On April 10, at the 2024 Digital Industry Chain Finance Industry Summit, MYbank announced that it would upgrade the Dayan system and apply the capabilities of AI large models to industrial chain finance for the first time.
It is reported that the large model is mainly used in two aspects. First, use the ability to extract knowledge to build an industrial chain map. After reading a large amount of commodity information and enterprise relationship information, the large model forms an industrial chain map, so that the small and micro enterprises in the upstream and downstream of the whole industrial chain can be "visible".
Through the ability of information analysis, the operation of small and micro enterprises can be evaluated in seconds. The large model is like a 24-hour intelligent production and research expert, reading research reports, judging industry trends, analyzing small and micro business data, and evaluating their creditworthiness.
It is worth noting that this application does not directly generate content and interact with users, but in the background, it becomes the "assistant" of the financial risk control system to help financial institutions identify small and micro enterprises.
(Feng Liang, President of MYbank, introduced the application of AI large model capabilities in the field of industrial chain finance)
At present, MYbank has built an industrial chain map of 9 industrial directions, including automobiles, medical care, construction, etc., through the large model, and identified more than 21 million small and micro enterprises in the upstream and downstream of the industrial chain, and the efficiency of WeChat portraits has been increased by 10 times. In the automotive and medical industry chains, MYbank has integrated it with the financial risk control system to improve the credit coverage and satisfaction of small and micro enterprises.
It is difficult to crack and identify AI large models, so that small and micro enterprises "do not drop the chain"
To understand the value of large-scale model applications, we first need to understand the dilemmas encountered by industrial chain finance.
At the summit, Feng Liang, President of MYbank, showed the story of a typical small and micro operator in the industrial chain. In Jiaxing, Zhejiang, Pan Yicheng runs a business called Chengda New Material Technology Co., Ltd. Its main business is the production of high-temperature nylon materials. The research and development of new materials takes a long time, requires capital investment, and the customer payment cycle is as long as 3 months, and the turnover needs funds, which is a common financial demand of small and micro operators. However, when Pan Yicheng applied for credit facilities from the bank a few years ago, he still had to mortgage his property and had to wait for a week.
Chengda is just a microcosm of the financial service dilemma of small and micro enterprises in the upstream and downstream of the industrial chain. "The portrait of Chengda in the eyes of financial institutions is like this - an ordinary small and micro enterprise, industrial and commercial information shows that it is engaged in machinery manufacturing, with an annual turnover of about 10 million. It's not clear what he produces and where he ends up. Therefore, in the absence of a brand enterprise guarantee, it is difficult for financial institutions to give a loan amount that meets the needs of the company. Feng Liang said.
In March 2023, MYbank's technical team began to try to solve Chengda's problem with a large model.
First of all, we need to "see" Chengda from the perspective of the industrial chain. Through the ability of knowledge extraction, the large model understands data from massive information, forms an industrial chain map, and then identifies the main business of small and micro enterprises through multimodal data fusion, collaborative reasoning and other technologies, and accurately mounts them to the industrial chain. For example, in the automobile industry chain, the large model "sees" engine manufacturers, 4S shops, bearing manufacturers and other links, and then sees which companies are distributed in each link and weaves them into a net. In the whole process, the large model is like a searchlight, illuminating every small and micro micro in the upstream and downstream of the industrial chain.
In this way, Dayan System can identify Chengda as located in the automotive industry chain, it has 12 patents, and it is a high-tech enterprise in Zhejiang Province. The nylon material produced by Chengda eventually went to BYD Auto and became the insulating layer to protect the core electronic devices of the connecting rod. Moreover, it is located in Jiaxing, Zhejiang, an emerging material production base, with the supply advantage of "near the water".
(Pan Yicheng, founder of Zhejiang Chengda)
After seeing it, you also need to see the business situation of small and micro enterprises. The large model can automatically read a large number of research reports and generate the most suitable business portrait to describe small and micro users. This process is like an AI production and research expert, researching the industry, forming trend judgments, analyzing various business data of small and micro enterprises, and forming a score of its business situation. These ratings, to a certain extent, represent his ability to repay.
When Cheng Da came to MYbank to apply for credit services, the risk control system combined with the industrial chain map built by the large model, knowing that it was located in BYD's supply chain, combined with the score of "AI production and research experts" and other financial risk control data for strict verification, and finally provided it with a pure credit loan of 2 million.
When capillary small and micro enterprises can obtain convenient and flexible financial support on a large scale, the capital financing and operation efficiency of the industrial chain will also be greatly improved. Taking the automotive industry chain as an example, MYbank's Dayan system has provided credit lines to more than 1 million small and micro enterprises. Among the users who have obtained financial services, 64% have obtained pure credit loans for the first time, and nearly 3% have become science and technology enterprises, and have obtained higher quotas. Because of the financial services approved in seconds, they have achieved "zero account period" in operation, can receive more orders, and the average monthly delivery volume has increased by 17%.
The key to the future of large-scale model industrial applications is to find scenarios
"We've found a direction, but we can't say we've dug up a lot of treasure right now. There is still a long way to go to explore the industrial application of large models. Fang Ke, a senior engineer at MYbank, told reporters.
It is worth noting that the Dayan system does not use the generation capabilities of large models to directly interact with customers. "The financial industry has extremely strict requirements for security and risk, and there are still many problems that need to be verified in the application of large models. For example, Fang Ke said that large models will have "hallucinations", and if the customer's business situation is misjudged, it may bring real money loss.
Therefore, in the practical application of the Dayan system, the industrial chain map drawn by the large model will provide customer identification, business scoring and portrait to the risk control system, but in the end, the loan amount obtained by small and micro operators is still the result of multi-dimensional cross-verification of the risk control system. "The risk control system is the 'nerve center' of the Dayan system and the 'goalkeeper' of the large model. ”
Why can MYbank apply the capabilities of large models to industrial chain finance scenarios? Fang Ke believes that this is due to the accumulation of MYbank's intelligent risk control system. On the first day of its establishment, MYbank adopted the method of big data risk control to provide 310 (3-minute application, 1-minute loan disbursement, and 0 manual intervention) services for small and micro operators. By the end of 2022, it has served more than 50 million small and micro operators. AI technologies, including graph computing, are also widely used in MYbank's risk control system to provide safe and convenient financial services for small and micro businesses.
"The accumulation of ultra-large-scale customer volume, data and risk control capabilities, and the accumulation of cognition of small and micro operators form the basis for MYbank to explore the financial application of the industrial chain with a large model. Fang Ke told reporters.
Wu Lianfeng, vice president and chief analyst of IDC China, said that the most important thing for high-value applications of large models is to find the right scenario. It must meet at least two conditions, one is the need to analyze massive amounts of data, and the other is that it requires a high level of expert knowledge. MYbank's Dayan System provides an innovative sample to build an industrial chain map through a large model and an expert-level credit score, so that better financial services can flow to the capillaries of the industrial chain. Under the industrial proposition of developing new quality productivity and "artificial intelligence +", this is a useful exploration to better serve the real economy and promote industrial upgrading with financial service innovation.