36Kr Exclusive | The core team is from Microsoft, and it has received nearly 100 million in investment. It aims to bridge the last mile of AI implementation in factories.
Author: Ou Xue
Editor: Yuan Silai
Yingke recently learned that Zhiyong Kaiwu, a company focusing on industrial agents and high - value applications, has completed nearly 100 million yuan in Series Angel + financing. This round was led by existing shareholder Ruifeng Capital, followed by Chuangxiang Investment. Meanwhile, it received strategic investments from the family office and executive team of strategic customer Luxshare Precision. The funds will be mainly used to build the world's first "Industrial Semantic Engine" for the industry and to develop intelligent agents for high - value positions in the manufacturing industry.
This is already the third round of financing for Zhiyong Kaiwu in the past year. In 2025, the company successively completed investments from Ruifeng Capital, Yuzi Zhangquan, and Saiyi Information, a leading enterprise in manufacturing digitalization among listed companies.
Zhiyong Kaiwu was established in Guangzhou in January 2024. It focuses on transforming complex industrial knowledge into AI - interpretable and executable logic through the "Industrial Semantic Engine + Industrial - grade Multi - agent System", reducing the threshold for AI implementation in industrial scenarios.
Yingke learned that the core team of Zhiyong Kaiwu has a strong "Microsoft gene". CEO Guan Zhen once served as the chief technical advisor of Microsoft China and is a well - known artificial intelligence application expert in China. COO Zhao Ming once served as the chief architect of the Microsoft Global CTO Office. CTO Zhang Shanyou is a 20 - time Microsoft Most Valuable Professional and a member of Tencent's Architect Hall of Fame. Other core members of the team also come from companies such as Microsoft, Alibaba, Tencent, and IBM.
Different from most software companies in the market that only stay at the digital level, Zhiyong Kaiwu has truly penetrated into the production line. The company's products can natively support industrial protocols such as OPC UA (based on the C# language) and achieve millisecond - level response. Its multi - agent system can already coordinate the robotic arms and machines on the production line to complete the autonomous planning, scheduling, and intelligent process optimization cooperation of complex tasks.
This round of funds will be mainly used to deepen the technological iteration of the industrial - grade multi - agent operating system Agent Foundry, build a foundation supported by the "Industrial Semantic Engine (Ontology & Harness)", and drive the large - scale application of "Position Intelligent Agents (NCREW)" in multiple high - value industrial scenarios.
Currently, Zhiyong Kaiwu has formed three major product matrices: First, multi - position intelligent agents: Standard AI employee products for enterprise end - users. Enterprises do not need to invest a large amount of resources to start research from the AI foundation. They only need to upload materials such as position SOPs, and the intelligent agents can complete the adaptation to multiple tasks within "N minutes" through self - learning. The Position Intelligent Agents (NCREW) are not simple chatbots but "virtual employees" that evolve based on enterprise SOPs and industrial logic. In specific scenarios, their comprehensive collaborative efficiency can be increased several times, and they can execute precisely without interruption 7x24 hours. The efficiency of a single position intelligent agent is equivalent to that of 6 - 10 human employees.
Second, high - value scenario products that can be empowered by artificial intelligence in the industrial manufacturing field. Based on the mining of industrial privatized data and scenarios of leading enterprises in the industry chain, it serves customers widely in fields such as BOM management, APS scheduling, supply chain prediction and control in the ICT, equipment manufacturing, and automotive industries.
Third, the golden foundation of the multi - agent operating system "Pilot navi" + Industrial Semantic Engine: The semantic engine is the "central brain" that carries the core business logic, decision - making mechanism, and domain mechanism of the enterprise. It solves the problem of inconsistent data semantics between different systems, structures implicit industrial knowledge, and ensures that the output of AI complies with manufacturing specifications and process constraints. It is the key to realizing the rapid migration of intelligent agents across enterprises. The multi - agent operating system is responsible for the understanding and decomposition of complex tasks, multi - agent collaborative planning, dynamic scheduling, and self - reflection and iteration, supporting the full - life - cycle evolution of intelligent agents.
Navi Labor Intelligent Agent Market (Source/Enterprise)
Guan Zhen told Yingke that the value of Zhiyong Kaiwu has been fully verified by leading customers. Taking Luxshare Precision, the strategic investor in this round, as an example, the intelligent agents of Zhiyong Kaiwu have been deeply applied in its multiple production lines.
