人工智能方案目錄

This solution is a BIM (Revit)-based AI quantity take-off and cost management system purpose-built for Hong Kong public works and Capital Works projects, built on three integrated modules: Cost Database, Quantity-Price Linkage Engine, and BIM Design Platform. It automates the complete pipeline from model-based quantity extraction to LMP rate application, outputting complete, audit-ready quantity-price reports in real time and in full compliance with HKSMM4/HKSMM5. Two distinct AI capabilities are embedded within the system's rule-driven architecture:

(1) Early-Stage AI Cost Benchmarking: At project inception, the user inputs a project overview comprising use type, GFA, structural form, finish standard, and major M&E systems. The AI large model applies semantic retrieval and cross-referencing across the historical cost database to automatically identify the most comparable completed projects and push recommended planned cost indicators as the design-stage baseline. This function directly delivers the AI-driven historical cost analysis capability for ensuring future Capital Works designs are cost-effective from inception.

(2) AI Semantic LMP Price Matching: During quantity take-off and bill compilation, the large model's semantic understanding automatically identifies and applies the most appropriate current LMP (Labour, Material and Plant) prices for every bill item, replacing manual rate-lookup with intelligent, context-aware matching. All computation remains rule-driven, transparent, and fully auditable by quantity surveyors.

Together, these two AI functions automate the most repetitive and error-prone stages of public works cost management, delivering measurable and verifiable improvements in accuracy and efficiency across estimation, tendering, and progress control.

Taking a Capital Works project at the design stage as a concrete example: designers continue to forward-model in Revit per standard specifications. No change to existing workflow, no additional modelling workload is introduced. Before quantity take-off begins, the AI large model has already retrieved and pushed the closest matching historical project cost indicators from the database as a planning baseline, delivering the same function envisaged by the Development Bureau's forthcoming Digital Cost Control Platform. After the BIM Quantity Engineer configures the HKSMM4/5 take-off rules, the system automatically extracts quantities from all modelled elements and supplements unmodelled components by rule. The large model then semantically matches and applies current LMP prices for every bill item, auto-generates a complete auditable quantity-price report, and immediately compares current design quantities against the planning baseline, triggering alerts for any item deviating beyond ±5%.

The measurable outcomes are: multi-day manual take-off and rate application is compressed to the hour level; human error is substantially reduced; ensures that the cost deviation rate meets the quality standards for cost estimation deliverables, as verified by testing; and the project team is equipped to identify cost risks and make design adjustments at the design stage, and supporting scheme comparison, tender cost control, and more effective stewardship of public funds. This directly responds to the objective of reducing construction costs and enhancing the cost-effectiveness of public works investment.