22 / Real Estate & Construction
AI Bill of Quantities Automation for Data-Centre Tenders
We turned data-centre tender drawings and 3D models into normalized bills of quantities automatically, including the items experts used to add from memory, with every AI line citing its source or waiting for sign-off.
- Client
- Sterling and Wilson Renewable Energy Ltd
- Stack & approach
- Data CenterCAD / BIM ExtractionRule EngineSemantic MatchingFastAPI
The problem
Data-centre tender estimation at SWDC was manual and knowledge-dependent. Drawings took days to convert into usable BOQs, CAD tools missed inferred items, and many line items depended on discipline-specific thumb rules held in expert heads.
Solution & architecture
An AI-assisted estimation engine for SWDC's workflow. It extracts structured quantities from 3D sources, applies SWDC-approved rules for items not explicitly drawn, normalizes multi-format BOQs, and semantically matches lines against historical rates and vendor quotes. Every AI action either answers with citations or creates a proposal requiring human approval.
Why it is built this way
Expert rules encoded
Inferred items no longer live only in heads
Cited or approval-gated
Every AI action is traceable
Deterministic calculators
Weights, amperage, MW-basis exposed
