金融科技正以前所未有的速度助推银行业数字化转型，在客户需求快速变化、产品创新周期不断压缩、技术架构面临持续升级的时代背景下，“敏捷”被视作数字化创新能力的代名词，成为衡量银行竞争优势的关键能力。数字化转型网szhzxw.cn
银保监会年初发布的数字化转型行业指导意见中指出，要建立快速响应需求的敏捷研发运维体系，推动科技管理敏捷转型，提高对大规模科技队伍和复杂技术工程的管理能力。农业银行主动拥抱数字化转型，构建“业、技、数”深度融合的敏捷研发体系，打造产品快速迭代创新能力。数字化转型网szhzxw.cn
一、不懈探索，寻求大行敏捷之道
1.重新定义数字化转型赋予敏捷研发的全新内涵。在数字化转型和金融科技双重作用下，银行业所追求的敏捷已从一种单纯迭代研发模式转向全方位敏捷化的IT基础实施和工程管理能力。银行正在将敏捷内化到技术基因，融入科技管理流程，敏捷研发将成为与数字化经营相适应的最主要创新模式。
2.量身打造独具国有大行特色的敏捷研发体系。敏捷的最终目的是提升响应力。落地敏捷没有固定套路和捷径，必须寻找适应自身经营特点和战略定位的道路。多年来，农业银行在敏捷研发方面开展了大量的理论研究和实践探索，我们的最大感悟是：敏捷研发从“小规模试点”到“大范围复制”之间存在量变到质变的差距，依赖于完全不同的底层逻辑。对于产品条线众多、组织层级复杂的国有大行而言，大规模可复制的敏捷研发能力源于技术架构、业务架构、研发工具、管理流程、协作机制五项要素的共同作用，进而实现系统化提升。只有以坚实的技术底座和先进的开发运维工具为基础，敏捷研发才能从“小而美”发展为“大而全”。数字化转型网szhzxw.cn
3.促进形成敏捷研发与安全发展之间的良性循环。高频引入大量新技术成为科技部门面临的常态，国有大行的内部经营策略和日益严峻的外部安全形势决定了敏捷绝不能以牺牲质量和影响安全为代价。敏捷与质量、敏捷与安全要形成互补关系，而非对立。一方面敏捷研发要以质量和安全为前提，另一方面敏捷研发要为质量和安全赋能。要通过快速反馈、质量内建，形成开发、测试、运维之间的质量闭环，及早发现质量问题，降低修复成本。数字化转型网szhzxw.cn
二、整体布局，构建敏捷研发体系
1.推动云原生技术栈转型，为敏捷交付提供坚实技术底座。全面转型云原生是提升敏捷交付能力的基础。采用云原生技术可简化大量异构应用的系统部署复杂度，从底层打通大型银行固有的跨产品线技术壁垒，使研发团队专注于业务逻辑开发。农业银行推进面向云原生2.0的分布式架构转型，构建立体化微服务技术体系。建设一体化金融云平台，利用容器化、声明式API和不可变基础设施等技术，打造运行安全稳定、资源灵活调度、应用敏捷弹性的IT基础设施云底座，自2022年起，基于Java技术栈的新增应用100%上云。打造以应用为中心的云原生技术栈，通过微服务、服务网格、Serverless等技术降低业务组件的共享复用成本，满足快速迭代、敏捷交付的效率要求。数字化转型网szhzxw.cn
2.引进自动化开发运维工具，为敏捷研发提供效能保障。建设DevOps工具链，加速软件交付。引入CI/CD流水线，将软件交付过程标准化、自动化，串联打通编译、构建、测试、部署环节，有效降低高频次迭代开发的“非增值性”固定成本。农业银行DevOps流水线已覆盖8个主流技术场景，接入应用系统模块超过300个，流水线总数超过8000条。建立适应农行实践的本地化持续交付标准，配套发布系列工具解决方案、规范手册和最佳实践。建设低代码开发平台，提高软件编码效率。封装公共组件、参数灵活配置，降低硬编码工作量，提升研发人员的单位产能。农业银行以轻云平台为低代码前端开发框架，实现研发快速上手，应用一键发布;后端开发推广BFC金融服务平台，集“参数化开发、共性组件复用、便捷测试、标准化交付”为一体，构建“37+1”总分创新研发模式，累计支持全行2000多个项目快速迭代开发。建立InnerSource内源技术生态，促进资源共享复用。依托农业银行“应用商店”充分盘活总分行研发过程资产，发挥“群智”效应，避免重复造轮子。数字化转型网szhzxw.cn
3.搭建企业级业务架构，为模块化产品设计提供统一标准。利用企业级业务架构建立产品创新的标准化设计语言，实现“书同文、车同轨”。持续完善企业级业务架构，厘清全行产品领域资源能力，推动企业级产品流程贯通、数据共享、能力互促。建立统一产品视图，维护30余项公有属性和300余项专有属性，形成15大类对客可售产品统一目录，采用“1+N”映射模式满足各领域产品目录的差异化使用需求。数字化转型网szhzxw.cn
