The development of large models has opened up a huge imagination space for enterprise application innovation. In the era of intelligence, the enterprise service model can be said to assume the role of the "operating system" of enterprise applications, making the intelligent capabilities of the technical base supporting enterprise applications more complete, promoting the upgrading of intelligent applications from the cognitive stage to the wisdom stage, helping enterprises to achieve intelligent operations, and making intelligence truly increase revenue for enterprises.
1The enterprise service model will become the "operating system" for enterprise-level AI applications
2023 is the first year of generative AI, and for companies that are actively seeking to transform their digital intelligence, the promise of generative AI technology is very enticing: generative AI tools can greatly simplify the process of low-end repetitive work, greatly reduce human input, stimulate content creativity, provide high-level business insights, assist in important decisions, and prevent risksWhen generative AI technology is fully integrated with the abundant data resources that enterprises already have, it can not only dig deep into the potential value of data, but also allow enterprises to focus more on business and product innovation with limited resources, laying the foundation for continuous competitiveness.
Dai Qionghai, an academician of the Chinese Academy of Engineering, once said in a public speech that broadening the boundaries of data, promoting algorithm innovation, and breaking the bottleneck of computing power will be the only way to bring application changes and lead breakthroughs in the foundation of artificial intelligence in the future. Generative AI based on large models is the latest development trend in AI technology and applications. Dai Qionghai**, the large model is expected to become a key basic platform for artificial intelligence applications in about 5 years, similar to the operating system of the PC era. As a leading enterprise digital intelligence software and service provider in the industry, Yonyou took the lead in proposing the concept of "digital intelligence enterprise" a few years ago, defining a new paradigm of digital intelligence business innovation and data-driven and intelligent operation driven by digital and intelligent technologies. Based on the accumulation of 35 years of experience in serving enterprises, Yonyou has integrated professional knowledge in various fields of enterprises and business knowhow in various industries, and launched the industry's first enterprise service model - Yongpt after pre-training and fine-tuning of field and industry data. Yongpt can understand and analyze all kinds of enterprise data, apply it to various business scenarios, and provide enterprises with intelligent human-machine collaboration, business insight, business decision support and intelligent operation services.
Yongpt is a good interpretation of the role of the "operating system" of enterprise intelligent applications. Yonyou believes that the role of large models should be integrated with the existing digital intelligence base of the enterprise, so that the capabilities and processes of the base can be upgraded in an all-round way. For enterprises, gaining generative AI capabilities is not the goal, but the ultimate goal is to reduce costs, increase efficiency, and accelerate innovation through the integration of leading technologies. The process of introducing generative AI should be "silent", and organizations prefer a smooth transformation process rather than a large-scale, rough escalation.
Yongpt provides functions such as data management, large model fine tuning, large model evaluation and optimization, large model inference and plug-in services through the large model service platform, providing stable and effective support for the construction and service of large models. Through the organic integration with other products of Yonyou BIP, Yongpt also innovatively combines the privatization data of the enterprise with the enterprise service model through a specific mechanism, which can not only solve the internal data security and privacy problems of the enterprise, but also make full use of the emerging reusable professional capabilities brought by the association of multiple fields and industries of the large model, so that the enterprise service is transformed from process-driven to semantic-driven based on the large model scheduling, bringing intelligent business operations and natural human-computer interaction to the enterprise. Intelligent knowledge generation and semantic application generation have become the source of energy for enterprise intelligent application innovation. 
Yongpt overall architecture.
2**The core technology of the industry's first enterprise service model.
Nowadays, the launch of a separate large-scale model product has required enterprises to have a considerable degree of technical strength, and the ability to integrate the large-scale model into the existing technology platform, and the comprehensive intelligent upgrade of the existing product matrix requires deep ability accumulation.
Yongpt's ability to empower enterprise intelligence is divided into two layers: general capabilities and application capabilities.
By optimizing the technical architecture and algorithms, Yonyou's general model base provides strong support for various mainstream general models in the industry, such as Baichuan Intelligence, Zhipu ChatGPT, Wenxin Yiyan, etc. This general-purpose large model base not only provides efficient computing and storage capabilities, but also has excellent scalability and flexibility, and can be customized according to different needs. To this end, Yonyou General Large Model Base also integrates a wealth of tools and libraries to help developers train, deploy, and optimize models more conveniently. Through integration and optimization with the industry's mainstream general large model, Yonyou General Model Base provides enterprises with more intelligent, efficient and reliable big data analysis and application services, helping enterprises in digital transformation and innovative development. 
Yongpt enterprise service model based on the general model.
