AI-based SaaS For Architectural Drawing Recognition

Challenge

Construction cost estimation is a complex, multi-stage process involving dozens of specialists and hundreds of architectural drawings. General contractors distribute project documentation across multiple trades (structural, finishes, MEP, and others), and each subcontractor prepares their own estimate based on material quantities and labor costs.

At the core of this workflow lies quantity takeoff: measuring walls, rooms, fixtures, and other elements directly from drawings. Traditionally, this process relies on dedicated desktop takeoff tools where estimators manually draw measurement lines and count objects. The resulting quantities are then transferred into estimation software to calculate costs.

Despite the maturity of construction estimation software, takeoff remains highly manual, slow, and error-prone. Existing tools separate takeoff and estimation into different applications, creating fragmented workflows and operational friction.

Our client, an established provider of construction estimation software, approached us to build an AI-powered takeoff SaaS solution and integrate it into their core estimation product, transforming two disconnected tools into a single, cloud-based cost estimation ecosystem.

Solution

We developed a cloud-based AI takeoff SaaS platform that automatically processes architectural drawings, extracts quantities, and delivers structured data directly into the construction cost estimation workflow.

The solution operates as a core module within a broader SaaS cost estimation system, enabling general contractors and subcontractors to perform takeoff, validation, and estimation inside a single unified product.

Universal Architectural Plan Solution

The platform supports a wide range of architectural and construction drawings, including:

• Floor plans,
• Reflected ceiling plan (RCP),
• Electrical plans,
• Mechanical plans,
• Detail architecture drawings,
• Demolition plans,
• Interior elevation plans.

The AI automatically detects drawing types, scale, and layout, supports multiple plans per page, and generates a structured table of contents for large multi-page PDF documents. This allows users to efficiently navigate and process complex project documentation directly in the browser.

AI-Based Quantity Takeoff

Using computer vision models, the SaaS platform automatically detects and measures construction elements such as:

• Walls, doors, and windows
• Rooms and enclosed spaces
• Electrical and plumbing symbols
• Ceiling hatchings and technical annotations

The system reconstructs room geometry from detected walls, calculates room areas (with and without wall thickness), wall lengths, and produces a complete bill of quantities. All measurements are stored in a structured format and can be immediately consumed by estimation software.

Detection of Custom Symbols and Labels

Users can select any symbol or label on a drawing, and the system will automatically detect and count similar elements across the entire document.

Detected symbols, such as wall type labels or technical annotations, are grouped into configurable categories and can be linked to structured spreadsheets containing additional metadata. This allows takeoff data to be enriched with trade-specific attributes and pricing logic.

Complex Spreadsheet Detection

Architectural and construction drawings often include embedded multi-page spreadsheets with merged cells and non-standard layouts. Off-the-shelf OCR solutions frequently break these structures and corrupt the data.

We built a dedicated SaaS subsystem that converts PDF-based spreadsheets into Excel-compatible formats while preserving layout, merged cells, and data integrity. This ensures schedules and specifications can be reused reliably within the estimation workflow.

Furniture Detection and Catalog Matching Module

As part of the extended takeoff workflow, we introduced a new AI-powered SaaS module for furniture detection and catalog matching, designed for furniture contractors and suppliers.

The system automatically identifies furniture items on drawings, such as cabinets, wardrobes, and built-in units, extracts their dimensions and structural characteristics, and matches them against a manufacturer’s product catalog.

The pipeline includes:

• PDF furniture catalog preprocessing and document structuring
• Extraction of dimensions and annotations
• Computer vision-based furniture detection
• LLM-based semantic analysis of furniture attributes (size, shelves, doors, locks, mounting type)
• AI-driven search across a structured product catalog

The solution combines multiple computer vision models with large language models to generate accurate furniture descriptions and identify the closest matching catalog items. This eliminates manual visual comparison and accelerates furniture-related cost estimation.

Feature Rich User Interface

The system is delivered as a full-featured SaaS application tailored for construction professionals:

• Centralized cloud storage for drawings
• Automated document structuring and navigation
• Manual review and correction of AI detections
• Custom labeling and object configuration
• Flexible report generation and data export

Users can generate detailed, data-rich reports that include calculated quantities, areas, lengths, and other metrics required for accurate cost estimation.

Results

The solution became a core AI takeoff module within a larger SaaS construction cost estimation ecosystem, unifying takeoff and estimation into a single cloud-based workflow.

It significantly reduced manual measurement effort, improved estimation accuracy, and accelerated project cost analysis for general contractors and subcontractors. The platform has been successfully adopted in real-world construction workflows and continues to evolve.