Over the past few decades, the mainland banking industry has been on the path of digital transformation. From the initial electronicization to informatization and then to digitalization, it is now moving towards the stage of intelligence.
In the face of the rapid development of science and technology, how should we promote digital transformation in order to enable banks to develop better in the digital era? At present, the explosion of large models has triggered a new round of technological revolution, how to embrace large model technology for the banking industry with high requirements for accuracy, reliability and security? Under the general trend of digital transformation of banks, how to get out of their own characteristic transformation road?
Recently, around the above topics, the reporter of "21st Century Business Herald" conducted an exclusive interview with Zhang Bin, chief information officer of China Minsheng Bank. Zhang Bin believes that the digital transformation of commercial banks is not only a process of improving technology and data capabilities, but also a process of organizational and cultural change. "The large-scale model that has attracted much attention this year is comparable to the invention of mobile phones and the Internet, which will inject new momentum into the digital transformation of banks.
Zhang Bin. Data map
On Digital Transformation: Minsheng Bank's Exploration
21st Century: Digital transformation is the trend of the times, and different banks have different transformation paths. What are the unique features of Minsheng Bank?
Zhang Bin: I think the characteristics of Minsheng Bank's digital transformation can be summarized into three points: first, from the very beginning, the digital transformation is regarded as an operation and management reform driven by digital technology, the second is to take digital finance as the core component of the bank's strategy, with the top leader in charge, top-level planning, and overall promotion, and third, it is guided by a clear strategy and methodology, supported by corresponding technology planning and data strategy. It is precisely these characteristics that have enabled Minsheng Bank's digital transformation to advance solidly and rapidly, and continue to achieve new results.
21st Century: Why do you mention business management and strategy in the first place? The topic of digital transformation often leads to a discussion about technology.
Zhang Bin: Indeed, when it comes to the digital transformation of banks, technology is very important, and data is also very important as a new factor of production. I would like to emphasize that digital transformation is not a simple technology project, nor is it a simple technology project, nor does it mainly use technology to solve the problem of improving the efficiency of internal processes, but to use digital technology and data elements to promote the innovation and transformation of business and business models, and achieve long-term sustainable development of banks by adapting to customer requirements and environmental changes. Only by accurately understanding and grasping the essence of digital transformation can we start with the end in mind, clarify the correct goal of transformation, and formulate corresponding transformation strategies and implementation paths.
In 2021, the new board of directors of Minsheng Bank formulated a new five-year development plan, adhering to the corporate mission of "serving the public and caring for people's livelihood", clarifying the strategic positioning of "a bank for private enterprises, an agile and open bank, and a bank serving with heart", and at the same time put forward the goal of "building a digital bank with agility and efficiency, ultimate experience, and value growth". Guided by the new five-year development plan, Minsheng Bank's digital transformation work has been fully launched. Digital transformation is regarded as a comprehensive and profound operation and management reform of the bank, and digital finance has become a core component of Minsheng Bank's strategy, and top-level planning and overall layout are carried out.
In terms of organizational system, the head office has set up a leading group for digital finance composed of all bank leaders, which is responsible for the decision-making and deployment of major digital issues of the bank, and the development planning department of the head office assumes the responsibility of the secretariat of the leading group, which is responsible for the organization, promotion and tracking of the daily work of digital transformation.
In addition, it took the lead in setting up two digital "pioneer" departments, the Data Management Department and the Ecological Finance Department, to enhance the intensive development and utilization capabilities of data elements and the agile innovation ability of customer-centricity. The Data Management Department is responsible for leading the bank's data governance and data asset management, maximizing the value of data through the use of analytical tools, algorithms and models, while the Eco-Finance Department is based on the bank's strategic, cross-business domain and cross-functional team scenario financial product and business model incubation, integrating financial services into government affairs and people's livelihood, business operations and personal life.
