The Public Resources Trading Supervision Committee Office of Zhejiang Province has recently issued the "Work Plan for the Pilot Promotion and Application of AI-Assisted Bid Evaluation in the Provincial Water Conservancy Bidding Field (Trial)" along with its supporting rules. This marks the official full-scale promotion and application of the water conservancy project AI-assisted bid evaluation system, which was jointly developed by the Provincial Water Resources Department and units including the Qiantang River Laboratory in Binjiang District, Hangzhou.

The system adopts a lightweight technical architecture combining "general large models with industry review rules". Built on the Tongyi Qianwen large model foundation, it integrates high-quality corpus compiled by water conservancy expert teams and industry review rules, while incorporating natural language processing, machine learning, and optical character recognition technologies. This enables a full-process review covering intelligent verification, analysis, bid clearing, evaluation, and report generation. The system includes dedicated functions addressing the three core aspects of bid evaluation. For compliance review, it refines bidding requirements into 26 categories with 122 precise detection points, standardizing review procedures and achieving automated compliance screening with 99% accuracy in identifying qualified bidders. For technical bid review, the system establishes over 2,000 review rules covering 15 major building types including dams, embankments, seawalls, and river management projects. It performs intelligent analysis and scoring of key content such as construction organization designs and critical technical proposals, ensuring consistent professional standards. For credit bid review, through data sharing with the provincial water conservancy project management system (Transparent Engineering), it enables intelligent comparison and verification of performance records and credit information, achieving 100% accuracy.

In April of this year, the system completed its first real-world application during the Chun'an mountain flood channel project, followed by deployments in Fuyang, Yuhang, Shangyu, and other locations. The integration of AI makes bid evaluation more standardized, efficient, and transparent, injecting new momentum into regulating bidding practices. After training on 150,000 high-standard professional corpus entries, the system functions as a "senior evaluation expert". By providing review point prompts, it compresses the discretionary space of evaluation experts, further standardizing their conduct. Actual measurements show the average deviation rate between evaluation experts and the intelligent agent is only 5.1%, representing a reduction of over 75% compared to pre-AI assistance levels, resulting in more scientific and rational expert scoring. The improvement in evaluation efficiency is also immediately apparent. Taking 100 bidding entities as an example, traditional expert evaluation requires 10 hours, while the intelligent system can generate compliance review results in just 10 minutes. Reading time for a single technical bid report is compressed to 2 minutes. This not only frees experts from routine compliance verification, allowing them to focus more on technical proposal analysis, but also provides references before evaluation, effectively enhancing overall review quality and efficiency.

Beyond improvements on the evaluation side, the system also drives information governance on the bidding side. Market participants classify and file information such as qualifications and performance records in the main database, completing structured data entry. Bid document creation directly draws on main database data, enabling online traceability and full-process recording. Combined with post-bid public disclosure and correspondence verification measures, this reduces the space for false performance claims at the source, effectively maintaining fair market competition. The issued work plan and supporting regulations clarify operational rules for data interaction, compliance review, technical bid evaluation, and expert behavior supervision and verification, providing a clear implementation path and regulatory basis for the promotion of AI-assisted bid evaluation. According to the promotion schedule outlined in the plan, full coverage of AI-assisted bid evaluation for all water conservancy project bidding will be gradually achieved by the end of 2026, further optimizing the business environment in the water conservancy construction sector and safeguarding engineering quality standards from the source.