"AI+" signifies the deep integration and application of artificial intelligence with other industries or fields, which means empowering traditional industries with new capabilities through AI technology, thereby generating new value and business models. This indicates that AI is not merely an isolated technology, but a powerful enabling tool that can be combined with various industries to improve work efficiency, optimize business processes, and create entirely new products and services.
Readers can understand "AI+" as "AI+X", where "X" represents any industry or field, such as "AI+Healthcare" (covering application scenarios like AI-assisted diagnosis, drug R&D, and personalized medicine), "AI+Finance" (including intelligent risk control, fraud detection, and robo-advisory), and "AI+Education" (such as personalized learning, intelligent tutoring, and automatic grading), among others.
"AI+" reflects the development trend of the deep integration of AI with various industries, which can strongly promote the transformation and upgrading of traditional industries while giving rise to new business models and economic growth points. This is not a simple technological superposition, but a comprehensive revolution of traditional industry processes and models, with the ultimate goal of reducing costs, improving efficiency, enhancing user experience, and creating new value.
This section will focus on how to promote the organic integration of AI technology and the business middle platform, thereby helping enterprises achieve innovative development while improving user experience and productivity.
Intelligent Business Acceptance
Cloud service acceptance business is more complex than general commodity sales, mainly reflected in the following aspects:
1) Complex product system with high costs for flexible configuration and selection. Cloud service products cover multiple technical fields such as computing, storage, networking, databases, AI (Artificial Intelligence), and security. Each field is further divided into various types (for example, the computing field includes general-purpose, memory-optimized, and high-performance computing instances). Users need to match their needs from a massive range of products. While the flexibility of cloud products (such as custom configuration of parameters like CPU/memory/storage/network bandwidth) is an advantage, it also requires users to deeply understand the business impact of technical parameters to balance performance and cost. This poses a high threshold for users without a technical background, easily leading to configuration errors or resource waste.
2) Diverse pricing models with significant difficulty in cost prediction. Cloud products generally support multiple billing methods such as pay-as-you-go, reserved instances, and monthly/yearly subscriptions. The discount rules, applicable scenarios, and cost structures for each method vary significantly. For example, reserved instances can enjoy higher discounts but require long-term resource locking, while pay-as-you-go offers high flexibility but with higher unit costs. If users lack long-term analysis of their business workloads, especially when using them for the first time, they are highly likely to incur cost overruns or low resource utilization due to improper selection of billing methods, making it difficult to accurately estimate the Total Cost of Ownership (TCO).
3) Strict compliance requirements with a high threshold for understanding contract terms. Cloud service contracts involve a large number of legal terms and technical specifications, including SLA (Service Level Agreement), data sovereignty attribution, privacy protection clauses, and detailed rules for liability division. Users need to accurately understand the rights and obligations in the terms (such as the definition of liability for data breaches and the compensation mechanism for service interruptions) to avoid legal risks. Faced with the complexity of cloud service acceptance business, most cloud service providers establish the position of "Account Manager" to provide pre-sales consulting and after-sales support for customers. Taking CTYun (Tianyi Cloud) as an example, account managers not only undertake problem-solving responsibilities but also provide proxy ordering services.
However, this model essentially shifts the operational pressure from the user side to the service side, without fundamentally solving the core problems such as complex product selection, difficulty in understanding pricing, and obscure contract terms. Based on this, the concept of realizing the intelligence of the acceptance process relying on AI technology emerged, aiming to generate precise product configuration solutions based on user needs through guided interactive processes.
After in-depth research, an intelligent acceptance solution based on AI large language models and centered on natural language interaction—the "Intelligent Assistant"—was proposed. This solution is committed to achieving the following goals:
1) Personalized service upgrade. Abandon the standardized service model, build a personalized intelligent assistant system, and provide each account manager with work assistance and reminders tailored to their functions and interests, achieving a highly personalized service experience for every individual.
2) Enhancement of self-service acceptance capabilities. Strengthen the self-service capabilities of account managers, promote the acceptance process towards 100% self-service, and reduce reliance on manual operations.
3) Process automation revolution. Upgrade the traditional manual click-driven mode to a proactive guidance mode driven by large models, simplify operation processes, lower the usage threshold, and significantly improve the user operation experience.
4) Conversational business processing. Guide customers to complete the full-process acceptance through natural language interaction, while providing real-time answers to professional questions such as legal terms and SLAs.
5) Best practice recommendation. Based on user needs, budgets, and business scenarios, generate optimal solutions covering deployment plans, cost estimation, cloud product portfolios, and more. The implementation of the Intelligent Assistant deeply relies on three major platforms: the business middle platform, the data middle platform, and the AI middle platform. Its technical architecture is shown in Figure 1. Among them, the AI middle platform and the business middle platform form a synergistic and symbiotic technical closed loop:
On the one hand, AI model training relies on standardized business data output by the business middle platform (such as order information, resource status, account systems, and knowledge base documents). After completing data access, cleaning, desensitization, and structured storage through the data middle platform, a high-quality training dataset is formed, supporting the model in accurately parsing user intentions and generating compliant solutions.
On the other hand, the business middle platform calls the intelligent acceptance model of the AI middle platform through API interfaces, embedding intelligent capabilities such as natural language interaction and configuration recommendation into the acceptance process, realizing the upgrade from "manual operation" to "model-driven".
Figure 1 Technical Architecture of the Intelligent Assistant
Readers may be more concerned about the source of the large model on which the Intelligent Assistant relies. This model is an enterprise-level AI product independently developed by CTYun, and its full lifecycle management and continuous optimization are handled by a dedicated AI middle platform team.