In terms of talent team building, we will pay equal attention to increasing external introduction and strengthening internal training. While continuing to introduce high-quality digital talents, we have carried out hierarchical training on digital leadership, digital professional ability and digital general knowledge ability for all cadres and employees of the bank, laying the intellectual and skill foundation for digital transformation.
21st Century: What exactly are the strategies and methodologies you mentioned, and how can technology planning and data strategy support digital transformation?
Zhang Bin: Minsheng Bank's digital finance strategy is "ecological banks seek breakthroughs, and smart banks go to the next level". Building a smart bank refers to the use of digital technology to comprehensively transform and upgrade the value chain activities of commercial banks, improve customer experience and process efficiency by reshaping the customer journey end-to-end, or improve the quality and efficiency of bank operation and management through digitalization and intelligence. The construction of an eco-bank refers to exploring a new model of scenario-based finance, and integrating financial services into government public platforms, enterprise production and operation, and personal life scenarios by entering or jointly building an ecosystem. The digital finance strategy has made the bank's vision of digital banking more concrete, and has played an important guiding role in promoting digital transformation in a comprehensive and orderly manner.
Based on the strategy of Minsheng Bank and in accordance with the relevant plans and guidelines of the regulators, we have formulated a three-year technology development plan and the first data strategy of the bank. With the goal of "entering the leading ranks of domestic commercial banks in terms of scientific and technological level", the science and technology plan has formulated seven enterprise-level digital solutions, including digital marketing, digital risk control, digital operation and digital decision-making, and planned the middle-platform application architecture, model-based data architecture, cloud native technology architecture, three-dimensional digital security architecture and green data center layout, and clarified the direction of industry and technology integration, agile and efficient science and technology governance. With the vision of "using data insight, making decisions with data, and managing with data, to become a data-driven bank", the data strategy follows the five principles of "data is visible and accessible, data is available and understandable, data is easily connected and shared, data can be empowered and value-added, and data is safe and trustworthy", and further clarifies the important areas and tasks of data capability improvement and empowerment business development.
The science and technology planning and data strategy is formulated in accordance with the bank-wide strategy, serves the implementation of the strategy and the achievement of goals, and is fully discussed and jointly compiled by scientific and technical personnel and business personnel, which has strong guidance for the development of specific work. Judging from the work practice in the past year, the scientific and technological work of Minsheng Bank has changed from the previous focus on strengthening shortcomings and supporting key projects to a new stage of goal-oriented and comprehensive and coordinated promotion. The continuous improvement of technological and data capabilities has played an important role in supporting and promoting the digital transformation of the whole bank.
21st Century: In addition to the top-level design, strategy, and methodology of digital transformation, you also mentioned management and cultural change.
Zhang Bin: Since digital transformation is a change in the operation and management of banks, successful digital transformation is inseparable from changes in organizational structure, institutional mechanisms and culture. As we mentioned earlier, Minsheng Bank has established a leading group for digital finance work, added a new digital pioneer department, and enriched the digital finance talent team, which are the changes in the organizational system and talent team.
In order to promote digital transformation, Minsheng Bank has established and promoted an agile project management mechanism. The leading group for digital finance conducts project decision-making and regular review of major digital transformation projects across the bank, and the business, risk, and technology departments select key personnel to form an agile project team to focus on tackling key problems and iterate rapidly, while providing centralized office space, special technology and financial resources, and exclusive assessment policies for the project team to support and guarantee. In order to fully support business agility, the technology line has launched an agile and lean dual-mode R&D system to support the integrated process of business development and operation.
As we all know, culture is the most important influencing factor for digital transformation. In the process of digital transformation, Minsheng Bank adheres to the values of "customer-oriented and people-oriented", advocates the cultural concept of "collaboration, innovation, openness, inclusiveness and pragmatism" throughout the bank, and continuously embeds digital genes into the thoughts and behaviors of all employees.
One of the keys to the success of digital transformation is to improve the digital literacy of all employees across the bank, so that the use of digital technology to improve work efficiency and customer experience becomes a conscious action of everyone.
