On-Premise Data Preparation for Construction Intelligence
Ertas Data Suite gives construction and engineering firms a secure, desktop-based pipeline to prepare project data — specifications, RFIs, submittals, safety reports — for AI training without sending proprietary project information to external services.
The Challenges You Face
Project Data Is Sensitive and Contractually Protected
Construction contracts routinely include confidentiality clauses that prohibit sharing project documents with unauthorized third parties. Uploading specifications, drawings, and cost estimates to cloud AI services can violate these agreements and expose firms to legal liability.
Construction Documents Are Uniquely Complex
Specifications follow CSI MasterFormat, drawings contain technical notations, RFIs reference multiple documents simultaneously, and submittals mix structured data with free-text descriptions. Generic data tools cannot parse these domain-specific formats effectively.
Institutional Knowledge Walks Out the Door
When experienced project managers and superintendents retire, decades of knowledge about project risks, cost estimation patterns, and safety procedures leaves with them. Capturing this expertise in AI-readable training data requires tools that domain experts can actually use.
Data Is Scattered Across Disconnected Systems
Project data lives in Procore, PlanGrid, Bluebeam, email archives, shared drives, and paper files. Consolidating this information for any purpose — let alone AI training — requires manual collection that nobody has time for.
How Ertas Solves This
Ertas Data Suite is a native desktop application purpose-built for turning messy, multi-source project data into clean, labeled training datasets. The Ingest module pulls data from PDFs, spreadsheets, CSV exports from project management platforms, and even scanned documents. The Clean module normalizes formatting, extracts relevant sections, and handles the domain-specific structures common in construction documentation.
The Label module lets project managers, estimators, and safety professionals annotate documents with their domain expertise — tagging risk factors, categorizing specification sections, or classifying RFI patterns. Because Data Suite runs entirely on the local machine, no project data leaves your network, and no cloud service ever sees your proprietary information.
The Augment module generates training data variations to fill gaps in underrepresented categories, and the Export module produces versioned datasets ready for model training — with a complete audit trail documenting every transformation from raw input to final output.
Key Features for Construction & Engineering Firms
Multi-Format Construction Document Ingestion
The Ingest module handles PDFs (including scanned blueprint sheets), DOCX specifications, CSV exports from project management tools, and structured data from estimating software — normalizing everything into a consistent format for processing.
Domain-Expert Labeling Interface
The Label module is designed for construction professionals, not data scientists. Project managers tag documents using terminology and categories they already understand — CSI divisions, project phases, risk levels, trade classifications.
Complete Data Sovereignty
Data Suite operates as an air-gapped desktop application. Install it on a project office workstation and process sensitive project data without any network connectivity. Contractual confidentiality obligations are satisfied by design.
Provenance-Tracked Exports
Every exported dataset includes complete lineage metadata — which source documents were ingested, what cleaning rules were applied, who created which labels, and what augmentation strategies were used. This traceability supports both internal quality assurance and external audits.
Why It Works
- Construction firms have used Data Suite to consolidate and label RFI data from multiple projects, building training sets for AI models that predict RFI categories and route them to the correct trade — reducing response times by 30%.
- The air-gapped architecture ensures compliance with typical AIA and ConsensusDocs confidentiality provisions without requiring legal review of third-party data processing agreements.
- Safety teams have prepared incident report training data using Data Suite's labeling interface, enabling AI models that flag high-risk conditions from daily field reports before incidents occur.
- Data Suite processes scanned specification documents and extracts structured text with section-level granularity, handling the multi-column layouts and reference-heavy formatting typical of CSI MasterFormat documents.
- The audit trail provides the documentation needed for ISO 19650 BIM compliance when AI is used in information management workflows.
Example Workflow
A general contractor wants to build an AI system that automatically classifies incoming submittals and routes them to the correct reviewer. A project engineer opens Ertas Data Suite on a workstation in the project office, ingests 5,000 historical submittals from a Procore CSV export and associated PDF attachments.
The Clean module normalizes submittal descriptions and extracts key metadata — specification section references, trade information, and product types. Senior project managers use the Label module to classify submittals by category, urgency, and responsible reviewer. The Augment module generates additional examples for under-represented categories.
The Export module produces a versioned JSONL dataset with complete provenance. The firm's technology team uses this dataset to train a classification model that automatically triages new submittals — saving project engineers hours of manual routing work on every new project.
Related Resources
Ship AI that runs on your users' devices.
Early bird pricing starts at $14.50/mo — locked in for life. Plans for builders and agencies.