The transition from traditional chatbots to agentic AI is no longer a distant prospect; experts predict it could fundamentally transform industrial workflows. At the IDC CIO Summit in Shenzhen, Liu Xiangyang, chief information security officer at Midea Group, stated that existing upper-layer software, which includes both consumer-facing applications and internal management systems, is on the verge of being replaced by these intelligent agents.
Du Yanze, senior research manager at IDC, elaborated on the potential of agentic AI, forecasting that 90% of an AI agent's value will come from its specialized industrial expertise. He pointed out that these agents could significantly reduce supply-chain order processing times from hours to just minutes, showcasing their ability to improve operational efficiency.
Despite the enthusiasm surrounding agentic AI, concerns about its use in critical sectors remain. Experts at the summit warned that industries such as healthcare and aerospace may face substantial risks, where deploying autonomous agents could result in serious safety issues. The discussion highlighted the need for strict verification processes and the establishment of trust before widespread implementation in these high-stakes environments.
China's ambitions in this area are reflected in its 'AI Plus' strategy, which aims for adoption rates exceeding 70% by 2027 and more than 90% by 2030. However, significant gaps in the country’s industrial software capabilities could impede these targets. The disparity between policymakers' goals and current software capabilities in advanced manufacturing raises questions about the feasibility of achieving such rapid integration of AI technologies.
The Technical Landscape of Agentic AI
The shift toward agentic AI signifies a critical move from generic language processing to the need for domain-specific knowledge. This change underscores the importance of high-quality, curated datasets tailored for specific industries. Achieving effectiveness in agentic AI also requires explainability, particularly regarding operational metrics, and effective integration with existing control systems like Manufacturing Execution Systems (MES) and Supervisory Control and Data Acquisition (SCADA) systems.
As the complexity of deploying these systems increases, ensuring safety becomes more challenging. The lack of training data, especially for rare failure scenarios, complicates the verification process and poses risks to operational reliability.
Observations and Future Considerations
The current situation reveals a tension between China's ambitious adoption targets and the reality of existing software limitations. Practitioners may need to engage in extended pilot programs, implement rigorous change-control measures, and conduct thorough validations before deploying agentic AI solutions in safety-critical workflows.
Looking ahead, stakeholders should keep a close watch on regulatory guidance concerning the deployment of AI agents in regulated sectors. The availability of domain-specific benchmark datasets will also be crucial in shaping the future of agentic AI. Insights from pilot results by major manufacturers, disclosures from vendors about verification methods, and successful case studies that link agents to operational control systems will provide valuable information on the practical applications of this technology.
As the industry navigates these developments, the focus will remain on balancing innovation with the need for trust, particularly in sectors where the stakes are highest.
The stories that move AI & crypto markets — before the market reacts.
Free. 7am ET. Five stories. 62,400 readers.