Let's take an example of an office system. The office system supports the daily operation of the enterprise and is used by all employees, and to some extent, it can be said that the enterprise office system reflects the culture of an enterprise. Conversely, the office system can influence and shape the corporate culture, or support and shape a behavior model that reflects the company's values. Minsheng Bank strategically introduced the Feishu system for localized deployment, fully integrated with the bank's self-developed system, and created a new office platform "i Minsheng", which is not only an important measure of digital transformation, but also a profound change in corporate culture. After the launch of "i Minsheng" in the whole group, the collaborative efficiency and office experience of employees have been greatly improved, and the working mode has also quietly changed, and agility, efficiency, openness and flatness have gradually become a consensus. At the same time, "i people's livelihood" also promotes the precipitation and sharing of knowledge, which is conducive to the cultivation of digital talents.
21st Century: Digital transformation is a complex system engineering, what are the main results achieved after so much work?
Zhang Bin: Through the joint efforts of the whole bank, the construction of ecological bank and smart bank has achieved remarkable results.
The first is to iteratively innovate ecological financial products and services, and the integrated comprehensive service capabilities of large, medium, small and micro enterprises have been significantly enhanced. Relying on the transaction data, scenario data and enterprise relationship graph of the supply chain industrial chain, we have built a big data intelligent risk control system, and innovatively launched online data credit enhancement products such as "order e" and "procurement e", as well as special products of "export e-financing". As of the end of September, the financing balance of ecological finance products has exceeded 100 billion yuan, an increase of more than 30% from the beginning of the year. For small, medium and micro enterprises, it has created a one-stop digital platform for enterprise operation and management, "Minsheng e-home", covering personnel, finance, salary, taxation and other functions.
Second, the construction of smart banks has been comprehensively promoted, and the operation and management level of the whole bank has been greatly improved. In terms of marketing, we have built a smart marketing brain, and used tools such as enterprise-level customer data platform (CDP), intelligent delivery and AB experiment to continuously improve our precision and intelligent marketing and omni-channel operation capabilities. In terms of risk control, we have built an intelligent risk control system, using big data, risk mapping and machine Xi technology to accurately identify and warn various risks, and built a unified data base and intelligent monitoring and early warning model for anti-money laundering and anti-fraud to protect the safety of customers' funds and business compliance. In terms of operations, the company has greatly improved business processing efficiency and customer satisfaction through straight-through acceptance, centralized operations, RPA, and remote services.
In the process of digital finance construction, Minsheng Bank has always taken the use of financial technology to better serve the real economy, support the growth of science and technology enterprises, enhance financial inclusion, and help rural revitalization as the starting point and end point of its work. For "specialized, special and new" science and technology enterprises, it has launched a fully online credit loan product "Yichuang E Loan", and at the same time created four platforms of "science and technology innovation evaluation, science and technology innovation products, science and technology innovation industry research and science and technology innovation ecology" to support the development and growth of science and technology enterprises. For small and micro customers, we have built an active credit intelligent decision-making system based on big data analysis, and launched the "Minsheng Hui" (legal person + individual) active credit product that is suitable for small and micro customers to help every small and micro dream. Facing the three rural areas, using big data technology, we have innovatively launched credit products that benefit farmers, such as "Agricultural Loan Connect" and "Cotton Farmer Loan", to serve rural long-tail customers with higher efficiency and lower marginal cost and help rural revitalization.
In the future, Minsheng Bank will fully implement the spirit of the Central Financial Work Conference, and on the basis of the results it has achieved, devote itself to the work of digital finance with a higher position, higher requirements and more enthusiasm.
21st Century: When it comes to the effectiveness of digital transformation, how do you see the relationship between technology and business?
