AI Building-Code Review — Accessibility Compliance

Context

A joint R&D engagement with a major general contractor's site teams and a national construction research institute. BIM is increasingly mandated in Korea, but 2D drawings handed to contractors often accumulate design changes and arrive without thorough code or consistency review — and manually checking hundreds of drawings against regulation is not realistic on large projects.

The bottleneck

Code-compliance review depends on engineers manually cross-checking drawings against regulation. On large projects — hundreds of drawings plus mid-construction changes — a full review is practically impossible, yet missed compliance creates rework and approval risk at handover.

What we built

A pipeline built on BIBIM AI: building drawings are cleaned, converted to BIM (geometry information, not OCR or image recognition), checked by accessibility-code logic embedded in the agent, and returned as a report, a pass/fail checklist, and CAD markups placed at the exact location on the drawing. The first application targets the Act on the Guarantee of Promotion of Convenience for Persons with Disabilities, the Elderly, and Pregnant Women — starting with geometry-determinable rules: ramp slope (≤ 1/12) and clear width (≥ 1.2m), floor level differences (≤ 2cm), handrail conditions, ramp landings (1.5m × 1.5m), door clear widths, accessible restrooms, tactile paving, and accessible parking.

Engagement

Outcome

Joint R&D PoC underway — validating automated-judgment accuracy against manual review, time reduction (N drawings reviewed within a day), and generalizability across multiple sites. The roadmap extends from accessibility code to in-house technical standards and automatic quantity take-off.

Tech stack