4.运用精益理念重塑研发流程，提升复杂技术工程管理能力。建立适配敏捷化技术体系的项目管理机制，构建以价值为导向、以迭代为特征的研发流程。借鉴“精益生产”方法论，形成涵盖“一套敏捷研发流程+两个能力模型+三类知识集+四项行动指引”的“1234”敏捷项目管理机制。将MVP拆分、用户故事等方法应用于需求研制过程，超过20%的项目实现需求条目化;将“四会”、团队契约等方法融入迭代开发过程，迭代周期控制在2~4周。
5.健全多方协同创新机制，提高大规模技术队伍管理水平。针对大型银行研发队伍规模大、分支机构多等组织架构特点，提升业技数融合、总分支联动的协同创新能力。推动科技前台化，组建上千人的售前工程师团队，建立覆盖产品创新全生命周期、客户营销全过程的业技融合机制。围绕重要客户和重大项目成立相对固定的柔性团队，“客户、技术、产品”多方高效协同作战。推行金融科技创新实验室、多级联动创新支持等，着力解决跨地域、跨组织层级的大规模敏捷协作难题。数字化转型网szhzxw.cn
三、积厚成势，激发数字化转型新动能
1.业务响应能力提升。借助需求条目化、组建柔性团队、接入DevOps流水线等手段，敏捷研发机制显著加快了端到端的产品创新效率，缩短了产品交付周期。相较传统瀑布模式，敏捷项目需求研制周期缩短50%以上，关键领域产品具备每周多次投产发布能力。
2.科技供给能力提升。依托基础设施云化和低代码平台推广，科技研发产能不断增大，交付项目总量保持20%以上年均增幅，单人研发效率连年处于高位。通过新生产力工具的全面应用，突破了传统手工作业模式下的研发效能极限。
3.价值创造能力提升。聚焦高价值业务领域的资源倾斜，网络金融等重点领域超过50%的项目实行敏捷管理模式。以掌银为例，2021年按照“敏捷迭代，快速交付”模式，先后发布五大里程碑版本，助力掌银MAU突破1.5亿，位居同业首位。实施客户体验优化敏捷项目群，为掌银客户自助注册等重点应用，建立从客户行为采集到产品功能优化的互动闭环，先后完成12轮次迭代优化，掌银自助注册成功率提升近5倍。数字化转型网szhzxw.cn
二、夯实基础，敏捷研发高质量发展
1.加速技术升级。坚持“把复杂留给自己，把简单留给用户”，做厚敏捷研发技术密度，锤炼用标准化产品解决个性化需求的能力。按新一代技术栈建设规划，完善人机交互、应用协同等6大子域平台的全栈式技术中台;继续体系化布局基础设施云原生、架构云原生、开发运维云原生与安全云原生，推进技术体系从OnCloud向InCloud全面转型。构建总分行一体化敏捷研发体系，加快向分行推广DevOps自动化工具集和CI/CD流水线。
2.改进研发质量。依托DevOps质量门禁，实现更精准高效的质量管控。将自动化检查、测试工具嵌入敏捷研发过程，前移质量管控环节，建立“组织级+自定义”的多级质量管控体系。通过数据沉淀和度量分析，促进研发效能改进。基于质量门禁收集度量数据，量化分析研发测试过程。加快探索研发上云(CloudIDE)，将DevOps工具链由软件交付阶段向前延伸到编码阶段，强化端到端质量控制。数字化转型网szhzxw.cn
3.保障安全运行。筑牢敏捷研发安全底线，保障金融科技高质量发展。完善灰度发布和蓝绿发布环境，健全发布机制，在保持生产运行稳定的同时，支持对用户体验的快速反馈。加快实现“运行、运维、运营”领域的数字化管理，利用算法分析引擎、BI可视化分析引擎赋能智能运维场景。健全SDL体系，将安全规范内置到敏捷项目管理各环节，加固软件研发全过程的安全性。建设DecSevOps，将适配敏捷研发技术体系的安全工具嵌入DevOps工具链，实现效率与安全并重。
翻译：数字化转型网szhzxw.cn
Fintech is boosting the digital transformation of banking industry at an unprecedented speed. Under the background of rapid changes in customer demand, continuous compression of product innovation cycle and continuous upgrading of technical architecture, “agility” is regarded as a synonym for digital innovation ability and a key ability to measure banks’ competitive advantages.
