During the 9th Critical Information Infrastructure Independent Security Innovation Forum, Sointech showcased the deployment depth of domestic Agentic AI in real business scenarios through its manufacturing R&D and enterprise knowledge processing practices.
On April 19, the 9th Critical Information Infrastructure Independent Security Innovation Forum was held in Beijing, where the "2025 Cyberspace Independent Innovation Survey Report" was officially released. As one of the co-authoring units, Sointech deeply participated in the related content building. In "Chapter 8: Deepening AI Applications, Reshaping Business Value," Sointech, around its practical experience of "R&D and Production: From Single-Point Tools to Systematic Collaboration," provided a reference sample from a front-line business environment for AI applications in manufacturing R&D scenarios. Sointech's case content was also included in the report's appendix.
More notably, Sointech had 3 cases selected into the AI application content of the "2025 Cyberspace Independent Innovation Survey Report": the Sointech Enterprise Document Intelligence System (ADS) reshaping the paradigm of high-complexity knowledge processing with agentic workflows; the SOIN ALA intelligent platform empowering the reshaping of advanced manufacturing R&D processes with AI agents; and the Sointech PLM intelligent assistant reshaping the paradigm of intelligent component selection in manufacturing with AI. In the case compilation in the report's appendix, Sointech's three cases are continuously included as Cases 32–34 of Chapter 9, covering key directions such as enterprise document intelligence, ALM/PLM R&D process intelligence, and intelligent component selection in manufacturing.
For Sointech, this co-authorship and selection is not just a brand-level presentation, but a phased embodiment of the company's long-term focus on enterprise-grade Agentic AI productization, engineering and industry deployment capability. All three cases in the report are built on Sointech's proprietary Agentic Workflow Engine (AWE) as the core foundation, embedding AI capabilities into real business processes through the approach of "engine + professional agents + industry workflows," rather than staying at the level of generic Q&A or single-point tools. Whether the requirements, testing and traceability closed loop in ALM scenarios, component selection and engineering knowledge invocation in PLM scenarios, or parsing, comparison, proofreading and translation in complex document scenarios, Sointech is advancing AI deployment in ways closer to the enterprise production environment.
In the advanced manufacturing R&D process scenario, the report shows that the SOIN ALA intelligent platform has built an enhanced closed loop of "extraction — review — generation — analysis — traceability — optimization" around key stages such as requirements extraction, quality review, test generation, change impact analysis and traceability matrix maintenance. Among them, atomic-level requirement unit automatic extraction accuracy reaches 94.7%, requirements review efficiency improves 5–10x, test scenario coverage rises above 90%, and requirements change impact analysis time is shortened from 4 hours manually to 5 minutes. This means that in high-compliance, high-complexity R&D systems, AI is no longer just an auxiliary tool, but is beginning to become an engineering capability that genuinely participates in process execution and quality improvement.
In the intelligent component selection scenario for manufacturing, the Sointech PLM intelligent assistant combines multimodal engineering document parsing, knowledge structuring and condition-driven algorithms. According to the report, the system achieves document text parsing accuracy of 94.7%, chart data parsing accuracy above 93%, and table structure restoration rate of 95%; after inputting condition parameters such as rotational speed, load and temperature, it can complete knowledge graph comparison and return recommended models within 3 seconds, with customer-side engineers' adoption rate of agent recommendations reaching 88%. This shows that AI is no longer just an information retrieval tool, but is becoming an "embedded design consultant" with engineering computing and decision-support capabilities.
In the enterprise document intelligence scenario, Sointech's ADS enterprise document intelligence system further proves the production-grade capability of agentic workflows in high-complexity knowledge processing. The report shows that in complex financial and technical document processing, ADS achieved document text parsing accuracy of 97.4%, table structure restoration rate of 95%, and chart data parsing accuracy above 93%; cross-version difference comparison time for an 800-page report was compressed from 2–3 hours to 5–10 minutes, key terminology translation consistency improved to 99%, and the overall report preparation cycle shortened from 8–12 days to 2–4 days. For enterprise scenarios requiring high accuracy, high consistency and high audit requirements, this is not only an efficiency improvement, but a simultaneous improvement in process quality and risk control capability.
From the common characteristics shown in the report's cases, what Sointech adheres to is not simply "wiring AI into the business," but continuously polishing product capabilities that are deployable, governable, collaborative and auditable around real enterprise processes. On one hand, Sointech achieves multi-agent scheduling, workflow orchestration and multimodal data processing through its proprietary AWE engine; on the other hand, it has formed implementation paths of progressive enhancement, compatibility with existing systems, support for private deployment and open integration in specific industries. These capabilities enable Sointech to deliver more stable, more credible, more business-penetrating AI solutions in professional scenarios such as manufacturing R&D and complex document processing.
Participating as a co-authoring unit in the "2025 Cyberspace Independent Innovation Survey Report" and being included with multiple practice cases is, for Sointech, a validation from the dual dimensions of industry research and real scenarios. In the future, Sointech will continue to be driven by real business scenarios, with independent and controllable Agentic AI technology as the foundation, focusing on manufacturing R&D, enterprise knowledge processing and industry intelligent upgrading, continuously advancing AI from "able to see and speak" to "able to execute, collaborate and deliver," truly deploying AI capabilities with professional depth into front-line enterprise processes.