URUMQI, China, Aug. 8, 2025 /PRNewswire/ — Recently, as one of the first batch of specialized scenario pilot units for the innovative application of artificial intelligence large models by State Grid Corporation of China in 2025, the financial risk penetration supervision scenario in the material field, constructed based on the Guangming large model by State Grid Xinjiang Information & Telecommunication Company, has received recognition from State Grid Corporation of China and was commended at the monthly AI meeting in July.
In 2025, State Grid Corporation of China proposed to focus on overarching and comprehensive management requirements that impact the high-quality development of the power grid and the company. Centering on three key areas—investment performance evaluation, financial penetration supervision, and asset life cycle management—the initiative aims to achieve top-down, full-level, full-chain, full-process, and full-element penetration across domains such as development, finance, and equipment. In the area of financial penetration supervision, efforts are concentrated on enhancing the effectiveness of supervision through the three core directions of risk prevention and control, fund penetration, and lean management. This enables online monitoring and dynamic risk control in key areas, continuously improving the level of lean management.
As a pilot unit for financial penetration supervision, since undertaking the pilot work, State Grid Xinjiang Information & Telecommunication Company has placed great emphasis on its implementation. A dedicated working group was established to strengthen organizational and technical support. Within 60 days, a financial risk penetration supervision scenario in the material field was developed based on the Guangming large model and has since been deployed and applied within the Finance and Assets Department of State Grid Xinjiang Electric Power and the Material Company. The scenario addresses the challenges of scattered cross-system data and low OCR model recognition accuracy in the clearance of material payables. By integrating data from systems such as the digital legal system platform, ECP2.0, ERP, and the smart shared financial platform, the scenario connects the entire chain of information including contracts, performance documents, invoices, and payments. Leveraging semantic understanding, multimodal analysis, and retrieval-augmented generation (RAG) technologies of the Guangming Large Model, five intelligent entities—doubtful point inquiry, supervision distribution, payment review, supporting evidence review, and report generation—have been constructed, enabling full-process intelligent supervision of debt collection. The scenario is accessed over 400 times daily on average, reducing the time required for data analysis of the receivables and payables ledger from 1–2 days to real-time issuance, and increasing debt collection efficiency by 50%.
Moving forward, State Grid Xinjiang Information & Telecommunication Company will continue to deepen the application and iterative development of the Guangming Large Model in the material financial risk supervision scenario, accelerate the promotion and implementation of the scenario across Xinjiang, and leverage artificial intelligence to enhance financial risk prevention and control capabilities.