At the beginning of the year, China Banking and Insurance Regulatory Commission issued a guideline on the digital transformation industry, pointing out that it is necessary to establish an agile R&D operation and maintenance system that responds quickly to demands, promote the agile transformation of science and technology management, and improve the management ability of large-scale science and technology teams and complex technical projects. Agricultural Bank of China actively embraces digital transformation, builds an agile research and development system with deep integration of “industry, technology and data”, and builds rapid product iteration and innovation capability.数字化转型网szhzxw.cn
Unremitting exploration, seeking a big line agile way数字化转型网szhzxw.cn
1, Redefining digital transformation gives new connotation to agile R&D.
Under the dual role of digital transformation and fintech, the agility pursued by the banking industry has changed from a simple iterative research and development mode to an all-round agile IT foundation implementation and project management ability. Banks are internalizing agility into technical genes and technology management processes. And agile R&D will become the most important innovation mode suitable for digital management.
2, Create an agile R&D system with unique national characteristics. 数字化转型网szhzxw.cn
The ultimate goal of agility is to improve responsiveness. There is no set formula or shortcut to landing agile. It must find a way to adapt to its own business characteristics and strategic positioning. Over the years, Agricultural Bank of China has carried out a lot of theoretical research and practical exploration in agile R&D. Our biggest insight is that there is a quantitative change to qualitative change in agile R&D from “small-scale pilot” to “large-scale replication”. Which depends on completely different underlying logic.
For large state-owned banks with many product lines and complex organizational levels. Large-scale replicable agile R&D capabilities are derived from the combined action of five elements: technical architecture, business architecture, R&D tools, management process and collaboration mechanism. And thus achieve systematic improvement. Only with a solid technical base and advanced development and operation tools as the foundation. Agile R & D can develop from “small and beautiful” to “large and complete”.
3, Promote a virtuous cycle between agile R&D and safe development. 数字化转型网szhzxw.cn
The introduction of a large number of new technologies at high frequency has become the norm for the science and technology sector. The internal business strategy of the state and the increasingly severe external security situation decide that agility should not be sacrificed at the cost of quality and safety. Agile and quality, Agile and security should be complementary, not opposed. On the one hand, agile R & D should take quality and safety as the premise; on the other hand, agile R & D should empower quality and safety. Through rapid feedback and quality build-in, a quality closed loop between development, testing, operation and maintenance should be formed to find quality problems early and reduce repair costs.
Overall layout, the construction of agile R & D system
1, Promote the transformation of cloud native technology stack to provide a solid technical base for agile delivery.
The full transformation of cloud native is the foundation of agile delivery. The adoption of cloud native technology can simplify the system deployment complexity of a large number of heterogeneous applications, break through the inherent cross-product line technical barriers of large banks from the bottom, and enable the R&D team to focus on the development of business logic. Agricultural Bank of China promotes the distributed architecture transformation of cloud native 2.0 and builds a three-dimensional micro-service technology system.
