From October 14th to 16th, the 2026 Artificial Intelligence Industry Conference will be held for the first time in Jinan, Shandong. This conference will invite Nobel laureates, academicians of the Chinese Academy of Sciences and Chinese Academy of Engineering, and other top experts in the industry to share on stage, focusing on building a comprehensive industry platform for high-end dialogue, achievement debut, and scenario supply and demand docking in the field of artificial intelligence in China.

Shandong has a complete range of industrial categories and rich application scenarios, which are promoting the all-round and deep empowerment of the manufacturing industry by artificial intelligence, and exploring the leap from "Shandong Manufacturing" to "Shandong Intelligent Manufacturing". As a result, a number of representative industrial vertical domain models have emerged in Shandong.

Recently, Kaos IoT Technology Co., Ltd. (hereinafter referred to as "Kaos") launched the industrial world model and full stack intelligent agent matrix globally, aiming to bridge the "last mile" of industrial AI scale implementation and initiate a paradigm revolution from "AI empowering industry" to "industrial AI native".

New Engine of Industrial Internet

When consumer models frequently make a name for themselves with smooth dialogue and creative generation, the public's attention is focused on the novel experience of C-end AI; Another more hardcore AI revolution is quietly happening in the roaring factory workshops and production lines. In this race to reshape the core competitiveness of China's manufacturing industry, Kaos has put forward its own development ideas - exploring the creation of "industrial intelligent partners" that are rooted in the workshop, follow causal logic, and execute reliably.

There is a fundamental difference between the industrial model and the consumer model. Consumer AI allows occasional mistakes, while industrial scenarios cannot tolerate 'AI illusions'. A wrong process parameter suggestion may result in a large number of defective products, equipment damage, and direct losses of hundreds of thousands or even tens of millions of yuan. What industrial AI pursues is not fancy words, but understanding device signals, understanding process mechanisms, and outputting safe and practical execution solutions.

This is precisely the important reason why Chaos constructed the "Industrial World Model" and "Intelligent Agent Matrix". Yu Jinze, Director of Data Intelligence Technology at Kaos, pointed out that the ability of general large models comes from language and text training, but they have never really entered the workshop. The model can memorize device parameters, but cannot understand thermodynamic constraints; It can generate smooth text, but cannot guarantee physical credibility. In precision industrial production, this "illusion" may lead to real quality hazards and safety risks.

If we compare the universal big model to a "liberal arts student" who is well read and skilled in written expression, the Kaos Tianzhi Industrial Big Model is more like a "chief engineer" who has been deeply involved in the manufacturing industry for decades, understands equipment, processes, and on-site operations. It is exploring the transformation of hidden process experience hidden in the minds of experienced professionals into replicable and iterative digital industrial assets, promoting the domestic manufacturing industry to move from traditional digital transformation to a new stage of "AI native organization".

Unlike other industrial models developed by enterprises, the Tianzhi Industrial Model has been fully applied and repeatedly verified in the connected factories and intelligent production lines of Kaos and its parent company Haier since its birth in September 2023.

At present, Kaos solutions have covered vertical industries such as home appliances, electronic products, automobiles, energy and chemical industries, and mechanical equipment. According to statistics, Tianzhi industrial model has integrated massive data, precipitated more than 4700 industrial mechanism models, more than 200 expert algorithms and 110+agent development tools, becoming the first vertical domain model based on the industrial Internet platform in China, and has been applied in 9 major industries and more than 40 scenarios.

Kaos' location in Shandong also provides it with rich application scenarios. As a major manufacturing province, Shandong has all 41 industrial categories and is one of the provinces with the most complete industrial system in the country. Among them, Shandong accounts for 197 out of 197 industrial categories and 603 out of 603 industrial subcategories. The output of over 100 key products ranks among the top three in the country, and the revenue of 18 major manufacturing categories ranks among the top five in the country. This enables Kaos to empower thousands of industries, obtain nutrients from them, and grow rapidly, with a strong industrial soil.

Wang Chao, technical director of Kaos Industrial Intelligence Research Institute, told the Economic Observer that the big model will be the core engine for the new round of development of industrial Internet. In the field of manufacturing, future big models can participate in various stages of production by scheduling all systems and data within the factory, achieving functions such as process optimization, transportation optimization, equipment operation and maintenance, and energy saving and consumption reduction in production scenarios.

Building an 'Industrial Brain'+'Digital Human'

Different from the scattered single point AI tools, Kaos, relying on the industrial Internet platform base and the "industrial brain" with Tianzhi industrial model as the core, began to focus on the research and development of industrial agents - shaping the three virtual roles of CEO "digital man", factory director "digital man" and energy station director "digital man".

These three "digital people" correspond to the three major business sectors of enterprise management decision-making, factory production scheduling, and refined carbon management, forming a complete closed loop of "strategic issuance - production dismantling - energy landing - data reflux - business optimization", attempting to solve the long-standing pain points of "having data but no decision-making", "having systems but no staff", "collaboration relying on communication", and "execution relying on human monitoring" in a large number of manufacturing enterprises.

For example, in the past, when companies were investigating business abnormalities, they needed to pull reports across multiple systems, which took two to three days to hold meetings and analyze; With the help of CEO digital human linkage indicator definition, ontology governance, and business decision-making intelligent agents, the speed of abnormal root cause localization has been improved by 65%, and the overall decision-making chain efficiency has been improved by 70%.

For example, in the factory level scenario, Kaos uses the digital director as the core to promote the upgrade of manual scheduling and passive scheduling to intelligent scheduling and predictive collaboration, creating an intelligent decision-making center for the factory.

Wang Chao explained that the factory often encounters situations where urgent orders are inserted, which disrupts the original production schedule. Planners need to communicate and coordinate repeatedly, and even a slight mistake can result in delayed delivery of orders. Kaos, based on the Tianzhi Industrial Big Model, has developed a flexible production exchange intelligent agent that can unify the scheduling of various executing intelligent agents, automatically complete system operations, task distribution, and personnel dispatch. Once there is a change in production scheduling, the system will automatically synchronize the entire chain of production lines, materials, and processes, and adjust each link in a coordinated manner. Ultimately achieving a 30% increase in production efficiency, effectively resolving planning conflicts caused by urgent order insertion, and ensuring timely delivery of orders.

Based on the multiple interconnected factories built by Kaos and Haier Group, Kaos has developed the COSMO iMOM twin manufacturing integrated platform, creating over 40 industrial intelligent agents covering high-value scenarios such as process optimization, equipment management, quality inspection, energy conservation, and carbon reduction. Each intelligent agent is an "AI expert" in a specific business field, capable of autonomously executing instructions and providing real-time feedback on data, thus forming an industrial closed loop.

Chen Lucheng, the chairman of Caos, pointed out that from "industrial Internet" to "AI+industrial Internet" and then to "industrial AI native", Caos's progress in these three stages is essentially a process of AI's role in the industrial scene deepening - from connecting tools to analysis tools, to "intelligent partners" who can make decisions and implement independently.

(Author Zhong Ang)