The construction industry is the least digitalised major industry on earth. The Construction sub-tree on know.2nth.ai is built around a thesis: that openBIM (IFC + open tooling) plus AI agents that read the model directly is the right architecture — cost-honest, vendor-portable, regulation-aware. Anchored in the Gridline Properties × 2nth.ai openBUILD AI partnership.
Five sub-skills across three bands — the open standards substrate, the operational layer, and the SA-specific compliance and partner work. openBIM is the centerpiece; everything else either consumes it or layers on top of it.
The cost-and-lock-in case in detail. Worked example for a 10-seat SA mid-market team: USD-billed proprietary licences vs ZAR-cost openBIM stack. FX exposure, data portability, post-contract data ownership.
From IFC to BOQ. AI takeoff engines that read IfcWall + Qto_WallBaseQuantities directly. Living rate database, regional adjustments, tender analysis with anomaly detection. Three weeks of QS work compressed to hours.
4D programme tied to the model. Material tracking with BIM-extracted quantities. Quality capture linked to IFC GUIDs. Safety intelligence from site photos. Digital RFIs with 5-day-to-5-hour response time. Cost control flagged 4-6 weeks ahead of overruns.
NHBRC, OHSA Construction Regulations 2014, SANS 10400, JBCC Edition 6.2 BIM provisions, NEC4 Y(UK)1, DPSA / NDPW handover. IFC property sets aligned to regulatory checks — queries, not paperwork.
The Gridline Properties pilot for openBUILD AI. From IFC takeoff through 4D programme + material tracking + handover. The case study every other openBIM-on-SA-mid-market project will read first.
Four positions that show up in every Construction sub-skill. They reflect a specific view: that the construction industry's data layer should be open, that AI agents are first-class consumers of that data, and that South African regulatory and commercial reality is a feature to design around, not an afterthought.
If the building's information lives in PDFs and spreadsheets, no AI can read it without scraping. If it lives in IFC, a Python loop reads it in three lines. Every Construction skill assumes IFC is the substrate.
buildingSMART standards (IFC, BCF, IDS) are ISO-grade open. Vendor APIs are rate-limited, billed, and version-controlled by the vendor's roadmap. Build on the standards; integrate with the vendors only where they earn it.
NHBRC, JBCC, NEC4 SA conditions, Section 25 procurement, B-BBEE, SANS 10400, OHSA CR 2014. SA construction reality differs from US/EU defaults; every leaf treats local regulation as the design constraint, not a footnote.
AI in construction is most valuable as a structured-data agent: takeoff, programme, cost, compliance. Computer vision on site photos is real but secondary. The leverage is in the data already produced by the design team — we just need to read it.
The Construction sub-tree leans heavily on Data (the model is queryable), Tech (Cloudflare R2 for IFC blob storage, agents that read it), and Design (the architectural authoring tools that produce it). Partners is where the Gridline Properties relationship lives.