Fit-out projects generate huge volumes of drawings, quantities, material requirements, supplier quotations and project documentation.
We built an AI-powered operational platform for a Hong Kong fit-out business to connect these fragmented workflows.
The challenge
The team relied on manual processes across drawing review, quantity takeoffs, material estimation, supplier requests, quotation comparison, tender preparation and project reporting. Information had to be repeatedly moved between documents, spreadsheets, email and internal systems.
What we built
The platform connects the project lifecycle.
Core capabilities include AI takeoff review, drawing validation, takeoff summaries, material estimation, RFQ workflows, a vendor portal, supplier quotation comparison, internal quotation workflow, project reporting, tender response automation and multilingual support.
How it works
The system analyzes construction drawings and produces structured takeoff information for review.
The AI accelerates the first pass. The QS or project professional retains final control.
Architecture
Approved material requirements move directly into procurement workflows.
Vendor portal
Supplier interactions were brought into the broader workflow through a dedicated vendor experience, creating a structured path from requirement to RFQ to supplier response to comparison to decision.
Multilingual workflows
The platform supports multilingual project information, including Chinese and Traditional Chinese documentation, which was important for the Hong Kong operating environment.
The AI approach
The system combines AI document and drawing analysis with deterministic workflow logic and human validation. The AI proposes. The project team approves.
Validation
Takeoff accuracy is this platform’s central commercial claim, so it is checked against manually produced takeoffs before results are trusted downstream. The AI handles counting and measuring against the drawing set; the estimator reviews flagged items, checks complex assemblies the system is less confident on, and applies a final check before a takeoff is used in a bid. Accuracy is tracked as agreement with the estimator’s final, reviewed numbers, not the AI’s raw first pass.
Independent testing of AI takeoff tools reports 95% to 99% accuracy on clean vector drawings, with 8% to 12% error rates on complex commercial sets with dense annotations. This engagement’s own agreement rate is a separate figure still to be run.
Security and governance
Drawings, bills of quantities and supplier pricing are commercially sensitive. Access to drawings, bills of quantities and vendor pricing is scoped by role and by project, so a coordinator on one job cannot see pricing on another, and a supplier only ever sees its own quote, never a competitor’s. What is sent to AI models is limited to the specific drawing or document being analysed at that moment, not stored project history.
Outcome
The platform transformed fragmented construction workflows into a connected operational system, linking what was designed, required, procured and quoted.
Why it matters
Construction AI becomes valuable when it is embedded in project operations. Not another chatbot. Not another document viewer. A system that turns project information into operational action.
Related work
Ready to Turn Your Vision Into Reality?
Ready to build your Web3 or AI solution? Let’s talk and get started today.