Zhang Bin: We often hear the term "science and technology leading", and I think the correct understanding of this sentence should mean that the whole bank actively embraces science and technology and regards science and technology as the primary productive force, rather than a narrow Ministry of Science and Technology or CIO leadership. To put it more concretely, embracing technology means that the whole bank should have a digital mindset, use technology to optimize internal processes, improve customer experience, and make business decisions more scientific and effective.
Technology and business should be a side-by-side convergence that requires different perspectives and complementary abilities of the two types of people to get things done. There are many ways to promote the integration of the two, for example, the technology department sends demand analysts and data analysts to the business department, and hands over part of the evaluation of the scientific and technical personnel to the business department.
Talking about the potential of large models: comparable to the invention of the Internet
21st Century: With everyone's enthusiasm for large models at an all-time high, what impact will this have on the digital transformation of banks?
Zhang Bin: From the perspective of its impact on the future, the large model is considered to be comparable to the invention of mobile phones and the Internet. Different from the phenomenon that has occurred many times in the field of artificial intelligence technology in the past, such as Alpha Go, which quickly disappeared from public view after attracting widespread attention for a period of time, large models have a wide range of application fields, can handle complex tasks, and have been practical in some fields, and at the same time, through the introduction of richer training data and model architecture improvement, large models are in the process of rapid improvement. Personally, I believe that the breakthrough of large models in the field of artificial intelligence has injected new momentum into the digital transformation of banks.
The comprehension and content generation capabilities of large models, as well as the resulting interaction capabilities, can enable the digital transformation of the financial industry to move to a higher level of intelligence. Based on computing power and generative models, the MaaS model as a service is becoming a new paradigm for building applications. In some way, the vast majority of applications will be reconstructed, and the impact will be huge.
21st Century: From a purely business perspective, what financial scenarios are the current large models more suitable for, and what pain points can be solved?
Zhang Bin: Large language models have a wide range of applications in the financial field, which can help improve work efficiency, such as helping individuals become super producers, and can also improve the experience, such as supporting multiple rounds of dialogue, and machines interacting in a way that is close to people.
In specific application scenarios, such as intelligent customer service, combined with digital humans, it can be multi-round, highly anthropomorphic Q&A; in the field of intelligent investment research, it can use its powerful analysis and refining capabilities and generation capabilities; in the field of program development, it can assist in coding, testing and completion; in the field of marketing, it can help precision marketing, including personalized content generation; in the field of operation, it can assist in manual interaction. In the field of risk control, it can realize the intelligent risk identification and legal compliance, and in the knowledge management field, it can realize automatic knowledge extraction, knowledge update and maintenance, and provide a better knowledge Q&A experience.
It should be noted that in the process of using large models and generative AI, the application of small models or discriminative AI cannot be ignored. For different problems, large models and small models have their own advantages and applicability, and they will be used in combination in more cases in the future.
21st Century: The financial industry has high requirements for accuracy, controllability and security, what risks may be brought by the use of large models, and how to prevent them?
Zhang Bin: This is a very important question, in October this year, the National Information Security Standardization Committee released a draft of the "Basic Security Requirements for Generative AI Services", which lists 31 major security risks of corpus and generated content in the appendix, which are divided into five categories, including commercial violations of laws and regulations, infringement of the legitimate rights and interests of others, and inability to meet specific service types. This shows that the security and risk issues that may be brought about by generative AI have attracted great national attention.
In response to the problems of fact illusion, response consistency, and accuracy of large models, the industry has some common countermeasures: first, in the selection of application scenarios, first inside and then outside, and internally as an employee assistant in different positions. The technical measures are mainly to optimize the data set to ensure the quality, quantity and category of the data, parameter adjustment and prompt word engineering, expert feedback, including data annotation, evaluation and calibration of the results, etc. In addition, large language models can be used in conjunction with vector databases and knowledge graphs to improve the accuracy of responses.
The above measures are aimed at the weaknesses of the large model technology itself, and the risks of ideology, values, racial bias, and malicious poisoning should be comprehensively solved by increasing the endogenous security and use safety of the model. For the selection of corpus in the pre-training stage, supervised fine-tuning, and security filtering of input and output, a complete set of security specifications and evaluation systems are required.