Yonyou Yongpt's general capability layer has complete semantic understanding capabilities, enabling it to understand and generate natural language text, perform inference, identify entities, manage context, integrate with knowledge bases, and interact with users in meaningful ways
Yongpt's content generation capabilities can be used not only to generate articles, reports, and press releases, but also to automated report generation, translation services, data visualization for natural language generation, and the development of virtual assistants and chatbots
Multi-turn conversational capabilities, the ability to participate in multiple rounds of conversations in a natural and fluid way, maintain contextual consistency, understand user needs, and generate targeted responses. This capability offers great potential for a wide range of application scenarios;
Knowledge Q&A capabilities that can extract information from a wide range of knowledge sources and answer user questions. It has a wide range of knowledge coverage, including a variety of textual materials, encyclopedias, research**, and a large body of knowledge related to the field of business management, and is also able to deal with complex multi-step problems;
The role-playing ability allows it to simulate different virtual characters, engage in realistic conversations and interactions with users, and create a variety of situations and emotional representations. It includes different roles such as simulated architects, consultants, and customer service, as well as personalized role-playing experiences in different scenarios such as daily office, decision analysis, and customer support
Logical reasoning ability, which can identify and apply various logical rules, answer questions that require logical reasoning, such as data analysis, execution suggestions, etc., can also be used to solve complex problems, verify hypotheses and infer conclusions, and help users understand the essence of the problem more deeply
Generative capabilities, which can generate programs, including segments, scripts, and algorithms for various programming languages. And it can be combined with the digital modeling capabilities of YonBuilder, a low-quality development platform of YonBuilder, to generate first-class results that conform to the logic of the YonBuilder framework, which is of great value for reducing the R&D threshold of R&D personnel. Image generation capabilities that can generate a diverse range of images, including images of different themes, styles, and styles. This makes it very flexible and adaptable to different creative needs.
In the enterprise service scenario, many business scenarios are very complex, and Yongpt proposes a solution of "decision GPT", that is, decomposing complex tasks into task chains and scheduling decisions. In the past, traditional AI technologies were more direct to provide results, while large models gave enterprises the ability to help deduce the best results throughout the complex business decision-making process. Generative AI technology based on large models can provide enterprise employees with knowledge graphs, creative references, decision-making corrections and other capabilities in areas that rely on innovation output, such as content creation, human-computer interaction, and product design, and upgrade the innovation production chain of enterprises, so that enterprises can continuously and stably output innovation results and make the best decisions.
In terms of application capabilities, Yongpt first solves the shortcomings of the general large model in the identification of intent slots. For the intent slots with a large number of candidates, the model training into the model in batches + retrieving the relevant candidate slot words according to semantics is used to identify the intent slots, which greatly improves the accuracy of intent recognition in enterprise application scenarios.
In the scenario of business knowledge query and Q&A, Yonyou Yongpt combines the relevant capabilities of intelligent search to solve the search query and Q&A of multi-modal data. For example, it can quickly index multiple data types, accurately locate the context, and solve the illusion of large models generating questions and answers. In the professional field, the large model combined with the capabilities of digital intelligence employees can train expert robots in professional directions and provide professional services for employees, such as official document experts, legal expert robots, etc.
With its strong data analysis and development capabilities, natural language processing capabilities, knowledge integration capabilities and application generation capabilities, Yongpt has provided strong support for enterprises to achieve digital and intelligent transformation, and has played an important role in the field of production, operation and operation management of many enterprises. Yongpt has formed a full-scenario large-scale model application in the business fields of finance, human resources, ** chain, procurement, manufacturing, marketing, R&D, projects, assets, collaboration, etc., through which the large model can better understand business needs, make decisions more accurately, and ensure the practicability and effectiveness of the model.
Yongpt large model is used in all scenarios.
For example, based on market changes and historical data, Yongpt can intelligently perceive the correlation and attribution of data in various fields of production, sales and inventory of enterprises, and simulate and adjust relevant strategies, and calculate the profitability data of the next business cycle based on multiple factorsAnother example is the probationary period evaluation based on Yongpt, which can automatically generate the probationary evaluation of the employee and submit it for review according to the employee's work performance, goal achievement, daily collaboration, special work and other behavioral dataThe chain collaboration based on Yongpt can achieve network optimization. According to the historical consumption and demand**, the safety stock of different warehouse materials is dynamically calculated, and the target inventory and estimated order quantity indicators are calculated and recommended in a certain period in the future according to the set service level, so as to reduce and optimize the inventory cost under the premise of ensuring the level of customer service.
It is worth mentioning that Yongpt attaches great importance to the security protection of enterprise privacy data, and uses a variety of security combination architectures to eliminate the worries of enterprises when using large models.