As a general-purpose intelligent foundation, its application scenarios are not limited to intelligent acceptance, but have also been deeply integrated into core business systems such as code management platforms, knowledge bases, and intelligent customer service, achieving cross-scenario reuse and synergistic efficiency of AI capabilities. The core logic for choosing an independently developed large model stems from the data security strategy. Taking the intelligent acceptance scenario as an example, model training requires calling highly sensitive data such as resource configuration and account systems. If third-party API interface services are adopted, although initial technical costs may be reduced, there is a significant risk of leakage of sensitive data during cross-platform transmission and storage. Although private deployment can partially solve the problem of data overflow, it has inherent shortcomings in system maintainability and business adaptability, and cannot meet the requirements of enterprise-level data security governance and technological autonomy.
By building a technical closed loop of "independently developed large model + privatized training environment", CTYun has realized the localized control of the entire link of "collection-training-application" for sensitive data, ensuring data sovereignty while ensuring the dynamic and precise matching of model iteration and business needs. One of the core functions of the Intelligent Assistant is to provide users with a personal assistant function, as shown in Figure 2. After entering the conversation page with AI, users can intuitively see the business information within their scope of concern or authority, specifically including key content such as pending approval of discount applications, cloud resources due for renewal, applicable available discounts, and product price reduction notifications.
Figure 2 Personal Assistant Function Provided by the Intelligent Assistant
Compared to business acceptance through dedicated front-end pages, the processing efficiency of the Intelligent Assistant has more advantages. Taking the resource renewal scenario as an example, users only need to provide the target resource ID to trigger the operation, as shown in Figure 3. The built-in intelligent verification mechanism of the system can automatically verify the accuracy of the input information, and only execute the subsequent process when the resource ID is legal and valid, avoiding operation failures caused by erroneous data.
In addition, permission control is also one of the core considerations in the design of the Intelligent Assistant. For example, Account Manager A can only perform renewal operations on the resources of customers under their name. The system achieves fine-grained access control through the permission management service of the business middle platform. Specifically, the business middle platform provides unified interfaces for identity authentication, role permission verification, and resource ownership determination. The Intelligent Assistant reuses the permission logic by calling these standardized services, without the need for repeated development.
Figure 3 Resource Renewal via the Intelligent Assistant
For various questions encountered by users during the acceptance process, such as process instructions, term explanations, and laws and regulations, professional answers can be obtained through the Intelligent Assistant. For users who are new to cloud products, the Intelligent Assistant provides a configuration recommendation function, which can generate adapted product configuration solutions based on user business needs and complete preliminary cost estimation, as shown in Figure 4.
Although the recommended configurations are for reference only and users still need to adjust them according to actual needs and budgets, compared to the traditional operation process that requires consulting multiple product documents or frequently consulting product managers, the Intelligent Assistant significantly reduces interpersonal communication costs and can maintain objective and neutral recommendation outputs.
Figure 4 Configuration Recommendation by the Intelligent Assistant Based on User Needs
The user group of the Intelligent Assistant is not limited to account managers; operation personnel can also use it to simplify business operation work. For example, when an exception occurs in the acceptance process, operation personnel can upload the error message text or screenshots to the Intelligent Assistant to trigger the preliminary problem diagnosis function, as shown in Figure 5.
Figure 5 Preliminary Diagnosis of Business Operation Problems via the Intelligent Assistant
The efficient operation of the Intelligent Assistant relies on continuous model training. In addition to high-quality business data, materials accumulated in the business middle platform, such as requirement specifications, detailed design documents, and Standard Operating Procedures (SOPs), can be converted into model training materials through uploading on the management page or importing via links.
After nearly two years of iterative optimization, in the business acceptance scenario, the Intelligent Assistant has achieved 100% coverage of business scenarios, accumulated a knowledge base of over 500 documents, and the daily processed acceptance orders account for 70% of the total acceptance volume, directly promoting an increase of over 50% in related work efficiency.
Currently, the R&D team is further optimizing the system by supplementing knowledge documents and expanding functional modules, with the goal of increasing its application coverage rate to over 90%.
The above content is excerpted from "Digital Transformation: Practice of Business Middle Platform Construction" by Sun Lianshan, Huang Chuanguang, Zhang Jie, and He Yanhui.
Recommended Reading
▊ "Digital Transformation: Practice of Business Middle Platform Construction"
Sun Lianshan, Huang Chuanguang, Zhang Jie, He Yanhui
In the book, CTYun systematically summarizes how to transform complex cloud and network resources into efficient and standardized digital services through the architectural decoupling and capability atomization of the business middle platform. This is not only a profound reflection of the transformation of China Telecom's IT architecture from "support-oriented" to "empowerment-oriented", but also a practical extraction of how the entire network can achieve the ultimate synergy between computing power and business logic. This practice of precipitating core technical capabilities, modularizing and encapsulating them, and transforming them into service value not only improves the transformation efficiency of scientific and technological achievements, but also provides a highly reference-worthy model for cultivating core competitiveness in the digital era. At the current critical node of enterprise digital transformation, this book provides a strong practical guidance model for practitioners.
This book is suitable for practitioners in related industries such as finance and technology to read and study, and also for teachers and students in related majors such as mathematics, finance, and computer science for reference reading.
Author: Li Shan
Managing Editor: Zhang Shuqian
Reviewer: Cao Xinyu