The achievements include: The efficiency of a single AI scheduler is equivalent to that of six human employees; the SOP automation rate reaches 80%; the on - the - job training time for tidal workers is shortened from 1.5 days to 2 hours; the processing speed of production line abnormalities is increased by 8 times; and the equipment maintenance efficiency is increased by 40%. Next, the company plans to launch multiple general - purpose position intelligent agents this year, covering core scenarios such as R & D, scheduling, supply chain risk, and quality in industrial enterprises. In addition, the company will continue to deepen cooperation with multiple leading enterprises in the industrial chain to jointly promote the formulation of industrial semantic engine standards for industries such as equipment manufacturing, automotive parts, and precision electronic processing.
The following is an excerpt from the interview between Yingke and Guan Zhen, CEO of Zhiyong Kaiwu:
Yingke: Compared with Internet giants and traditional software companies, what are the core advantages of Zhiyong Kaiwu?
Guan Zhen: The best business model of giants is to sell computing power and models. It is difficult for them to customize production line logic for a single factory, and the marginal cost is too high. Traditional software companies have historical burdens. It requires huge investment to develop a multi - agent system independently, and it is easy to fail. Our advantages lie in three points. First, we are truly on the production line. Our products natively support industrial protocols without translation and can achieve millisecond - level response, which is crucial in the industrial field. Second, the industrial semantic engine. We model 80% of the general industrial logic, and the remaining 20% of personalized knowledge is self - learned by AI, enabling rapid migration across enterprises. At the same time, we use the Harness framework to impose physical constraints on industrial processes, which means that AI will not have "hallucinations", and every action of it complies with safety and process specifications. This is the confidence for us to claim that "an agent can start working in half an hour". Third, multi - agent collaboration. We not only handle single - point tasks but also enable the robotic arms, machines, and quality inspection systems on the production line to collaborate autonomously like a team. The test standard is simple - whether you are truly on the production line and whether you can replace 6 or 10 people.
Yingke: Industrial scenarios are highly customized. How do you achieve standardization and large - scale implementation?
Guan Zhen: This is exactly the direction we are focusing on this year. We have launched a standard product called "Position Intelligent Agents". Enterprises do not need to start from selecting models and buying servers. You only need to tell AI what this position does and upload materials such as SOPs and historical data. It will complete the adaptation to the job within half an hour to a few hours through self - learning. We break down this ability into the "Industrial Semantic Engine" - it pre - abstracts 80% of the general logic, and the remaining 20% of the personalized part is automatically aligned by the parser. Just like a human employee who can start working after getting familiar with the system. We are not competing on model parameters but on the depth of industrial logic analysis. Giants are creating "generalists", while we are creating "closed - loop specialists".
Yingke: What is the commercialization path of Zhiyong Kaiwu?
Guan Zhen: We have a three - step plan. First, serve large leading enterprises in the industrial chain, providing platforms and high - value scenario creation. Currently, we have served leading customers such as Luxshare Precision. Second, cooperate deeply with ISVs (Independent Software Vendors) to help them upgrade their traditional software into AI - native applications. Third, launch standardized position intelligent agents for small and medium - sized enterprises and quickly cover the market through distribution channels. In 2026, Zhiyong Kaiwu will achieve large - scale business growth through a three - in - one business strategy of "led by leading enterprises in the industrial chain, empowered by ISVs, and channel penetration". Currently, the company's contract value has shown a ten - fold growth trend, and it is expected to cover more than a thousand manufacturing enterprises in the next two years.
Views of the investor:
Ling Xipu, Chairman of Ruifeng Capital: The core reason for Ruifeng to increase its investment this time is our implementation of the principle of "no empowerment, no investment". Different from financial investors, Ruifeng relies on its unique industrial ecosystem and resources of leading enterprises in the industrial chain to continuously provide Zhiyong Kaiwu with real factory scenarios and front - line demands. Ruifeng always screens, evaluates, and empowers post - investment projects through first - hand industrial information. This additional investment as an existing shareholder is a high recognition of Zhiyong Kaiwu's ability to implement in leading enterprises in the industrial chain and a strategic continuation of Ruifeng's construction of the "industrial ecosystem". The aim is to promote the invested enterprises from technology verification to large - scale commercial implementation through the dual support of capital and resources.