How can industrial software create value as general-purpose models become widespread?

  • CLS analysis holds that the rapid commoditization of general-purpose large-model capabilities is shifting enterprise software value from general functions and user interfaces toward vertical business execution, proprietary data and industry knowledge. For industrial software vendors, the competitive focus is accordingly shifting to their ability to complete specific production tasks.
  • Manufacturing agents are already being applied in research and design, process improvement, quality inspection, equipment operation and maintenance, and energy consumption management. The material identifies collaboration among multiple agents to complete an end-to-end business closed loop as a key path: value gains come from linking multiple stages into an operational process, rather than simply adding an AI function.

What must be connected to move from software interfaces to production lines?

  • A report by the China Academy of Information and Communications Technology says industrial intelligence needs further integration of perception, modeling, decision-making and execution. This is also a condition for moving from localized applications to an overall transformation of manufacturing: after a model makes a judgment, it must connect to on-site equipment and execute the decision to form an “equipment + data” closed loop.
  • Policy support is extending across this chain. The Implementation Opinions on the Special Action for “Artificial Intelligence + Manufacturing” support on-device inference chips, edge computing servers and a “cloud–edge–device” model system. An industrial internet platform action plan calls for the number of connected industrial devices to exceed 120 million units or sets by 2028, providing policy direction for connecting on-site data.
  • IDC expects user spending on embodied AI robots in China to exceed 11 billion dollars in 2026, with growth approaching 120%. The forecast reflects the direction of demand expansion for AI in the physical world. Its statistical scope is embodied AI robots, which differs from that of industrial software projects.

Which companies have made progress, and what do the delivery figures show?

  • By the end of the third quarter of 2026, SIE Information had cumulatively delivered 255 projects incorporating AI applications, representing more than 20% of all projects delivered. AI is embedded in customers’ business processes in forms including agents and model algorithms. On this basis, the material concludes that demand for intelligent upgrades to existing systems continues to emerge.
  • SIE Information’s connection point is its existing manufacturing services capability, spanning consulting and planning, software development and computing infrastructure operations, alongside the development of a “computing foundation–industrial large model–scenario-specific agent” system. The above figure of more than 20% is the share of projects incorporating AI applications by number and measures delivery coverage; revenue and profit contributions are measured on a different basis.
  • Heide Control, meanwhile, plans to advance the integration of digital foundations and AI applications in industries including shipbuilding and municipal construction, and to promote lifecycle services for industrial network products. Compared with SIE Information’s disclosed project deliveries, the material’s description of Heide Control focuses primarily on the direction of its business initiatives.