大唐电商：三位一体智能评标，护航合规招标采购全流程

Key takeaway 1: Datang E-commerce has overhauled procurement bid evaluation with a trinity of AI, institutional rules, and organizational support. For the general reader, the essential insight is how AI-assisted evaluation boosts efficiency, cuts costs, and mitigates risk.

1. Three AI tools: Bid document parsing and qualification review use OCR and NLP for second-level comparison, raising evaluation efficiency by 70%. An intelligent price-comparison system benchmarks quoted prices against historical and market references, reducing overall costs by 15%. AI anomaly detection monitors audio, video, and logs in real time, delivering 98% accuracy in risk alerts.

2. Institutional and organizational safeguards: The company revised management rules to position AI as an aid to human judgment, with experts retaining final decision-making authority. A cross-departmental team continuously trains and maintains the models, addressing concerns about machines calling the shots.

Key takeaway 2: Intelligent bid evaluation has already generated measurable procurement improvements.

1. Procurement lead times have fallen sharply: in 2025, the parent company's procurement cycle for tenders was 31 days and 7.9 days for non-tender procurement, down 36 days and 17 days year-on-year, respectively.

2. The platform ecosystem is substantial: annual transaction value exceeds RMB 100 billion, with over 170,000 registered suppliers, more than 1,000 procurement entities, and nearly 30,000 evaluation experts—a replicable model for large-enterprise procurement.

Key takeaway 1: Xiwei.com, operated by Datang E-commerce, is a representative example of brand building for a centrally administered enterprise (central SOE) procurement platform.

1. Brand history: The platform's predecessor went live in 2005 and, after nearly 20 years of upgrades, has evolved into a unified portal covering four modules—tender, non-tender, e-supermarket, and auction. It has also been recognized as a leading enterprise in digital and intelligent innovation among central SOE e-commerce platforms.

2. Channel value: With more than 170,000 registered suppliers, over 1,000 procurement entities, and over 50 external energy companies such as Shenergy and Shaanxi Coal Group joining, it is clearly a high-value B2B channel for entering energy-industry procurement.

Key takeaway 2: Intelligent bid evaluation is changing the logic of procurement pricing and product R&D.

1. Pricing impact: The intelligent price-comparison system analyzes quotation reasonableness against historical data and market reference prices, making it harder for suppliers to charge premiums through information asymmetry. Brand pricing must return to true costs and market conditions.

2. Product R&D direction: The platform's R&D focus is shifting toward AI-assisted evaluation, Xinchuang (domestic IT innovation) adaptation, and private deployment of DeepSeek. It uses 100% accuracy on objective scoring and 80% pass rate on subjective items as competitiveness indicators; buyers increasingly value compliance, efficiency, and data value.

Key takeaway 1: For sellers looking to enter the energy procurement market, Datang E-commerce is an incremental channel with annual transaction value above RMB 100 billion.

1. Market opportunity: In 2025, the platform's transaction value approached RMB 150 billion, with over 1,000 procurement entities. The e-supermarket carries more than 560,000 listed items; long-term agreement (LTA) transactions totaled RMB 5.177 billion, up 24.75% year-on-year, with on-time delivery above 99%. Sellers should pay particular attention to opportunities in LTAs and supermarket supply.

2. Auction channel: The platform used auctions to revitalize idle-asset transactions worth RMB 391 million. Companies seeking to dispose of or procure idle assets should watch this model.

Key takeaway 2: Intelligent bid evaluation has changed the rules for suppliers' bidding and pricing.

1. Stricter qualification review: The system automatically compares bid documents against tender requirements using OCR and NLP. Non-standard formatting and blurred qualification materials are prone to rejection, so bid documents must be standardized.

2. Reasonable pricing required: The intelligent price-comparison system assists evaluation by combining historical prices and market references, so pricing must be based on costs and market conditions. AI anomaly detection automatically warns of bid-rigging, and compliance is the bottom line.

Key takeaway 3: Policy support and entry paths.

1. The NDRC, together with seven other departments, has issued a document promoting AI applications in tendering and bidding. The platform is expanding smart management, preparation, review, Q&A, and evaluation scenarios; eligible suppliers should watch for policy dividends.

2. The platform welcomes external energy companies to join. Interested parties should monitor recruitment conditions and collaboration models, and leverage the shared service system to reach more procurement entities.

Key takeaway 1: Demand shifts at the energy central SOE procurement platform send clear order signals to factories.

1. Industry coverage and procurement scale: The platform covers power, coal, chemical, and other industries. The e-supermarket carries over 560,000 listed items; LTA transactions reached RMB 5.177 billion, with an LTA coverage rate of 58.02%. This indicates that central SOEs increasingly favor stable, long-term supply, so factories should pursue inclusion in the e-supermarket and LTA rosters.

