Internal: Enterprise’s own application of AI

1. Prioritize visual intelligence and gradually move towards data value mining

AI has rich application scenarios in various fields such as industrial R&D, production, supply chain, and operation management, and many of them have been verified and replicated in many industrial enterprises. For industrial enterprises that want to start AI applications, in addition to referring to successful practice scenarios in the same industry, starting from visual intelligence application scenarios, such as machine vision quality inspection, video security monitoring, etc., the application threshold is lower and it is easier to quickly see When results are achieved, priority can be given. On this basis, data intelligence scenarios, such as defect root cause analysis, process parameter optimization, equipment predictive maintenance, and intelligent equipment control, can improve business value even more, but the threshold and difficulty are also higher and can be gradually explored.

2. Properly coordinate the group’s platform construction and the subsidiary’s scenario application construction

Many group-type industrial enterprises are carrying out AI application and construction work at the group and branch levels at the same time. However, judging from IDC's actual research and exchanges, many group companies have a situation where group and subsidiary construction are not well synchronized. Some enterprise groups prefer to rely on their subordinate digital companies to build group-level AI platforms and development capabilities for application by branch companies. However, many branches and subsidiaries have built specific business applications based on AI based on their own needs, and have not been able to form good synergy with the group's AI platform. IDC recommends that when carrying out AI construction, the group and its subsidiaries should first fully communicate with each other about the application scenarios that the subsidiaries have already carried out, the actual needs of the subsidiaries' business for the platform, and other issues, and then better plan the AI applications.

3. Explore large model applications based on subdivided scenarios

The application path of large models in industries such as industry is quite different from the application of large models in consumer scenarios such as entertainment and office. The direct construction of industry-level large-scale models has high requirements on corporate capital investment, data and talent base, and is suitable for group companies with strong technical capabilities. More enterprises can consider selecting some subdivided application scenarios and carry out application exploration of large models around the scenarios.

4. Carefully consider building your own AI full-stack capabilities

Some group companies have many AI application scenarios in their many subsidiaries. If each scenario requires cooperation with external suppliers, it will take a long time and the cost is high. Therefore, they prefer to form teams, build platforms, and build their own full-stack AI capabilities. However, scenario-oriented AI application development requires not only an expert team that takes into account AI algorithms and business mechanisms, but also requires rich AI application development experience, sufficient data sample foundation and other capabilities. Small improvements in individual scenarios require strong algorithmic capabilities, which poses a huge challenge to companies with insufficient digital capabilities. It is recommended that group companies with strong data and algorithm talent foundations can consider building their own. More companies can more economically choose to cooperate with some service providers with relevant experience, starting with applications and gradually expanding their own AI capabilities and teams.

Externally: Facing industry empowerment

1. Strengthen the construction and opening of desensitized data sets

Industrial enterprises have accumulated abundant actual production data, but many lack algorithmic capabilities, and many technology suppliers with strong algorithmic capabilities lack actual data. In the past, limited by the level of enterprise data management and data security control requirements, data could not flow well from enterprises to technology suppliers. Get rid of data sensitivity through desensitization, and in compliance with data security requirements, open the company's data to the outside world with algorithmic capabilities, which will help the outside world help companies solve actual production problems and help companies realize business value improvements faster and better.

2. Cooperate with supplier partners to develop AI solutions to empower the industry

Many group-type industrial enterprises are leaders in the industry and are pioneers and benchmarks in AI applications. On the basis of their own applications, enterprises can cooperate with suppliers to organize application scenarios with industry commonalities into solutions, and through reasonable interest division models, output and empower solutions to other enterprises in the industry to help the entire industry realize AI reduces costs and increases efficiency.

AI applications in industrial enterprises have been developing for many years, and IDC has also tracked and researched many excellent suppliers. Each of them has rich practical experience in different fields of AI+ industry, which can be used as a reference for industrial enterprises that want to carry out AI applications. These suppliers include but are not limited to: Alibaba Cloud, Aqiu Technology, Baidu Smart Cloud, Innovation Qizhi, 4Paradigm, Gechuangdongzhi, Huawei Cloud, Jiyun Technology, Kaos, Kunlun Data, Inspur Industrial Internet, Luculent Intelligence, Merrill Lynch Data, Mecamander, Shanshu Technology, Shuzhilian, Tencent Cloud, Weiyi Intelligent Manufacturing, Xinghuan Technology, Xuelang Cloud, etc. (arranged in alphabetical order).