An integrated financial cloud platform will be built, and technologies such as containerization, declarative API and immutable infrastructure will be utilized to build a cloud base of IT infrastructure with secure and stable operation, flexible resource scheduling, and agile and flexible application. Since 2022, 100% of new applications based on Java technology stack will go to the cloud. Build an application-centered cloud native technology stack, reduce the cost of sharing and reuse of business components through micro-service, service grid, Serverless and other technologies, and meet the efficiency requirements of fast iteration and agile delivery.数字化转型网szhzxw.cn
2, Introduce automated development operation and maintenance tools to provide efficiency guarantee for agile R&D.
Build DevOps toolchains to speed up software delivery. CI/CD pipeline is introduced to standardize and automate the software delivery process, connect the links of compilation, construction, testing and deployment in tandem, and effectively reduce the “non-value-added” fixed costs of high-frequency iterative development. The DevOps pipeline of Agricultural Bank of China has covered 8 mainstream technical scenarios, and more than 300 application system modules have been connected, with a total of more than 8,000 pipelines. Establish local continuous delivery standards that adapt to ABC’s practices, and release a series of tool solutions, specification manuals and best practices. Construction of low code development platform, improve the efficiency of software coding. 数字化转型网szhzxw.cn
Encapsulation of common components, flexible configuration of parameters, reduce hard coding workload, improve the unit capacity of R&D personnel. Agricultural Bank of China takes the light cloud platform as the front-end development framework of low code to realize the quick start of research and development and one-click application release; Back-end development and promotion of BFC financial service platform, set “parametric development, common component reuse, convenient testing, standardized delivery” as one, to build a “37+1” total score innovative research and development mode, cumulated support for the rapid iterative development of more than 2,000 projects in the bank. Establish InnerSource technology ecology to facilitate resource sharing and reuse. Relying on the “App Store” of Agricultural Bank of China fully revitalize the R&D process assets of the head branch, give full play to the “group wisdom” effect, and avoid repeating the wheel.
3, Adopt the “1+N” mapping mode to meet the differentiated use requirements of product catalogs in various fields.
Build enterprise-class business architecture to provide unified standards for modular product design. Using enterprise business architecture to establish a standardized design language for product innovation, to achieve “the same book, the same track”. Continue to improve the enterprise-level business architecture, clarify the resources and capabilities of the whole bank in product areas. And promote the integration of enterprise-level product processes, data sharing, and mutual promotion of capabilities. Establish a unified product view, maintain more than 30 public attributes and more than 300 proprietary attributes. And form a unified catalog of 15 categories of products available to customers. Adopt the “1+N” mapping mode to meet the differentiated use requirements of product catalogs in various fields.数字化转型网szhzxw.cn
4, Apply lean concept to reshape R&D process and improve management ability of complex technical projects.
Establish a project management mechanism suitable for the agile technology system. And build a value-oriented and iterative research and development process. The “1234” Agile project management mechanism covering “one set of agile R & D process + two capability models + three knowledge sets + four action guidelines” is formed by referring to the “Lean production” methodology. Methods such as MVP split and user story were applied to the requirement development process. And more than 20% of the projects realized the requirement item; Methods such as “four meetings” and team contract are integrated into the iterative development process. And the iteration cycle is controlled within 2 to 4 weeks.
5, Improve the multi-party collaborative innovation mechanism and improve the management level of large-scale technical teams.
Aiming at the organizational structure features of large banks. Such as large R&D team size and multiple branches, the collaborative innovation capability of integration of industry technology and total branches will be improved. Promote the foreground of science and technology, set up a pre-sales engineer team of thousands. And establish an industry-technology integration mechanism covering the whole life cycle of product innovation and the whole process of customer marketing. Set up a relatively fixed flexible team around important customers and major projects. And cooperate effectively with “customers, technology and products”. Fintech innovation LABS and multi-level linkage innovation support have been implemented to solve the problems of large-scale agile collaboration across regions and organizational levels.
