Alibaba B-side case! Cainiao Intelligent Design Middle Platform Design Review

Before we start this topic, we need to have a basic understanding of the design platform. Readers can learn about Alibaba's thinking and tool progress on the design platform R&D model and design engineering through the article "SEE Conf Design Engineering Trilogy! Exploring the Thinking and Practice of "Production and Research Collaboration Model" in the New Environment". We manage multi-role collaboration and consumable materials in the product production chain, accelerate production efficiency and reduce costs, and at the same time, the product experience baseline can be guaranteed.

The focus of our discussion today is not on system product production, but on the user experience trends of B-side systems and Cainiao's entry point in this field. We will discuss this topic in two phases. The first phase will discuss the intelligent experience trends of B-side systems, and the second phase will discuss the changes brought about by the application of intelligent capabilities to the experience.

The trend and experience demands of B-side systems

First, let me take a look at the development history of the B-side system at the product level :

• In the 1980s, companies such as IBM had already provided computer-based management systems to implement closed-loop management systems for enterprise raw material management, production and processing management, and employee work management.
• In the 1990s, such management systems became more mature and could realize financial forecasting, production capacity and resource scheduling, and truly became a system tool that could be used for product quality management, resource management, and financial management, assisting enterprise managers in making decisions. We call such a system ERP (Enterprise Resource Planning).
• Since 2000, the maturity of Internet technology has enabled corporate systems to exchange data with supply chains and customer upstream and downstream systems, strengthened the connections between companies in all links of the supply chain, and made it easier for corporate decision makers to collaborate across enterprises.
• From 2010 to the present, cloud computing-based ERP products have gradually come onto the historical stage by building non-localized systems using technologies such as SaaS and PaaS. The ability to customize to meet the personalized business demands of enterprises is testing the scale of the system's "application market" and secondary development capabilities, among which secondary development capabilities have attracted industry attention with low-code and zero-code.

We can observe that the basic trend of the industry has four characteristics:

1. From solving the enterprise single point management problem to solving the enterprise organization management problem
2. From focusing on internal management and production efficiency to focusing on data and collaborative efficiency across the supply chain
3. From serving things (enterprise production, etc.) to serving people (enterprise decision makers, system users)
4. From focusing on the functions of system modules to focusing on the ability of system modules to be quickly and cost-effectively configured and implemented

Based on this, we have reason to believe that the evolution of traditional management systems to the digital age is currently underway. At present, we need to focus on the informatization of systems based on data processing and intelligent experience methods guided by the system's autonomous decision-making to achieve goals.

We organize our thoughts based on the role and purpose of the system in the production relations during the production process. We call the stage in which the system focuses on production management and human processing the traditional era, and the stage in which the system uses data processing capabilities to replace humans in making repetitive decisions the digital era.

It is not difficult to find that the systems of the traditional era require us to pay more attention to the capabilities provided by the system during production and development, and for large-scale production, it is best to use low-cost and fast development methods. In line with this development model, various technologies and design platforms have been bred. Typical ones include Salesforce open platform, Ant's AntD, etc. Cainiao's platform is called Cone, which means Cainiao One. Cone provides design engineering, standardization, low-code development, capability service platform, mini-programs and other capabilities, focusing on optimizing system production links, cost and efficiency, capability module reuse and experience standardization.

In the era of evolution, in addition to the above production factors, the middle platform must also pay attention to the changes in experience brought about by changes in the purpose of the system. In order to reduce the system's dependence on people and allow the system to make highly repetitive and low-risk decisions instead of humans, the middle platform needs to provide the ability to precipitate solutions and system judgment trigger conditions, as well as the interactive form of supporting intelligent experience.

Pioneer in intelligent transformation

It may sound like intelligence is still a long way off, but in fact, this era has already begun. Let's look at several examples of enterprise intelligent transformation:

In 2016, Google adjusted its corporate strategy from Mobile First to AI First, and as a result, the user experience of Google products gradually became more intelligent and humanized.

