1 reasoning model has become a turning point in the industry for the local deployment of AI in financial institutions. The second path involves developing financial domain-specific large models, which are more suitable for solving specific problems in the financial industry and are more aligned with industry needs and compliance requirements. In terms of application forms, AI intelligent agents, especially multi-agent collaboration, have become the focus of future research and development. Currently, large models are widely used in "short thought" financial scenarios such as understanding, question-answering, and information extraction, while intelligent agents are mainly aimed at high-planning scenarios that require AI to autonomously undertake long-process work tasks.