Build up momentum and stimulate new driving force for digital transformation数字化转型网szhzxw.cn
1, Service response capability is improved. Agile R&D significantly speeds up end-to-end product innovation efficiency and shortens product delivery cycles through requirements entry, flexible teams, and DevOps pipeline access. Compared with the traditional waterfall model, the development cycle of agile projects is more than 50% shorter. And the products in key areas can be put into production and released several times a week.
2, Improving the supply capacity of science and technology. Relying on cloud infrastructure and low-code platform promotion. The capacity of scientific and technological research and development has been increasing. The total number of projects delivered has maintained an average annual growth rate of more than 20%. And the single-person research and development efficiency has been at a high level for years. Through the comprehensive application of new productivity tools. The research and development efficiency limit of traditional manual operation mode has been broken.数字化转型网szhzxw.cn
3, Improved value creation ability. Focusing on the resource tilt of high-value business areas, more than 50% of projects in key areas such as network finance implemented agile management mode. Take Palm Silver as an example. In 2021, according to the “agile iteration, fast delivery” model, five milestone versions have been released successively, helping Palm Silver to exceed 150 million MAU, ranking first in the industry. The agile project group of customer experience optimization was implemented. And the interactive closed loop from customer behavior collection to product function optimization was established for key applications. Such as self-service registration of Palm Silver customers. 12 rounds of iterative optimization were completed successively. And the success rate of self-service registration of Palm Silver was increased nearly 5 times.
Solid foundation, agile research and development of high quality development
1, Accelerate technology upgrading. 数字化转型网szhzxw.cn
Adhere to the “complex to oneself, the simple to users”, do thick agile research and development technology density, temper the ability to use standardized products to solve personalized needs. According to the construction plan of new-generation technology stack. We will improve the full-stack technology center platform of six sub-domain platforms. Including human-computer interaction and application collaboration. Continue to systematize the layout of infrastructure cloud native, architecture cloud native, development and maintenance cloud native and security cloud native, and promote the comprehensive transformation of technology system from OnCloud to InCloud. Establish an integrated agile R&D system for the head office branch. And accelerate the promotion of DevOps automation tool set and CI/CD pipeline to branches.
2, Improve the quality of R&D.
Relying on DevOps quality access control, more accurate and efficient quality control can be achieved. The automated inspection and testing tools are embedded in the agile R&D process. And the quality control link is moved forward to establish a multilevel quality control system of “organization level + custom”. Promote R&D performance improvement through data precipitation and measurement analysis. Collect measurement data based on quality access control, and quantitatively analyze research and development testing process. Accelerate the exploration and development of CloudIDE, extend the DevOps tool chain from the software delivery stage to the coding stage, and strengthen end-to-end quality control.
3, Ensure safe operation. 数字化转型网szhzxw.cn
Secure the bottom line of agile R&D and ensure the high-quality development of fintech. Perfect gray release and blue and green release environment, perfect release mechanism. While maintaining stable production operation, support rapid feedback to user experience. Accelerate the realization of digital management in the field of “operation, operation and maintenance”, and use algorithm analysis engine and BI visual analysis engine to enable intelligent operation and maintenance scenes. Improve the SDL system, embed the safety standards into each link of agile project management. And reinforce the security of the whole process of software development. DecSevOps was constructed to embed security tools adapted to agile R&D technology system into the DevOps tool chain to achieve both efficiency and security.
本文由数字化转型网（www.szhzxw.cn）转载而成，来源：金融电子化；编辑/翻译：数字化转型网宁檬树。
免责声明: 本网站(http://www.szhzxw.cn/)内容主要来自原创、合作媒体供稿和第三方投稿，凡在本网站出现的信息，均仅供参考。本网站将尽力确保所提供信息的准确性及可靠性，但不保证有关资料的准确性及可靠性，读者在使用前请进一步核实，并对任何自主决定的行为负责。本网站对有关资料所引致的错误、不确或遗漏，概不负任何法律责任。
本网站刊载的所有内容(包括但不仅限文字、图片、LOGO、音频、视频、软件、程序等) 版权归原作者所有。任何单位或个人认为本网站中的内容可能涉嫌侵犯其知识产权或存在不实内容时，请及时通知本站，予以删除。