Google Photos provides more than 5 billion photos to be viewed by Google Photos every day. AI can help users identify, beautify and share photos more easily. It uses AI to segment images, automatically repair overexposed and underexposed photos, and can also perform color correction on photos. Similarly, Google Assistant uses deep learning Wavenet technology to provide 6 natural human voices that are difficult to distinguish between true and false, and provides services in more than 30 languages ​​in more than 80 countries and regions (no coverage in China). With this technology, people's experience in reading content will become easier and more humane. The most popular plug-ins in the Google Chrome plug-in market are almost all for Gmail smart replies. These plug-ins can distinguish the objects and contents of emails, automatically enter work-related documents into the corresponding modules of the system, and can automatically reply to emails according to preset conditions. This intelligent experience will be a boon for workers who need to handle a large number of customer emails on a daily basis.

IBM divides intelligence into three stages:

• The first phase builds process automation capabilities (RPA) based on cloud computing, the Internet of Things, data analysis and other capabilities to achieve process automation, so that humans only need to make decisions;
• The second stage is intelligent automation (AI+RPA), which combines artificial intelligence and automation technology to provide humans with a better customer experience. For example, AI can help create guidelines for using RPA to automate processes, and AI uses data to quantify and calculate process efficiency, and simplify it to achieve higher efficiency.
• The third stage is true business intelligence (AI). IMB integrates enterprise-level AI into open hybrid cloud solutions to achieve natural language processing, trust, automation, and the ability to run anywhere. I will not elaborate on this here.

In 2019, Philips applied artificial intelligence technology to the medical field. Through image recognition, voice recognition, language processing, data mining, and cognitive reasoning capabilities, it serves the fields of prevention, diagnosis, treatment, rehabilitation, drug research, and hospital management, enabling doctors to change from "intuitive medicine" that relies on experience to "precision medicine" that relies on data. We found that the biggest benefits of intelligent technology are actually product service quality and user experience.

B-side system intelligence map

If we are still using traditional systems today, how can we gradually make them intelligent? It is probably unrealistic to achieve everything in one go. What goals do we need to achieve by building them step by step?

Let's sort out our thoughts. There are three elements of an intelligent system:

1. System awareness - the system needs to know what is happening in the business
2. The system's decision-making ability - the system needs to decide what process needs to be executed (including whether human intervention is required)
3. The execution capability of the system - intelligent systems are good at replacing humans to do a large amount of low-risk and highly repetitive work, and they also have the ability to directly generate production value.

Then we can divide the degree of system intelligence according to three factors:

1. Whether the system can complete tasks automatically (process automation)
2. Whether the system can initiate and complete tasks autonomously (decision-making + process automation)
3. Whether the system can complete self-evolution (perception + analysis + optimization execution capabilities)

Based on this idea, we divide the degree of human participation in system execution and divide the system intelligence into 5 stages:

• Basic manual (S0): All information processing and task handling must be completed manually, and the system only stores and presents information, such as online forms.
• Assisted execution (S1): All tasks must be handled manually, but the system can assist in the process. The system provides data analysis assistance in task distribution, push or processing, such as recommendation and prediction.
• Conditional Intelligence (S2): Manually complete the setting of rules and processes, or complete the establishment of a knowledge base. The system can automatically complete specific task processing based on preset rules, experience or intelligent information decision analysis.
• Highly intelligent (S3): Humans only need to train the system to complete system optimization. Based on state perception and real-time analysis, the system can make autonomous decisions, explore tasks, and complete tasks automatically.
• Fully intelligent (S4): The system has cognitive capabilities, can evolve and improve autonomously, and can autonomously optimize business processes or propose new solutions.

During the construction process, how do we need to define the stage of system intelligence?

We have conducted a lot of theoretical verification and simplified it into the table above. We define the system's intelligence level by who completes the main work in terms of information presentation, task organization and distribution, task processing capabilities, rule and process setting, and autonomous optimization and upgrading.

Cainiao's system products are also gradually evolving towards higher-level intelligent capabilities through the capabilities of the middle platform. So what capabilities do we need to build to achieve the intelligent transformation of the system?

We have set high-priority necessary capabilities (sapphire blue) and sustainable capabilities (blue-gray) at each stage. We believe that these capabilities are the key to system intelligence, and the changes that come with them are disruptive changes in user experience.

Capacity building means cost. Based on Cainiao's 6 years of experience in implementing the experience design platform, not all systems need to pursue S4's full intelligence, but need to decide the level of system intelligence based on scenarios and needs, as needed. The more repetitive the process, the more intelligent it needs, and the higher the decision risk, the more the system needs human experience to make judgments. The system only needs to provide data analysis capabilities.