Building AI-Assisted Document Processing for Construction Workflows

Large construction programs like the Alinma Bank Data Center Tier IV project generate a constant stream of technical submittals, MIR/WIR records, and correspondence across Aconex (the industry-standard construction document platform) and SharePoint (used internally). The two systems don't share a common record format, which means reconciling what exists in one against the other has historically been a manual, error-prone process — and missing records tend to surface late, when they're contractually expensive.

The first piece of the automation toolset classifies incoming technical documents and routes them between Aconex and SharePoint automatically, rather than relying on someone manually reading and filing each one. Classification accuracy matters more than speed here — a misrouted technical submittal is worse than a slow one.

Technical submittals (MIR/WIR — Material/Work Inspection Requests) contain review comments that engineers previously extracted by hand into Excel reports. The automation extracts those review comments directly from the submittal documents into structured Excel output, which only works reliably if the extraction logic accounts for the layout and formatting variance across different submittal templates rather than assuming one fixed format.

Even with classification and extraction automated, records can still go missing between systems. The log-reconciliation component compares what exists in Aconex against what exists in SharePoint and flags discrepancies, so missing records get caught during routine reconciliation instead of during a contractual dispute months later.