2. Product requirements: On-time delivery has remained above 99%, reflecting high demands on supply-chain responsiveness and delivery capability. Factories need to strengthen production planning and inventory management.

Key takeaway 2: Digital capability is the entry ticket for factories to connect with intelligent procurement.

1. Intelligent parsing of bid documents requires standardized bid essentials and formatting, reducing rejection rates caused by format errors. Factories must improve the digital quality of product documentation and bid files.

2. Datang E-commerce offers replicable supply-chain digitalization solutions built on layered microservices and an AI capability center. Factories can apply the same architectural thinking to their own digital transformation and better integrate with central SOE procurement systems.

3. With the platform deploying DeepSeek large models, future parsing of bidding documents will become even more intelligent. Factories should prepare structured product data in advance.

Key takeaway 1: Service providers can regard Datang E-commerce's intelligent bid-evaluation system as a mature example of AI applied in the tendering and bidding domain.

1. Industry trend: Procurement is moving from process automation to AI-driven decision-making. In the Intelligence 3.0 era, AI-assisted evaluation for non-tender procurement has achieved 100% accuracy on objective scoring and 80% pass rates on subjective items, and national policies are promoting AI adoption.

2. Client pain points: Bid evaluation suffers from low efficiency, high costs, elevated integrity risks, information asymmetry, and prominent human interference. There are also bottleneck risks in critical core technologies.

Key takeaway 2: The replicable solution spans technology, institutions, and organization.

1. Technology layer: Three core tools—bid-document parsing (OCR/NLP), intelligent price comparison, and AI anomaly detection—combined with a cloud-native hybrid cloud architecture and an AI capability center, plus private deployment of the DeepSeek R1 70B large model.

2. Institutional and organizational layer: The company revised its intelligent bid-evaluation management rules to clarify AI's auxiliary role and experts' decision-making authority, and formed a cross-departmental operations team to maintain the models and keep AI aligned with business.

3. Replicability: The solution has won first prize in a national competition and is recognized as an industry best practice, providing a reference template for service providers working with large energy enterprises.

Key takeaway 1: Platform operators can learn from Datang E-commerce's two-decade platform architecture evolution.

1. Architecture model: It uses shared services and a unified portal to integrate four transaction modules—tender, non-tender, e-supermarket, and auction—into one platform. Integration with planning, legal, ERP, online banking, and logistics systems enables end-to-end integrated electronic transactions.

2. Innovative functions: Mobile QR-code signing and sealing, remote decentralized bid evaluation, one-click bid clearing, e-contracts, online logistics, internet finance, and online payment all enhance transaction transparency and convenience.

Key takeaway 2: Operational and merchant-acquisition essentials for platform operators.

1. Operations management: Build a cross-departmental intelligent bid-evaluation operations team responsible for model training, rule maintenance, and effectiveness evaluation; revise the intelligent bid-evaluation management measures to balance AI efficiency with compliance responsibilities.

2. Merchant acquisition strategy: The platform has over 170,000 registered suppliers and more than 1,000 procurement entities. External energy enterprises are an important growth source; its model of bringing on more than 50 external enterprises is worth referencing.

3. Risk control and trust: Meeting the national Class 3 Cybersecurity Level Protection, the three-star certification for e-tendering systems, and Xinchuang (domestic IT innovation) adaptation, while using AI anomaly detection to prevent bid-rigging, are critical to winning customers.

Key takeaway 1: The Datang E-commerce intelligent bid-evaluation case reflects a new industry trend of deep integration between AI large models and tendering and bidding institutions.

1. The intelligent bid-evaluation system adopts a trinity model of AI empowerment, institutional safeguards, and organizational innovation, in which AI supports analysis and experts retain final decision-making authority, forming a new mechanism of human-machine collaboration.

2. The technical architecture uses layered microservices and a cloud-native hybrid cloud, deploys DeepSeek R1 70B, and builds an AI capability center. AI-assisted evaluation for non-tender procurement achieves 100% accuracy on objective scoring.

Key takeaway 2: The case provides a practical reference for improving policies and regulations.

1. Intelligent bid evaluation raises the legal boundary issue of machines calling the shots. Datang E-commerce addressed this by revising management rules to clarify AI's auxiliary role and responsibility boundaries, a useful reference for policymakers.

2. The award-winning project closely aligns with the NDRC's Document No. 195 (2026), covering five smart scenarios—management, document preparation, review, Q&A, and evaluation—creating a closed loop for smart procurement.

Key takeaway 3: The business model and performance results have research value.

1. Shared services and a unified portal integrate four transaction modules. Through data sharing and an AI capability center, high-value data assets are accumulated, forming a sustainable platform model.

2. Performance data: 70% efficiency improvement, 15% overall cost reduction, and 98% risk-alert accuracy. In 2025, the parent company set six historical records in operating performance, demonstrating the business value of digital-intelligent procurement.