Construction project intelligence: a practical guide for AEC data teams

Construction project data has never been richer. Every BIM model contains thousands of element parameters. Every ERP transaction is timestamped and cost-coded. Every RFI, submittal, and transmittal is logged in project management platforms. The problem isn't data volume — it's that the data lives in five different systems that don't talk to each other, and producing a meaningful cross-system view requires a data engineer, a BI developer, and several weeks of integration work.

This guide is for AEC firms that want to get past the data silo problem and start asking the questions that actually matter for project control and delivery.

The five data sources that matter most

1. BIM models (Autodesk Construction Cloud / Speckle / direct IFC)

BIM models contain element-level information: quantities, materials, locations, assembly status, cost codes, and parameters that your project controls team assigns. For analytics, the most valuable fields are element counts by discipline, parameter completeness (are all required fields populated?), and the relationship between model elements and actual cost/schedule data.

2. Project management platforms (Procore, Aconex, ACC Issues)

RFIs, submittals, daily logs, meetings, punch lists, and transmittals. The most analytically valuable data: RFI response time by responsible party, RFI-to-change-order conversion rate, submittal overdue rates, and issue open/close trends by discipline.

3. ERP / cost management (Maconomy, Sage, JD Edwards, SAP)

Committed cost vs. budget, cost-to-complete forecasts, subcontractor payment status, and labour hours by cost code. This is where project profit or loss actually happens, but it's the hardest system to query because ERP schemas are famously complex.

4. GIS / PostGIS (site mapping, utility layers, survey data)

For infrastructure and civil projects: utility conflict zones, survey points, access constraints, environmental buffers. PostGIS spatial queries let you ask "which BIM elements intersect with the underground gas easement?" — questions that previously required a GIS specialist and a week's work.

5. Document stores (specs, contracts, RFI responses, change orders)

PDFs and DOCX files contain specifications, contract terms, and technical responses. Document RAG (retrieval-augmented generation) lets you ask "what does the contract say about liquidated damages for delay?" and get a cited answer with the exact paragraph reference.

The questions worth asking

Once you can query all five sources, the interesting questions are the cross-source ones:

Cost and BIM alignment:

• "Show me the cost per square meter for concrete elements, by floor level" — joins BIM quantities with ERP cost codes
• "Which discipline's elements have the highest cost-to-complete overrun this month?"
• "Find BIM elements in the mechanical discipline that have open purchase orders but no confirmed delivery date"

RFI and change order intelligence:

• "What's the average RFI response time by responsible party, and how has it trended over the last 3 months?"
• "Which subcontractors have the highest RFI-to-change-order conversion rate?"
• "Show me RFIs that have been open more than 14 days with no response, by responsible party"

Clash detection analytics:

• "How many clashes are open per discipline, and how has the count changed week over week?"
• "Which element types are involved in the most clashes — structural, MEP, or architectural?"
• "Show me the average clash resolution time by discipline"

Document compliance:

• "Which specification sections are referenced in the most RFIs?" (joins RFI data with extracted spec references)
• "Are there any contract clauses about liquidated damages that have been triggered?" (searches contract PDFs and joins with delay data)

IDS compliance (information delivery specification):

• "Which IFC elements fail the load-bearing wall specification in our IDS?" — runs a validation against the loaded IDS spec
• "What percentage of structural elements have all required parameters populated?"

A real workflow: weekly project health review

Here's what a weekly cross-system project health check looks like when all the data is connected:

Monday morning — 10 minutes:

1. "Show me cost-to-complete variance vs. budget by cost code, for this week vs. last week" → ERP query, bar chart with red/green variance
2. "Which RFIs have been open more than 7 days with no response?" → Procore query, table with responsible party column
3. "How many new clashes were detected this week, broken down by discipline?" → ACC Issues or BCF query, stat card
4. "Show me the top 5 submittals overdue this week" → Procore submittal log, table

This used to require four separate logins to four systems, four exports, and manual assembly into a slide deck. With a connected project intelligence platform, it takes 10 minutes and produces a shareable dashboard.

Common integration patterns

Pattern 1: BIM + ERP via cost code

BIM elements have user-defined parameters that include cost codes (your project controls team assigns these during model setup). ERP has transactions by cost code. The join key is the cost code — element.cost_code = erp.cost_code.

With a connected system: "show me labour hours this month per BIM element type, joined by cost code." Without: a two-day exercise to export both systems and build a VLOOKUP.

Pattern 2: RFI → Change Order → Cost Impact

The chain is: RFI is raised → RFI is approved → potential change event → change order → ERP cost entry. Each step lives in a different system. Tracing the full chain — "for every RFI that became a change order, what was the final cost impact?" — requires joining Procore (RFIs, change orders) with ERP (actual cost entries).