3. Build a solid foundation, and promote the comprehensive application of innovative technologies in enterprises.
As a part of Yonyou IUAP, the foundation of enterprise digital intelligence, Yongpt draws on Yonyou IUAP's leading technology and platform capabilities, and also adds a weight to the intelligent upgrade of Yonyou IUAP. Yonyou IUAP helps enterprises improve their ability to control digital and intelligent technology. Based on the technology platform, business middle platform, data middle platform, intelligent middle platform, low-quality development platform and connection integration platform, it provides enterprises with the ability to build a middle platform, the ability to open and integrate the interconnection of hybrid cloud in a multi-cloud environment, the ability to quickly build applications such as low-level development and self-service of digital intelligence under technology inclusiveness.
After more than 20 years of continuous innovation, yonyou IUAP has developed into a digital and intelligent foundation for enterprises with better business understanding, leading technology and complete system. Among them, the IUAP intelligent middle platform carries the intelligent capabilities of IUAP, and provides three types of AI products and services: digital intelligent employees, intelligent big search and intelligent services, based on large models and service platforms, and intelligent technologies integrated with AI algorithms and knowledge graphs, and then acts on the cloud in the fields of finance, human resources, and procurement, providing intelligent empowerment for the innovation and development of various industries. With the support of the IUAP intelligent middle platform, the whole process of R&D, operation and business of the enterprise and every employee can be upgraded intelligently, and daily behavior, key decisions and results output are supported by data and intelligent analysis.
In 2023, in addition to the breakthrough in the enterprise service model, Yonyou IUAP will also achieve the improvement and iteration of leading technologies in multiple dimensions.
YMC Cloud Monitoring Center: Health management experts who support cloud-on-cloud and on-premise integration and remote intelligent consultation.
The YMC cloud monitoring center is divided into two parts: real-time monitoring on the cloud and daily monitoring on the local side, providing advanced monitoring, remote consultation, expert analysis, operation and maintenance optimization and other capabilities. The cloud monitoring center can report system problems to multi-stakeholder groups in a timely manner, and supports efficient root cause analysis of the whole link, multiple scenarios, and multiple environments. It is worth mentioning that the cloud monitoring center can obtain the help of UFIDA's online experts at any time, conduct in-depth analysis of monitoring reports, and discuss countermeasures with the enterprise IT department to eliminate problems in the bud.
Tenant domain sharding: Provides a highly elastic resource isolation mode with higher performance, more flexible deployment, and lower TCO
Database is a key component of the digital intelligence foundation of enterprises, and the database user segmentation method directly affects the efficiency of database use. Yonyou IUAP platform provides YMS cloud middleware, and innovatively implements a shared database architecture of database sharding by domain. In this database sharding mode, the technology middle platform, application middle office, business middle platform, human resources, finance and other departments can scold the database according to their purpose and share massive data resources of the database, and the middleware also supports the exclusive mode of database sharding by tenant. This design results in higher performance and lower TCO, which can meet the needs of more flexible application deployment and maximize the use of the platform's microservices infrastructure resources.
Migration family: Personalized and highly reliable, speeding up application rollout and iterative upgrades.
In the enterprise application development process, applications often need to be migrated back and forth between the development environment, the test environment, and the production environment. Yonyou IUAP platform has developed a complete set of migration technology stacks, including development migration, environment migration, configuration migration and file migration, so that developers can smoothly and seamlessly share results in different environments, and greatly accelerate the speed of application launch and iterative upgrade while meeting personalized and highly reliable requirements.
In addition, yonyou IUAP pioneered a continuous delivery system that integrates cloud and off-cloud, allowing enterprise private cloud platforms to experience the update efficiency of the public cloud, and making cloud application upgrades as simple as AppStoreThe self-developed multi-dimensional data engine (integrated storage and computing) realizes 100% autonomy, security and controllability, and supports the rapid merging of "multi-criteria, multi-currency, and main tables" under the scale of hundreds of billions of data, and one-click table output. This technology has been verified in large central enterprises, and has realized the rapid consolidation and one-click output of the statements of 1,500 super-large enterprises with the scale of 1,500 subsidiariesIt has realized the safe and credible adaptation of localized information and innovation, and provided enterprise customers with stable, credible, independent and controllable digital intelligence platform services. For example, the multi-dimensional engine database has completed the localized chip test, achieving hundreds of billions of data, 10,000-level concurrent retrieval, and millisecond-level response.
Technology is the source of business innovation, and based on leading technology, Yonyou IUAP will precipitate technology, tools, platforms, services, as well as profound knowledge accumulation and practical experience, and help enterprises successfully implement digital intelligence in the way of digital intelligence base. The enterprise digital intelligence foundation provides a solid platform support for enterprises to realize data-driven and intelligent operations. At the same time, when AI enters the popularization of applications, the enterprise service model may become the "operating system" of the new era, injecting intelligent energy into the digital intelligence base, bringing more possibilities for enterprises to carry out intelligent application innovation, and helping enterprises to control the future of digital intelligence.