21st Century: What regulatory issues may the application of large models face?
Zhang Bin: For the application of financial technology, financial regulators have always emphasized the need to adhere to "integrity and innovation". Regulators are observing the development and application of large-scale model technology, while paying close attention to the risks that may be bringed, which I personally believe reflects the "risk-based" regulatory philosophy. As mentioned earlier, the security risks of large language model applications are real, and the supervision of new technology applications is conducive to the steady and long-term application of generative AI in banks.
Talking about technology exploration: the combination of open source and closed source
21st Century: Does Minsheng Bank have any research on large-scale model technology?
Zhang Bin: For the research and application development of large models, Minsheng Bank has a certain foundation and accumulation in terms of talent team, technology platform and data foundation.
First of all, we have a professional NLP technical team, including algorithm experts introduced from leading Internet companies and with large model theoretical foundation and construction experience. NLP technology is highly related to the knowledge and skills required for data preparation, model training and tuning, and scenario application of large models.
Minsheng Bank has built an AI platform, including the NLP module and data annotation module, which provides good technical support for the implementation of large models in scenarios, and the data lakehouse platform that has been put into production provides a good foundation for efficient data pipeline construction. Widely developed NLP applications, such as knowledge bases, have accumulated a large number of preprocessed corpora, which can be directly used for the fine-tuning of large language models.
The cross-departmental large-scale model project team established by Minsheng Bank is directly led and promoted by the Digital Finance Leading Group of the Head Office. Through the construction of computing resource base, the introduction of model training and inference framework, the selection of large models, and the construction of application scenarios, a number of work has been carried out in an orderly manner, including six internal scenario applications such as knowledge base, code assistance, and intelligent agents.
21st Century: For the application of large models, which direction will Minsheng Bank choose in terms of technology selection path?
Zhang Bin: In terms of planning and strategy, we are scenario first inside and then outside, open source first, and open source and closed source combined.
Let's first use open source to build a complete framework, such as data preparation, data tuning, etc., and the model itself is pluggable. The iteration speed of the large model is too fast, we are not in a hurry to fix the large model, but first establish the overall capability of Minsheng Bank and find the application scenarios.
21st Century: From the perspective of banks, how do you view open source?
Zhang Bin: The logic of open source is very clear, the purpose of open source is mainly to rapidly expand the ecosystem, and closed source is more about "poking the ceiling" to promote the continuous development of large model capabilities. Open source, where the overall risk is manageable by controlling use cases and taking the security measures mentioned earlier, and there is also an open source commercial version that provides support and services. Closed-source products should still be used in some scenarios, mainly based on capabilities and cost performance.
21st Century: As mentioned earlier, generative AI has added new momentum to digital transformation, what other new technologies are worth looking forward to?
Zhang Bin: From the perspective of the combination of technology, the financial industry can divide digital transformation into two levels. One is the combination of traditional financial products and services with the Internet (web 2.0) and financial technology, which is manifested in the online, digital and intelligent financial services. The other is digitally native financial products and services, which are built on new digital infrastructures such as blockchain and next-generation Internet (Web 3.0), which can serve the development of the real economy and the digital economy more efficiently.
AIGC will not only help the first level of digital transformation move towards intelligence, but also greatly contribute to the evolution of web 3.0/metaverse. Today, the digital yuan has developed to the application stage of smart contracts, and it may be expected that the digital transformation of finance will be moving towards intelligence, and it is possible to develop more digital-native financial services with the evolution of web3.0.
For more information, please download the 21 Finance APP
![Zhang Bin, deputy general manager of Jinlong Automobile, was promoted all the way to the promotion, but now he is 58 years old[fig]](data:image/gif;base64,R0lGODlhAQABAIAAAP///wAAACwAAAAAAQABAAACAkQBADs=)