Pattern 3: Document clause → contract compliance

Upload the signed contract PDF as a document source. Ask natural language questions: "what is our notice period for force majeure events?" — the system retrieves the exact clause with a page reference. This isn't a database query; it's document RAG. But it sits alongside database queries in the same interface, which means you can ask "has the notice period been triggered in any of the project correspondence?" — searching both the contract text and the RFI/transmittal logs.

What BIM data actually produces useful analytics

Not all BIM data is equally valuable for analytics. The highest-value data from models:

Element counts by discipline and level — feeds into resource planning and cost forecasting by location. "How many MEP elements are installed on Level 3 vs. total planned?" connects model status to schedule.

Parameter completeness rates — "what percentage of structural elements have the fire rating parameter populated?" These questions drive model quality enforcement without a manual QA audit.

Element-level status parameters — teams that use model elements as progress tracking records (not all do, but many large GC firms now mandate it) can produce earned value analysis directly from the BIM model.

Clash volumes and trends — clash data from Autodesk or from BCF files shows where coordination is breaking down before it becomes a cost problem.

COBie data quality — for FM handover projects, COBie spreadsheet coverage rates (which equipment has manufacturer, model, warranty data) determine how useful the handover will be.

What requires more than just analytics

Data-driven project management doesn't remove the need for domain judgment. The analytics surface the problems — the project manager still decides what to do about them.

Some things AI tools specifically can't do:

• Interpret contract terms in context of a dispute — get a lawyer. AI can retrieve and summarize clauses, not evaluate their applicability to a specific situation.
• Determine if a design decision is structurally sound — that's what structural engineers are for. AI can query whether a given element meets a specified parameter threshold; it can't verify the parameter threshold is correct.
• Predict subcontractor default risk — payment data and milestone data are signals, not predictions. The judgment call is human.

The value proposition is time: the project controls team spends less time assembling data from five systems and more time interpreting the numbers and making decisions.

Getting started: a practical path

Week 1: Connect your Procore (or Aconex) data. Start with RFI analytics — response time, volume by discipline, conversion to change orders. This is the lowest-friction win and immediately valuable.

Week 2: Connect your ERP (or cost management export). Add cost variance analysis. Cross-reference with RFI-driven change orders.

Week 3: Upload your BCF file (clash data) or connect Autodesk Construction Cloud. Add clash analytics.

Week 4: Upload your BIM model (IFC) or connect ACC for element-level data. Start asking cross-source questions.

Month 2: Add document RAG (contract PDF, specification documents). Ask questions that span structured data and contract terms.

This progression works because each step delivers value independently. You don't need all five sources connected before you see ROI.

FAQ

What's the difference between SiteMind and iDBQuery for construction? SiteMind is the construction-vertical brand built on iDBQuery. It has the same AI engine plus pre-built connectors for Autodesk Construction Cloud, Speckle, IFC, BCF, IDS, COBie, Maconomy, Procore, and Aconex — plus PostGIS integration for site GIS layers. Non-construction users see iDBQuery; construction users see SiteMind.

Does iDBQuery / SiteMind write data back to Procore or Aconex? RFI creation and submittal status updates are available as opt-in write-back actions on Procore. Aconex transmittal status write-back is also available. All write actions are logged in an immutable audit trail. BIM sources are strictly read-only.

How does BIM element querying work with IFC files? Upload an IFC file directly. iDBQuery parses it using ifcopenshell, extracts elements and properties into a queryable schema (equivalent to what Autodesk ACC exposes), and makes them available for natural language queries. You can ask "how many IfcWall elements with a fire rating parameter of 2 hours are on Level 3?" and get a count in seconds.

Can I connect Procore and Autodesk ACC in the same project? Yes. Add both as sources to the same iDBQuery project and ask cross-source questions. "Show me BIM elements on Level 3 that have open RFIs in Procore" requires linking the element's IfcGuid (from ACC) with the RFI's linked element reference (from Procore) — the AI handles this if both sources are connected.

What ERP systems are supported? Native connectors: Deltek Maconomy, Procore (cost management), Aconex. For other ERPs (Sage, JD Edwards, SAP, Oracle Fusion), iDBQuery connects to the underlying database or works from exported data. Direct REST API connectors for additional ERPs are on the roadmap.

Is this suitable for small firms (under 50 employees)? Yes. The most valuable workflows — RFI analytics, cost variance, clash trends — are valuable even on small projects. The free tier (3 sources, 5 reports) covers a single-project deployment at no cost.