AI Agents
TL;DR: The Aedis AI Assistant uses seven specialized agents, each with unique skills, to deliver accurate, context-aware responses. The Primary Agent coordinates all specialist agents to handle your requests intelligently.
Multi-Agent Architecture
The Aedis AI Assistant is powered by seven specialized agents, each with a dedicated responsibility and skillset. This multi-agent approach ensures more reliable, domain-specific answers and significantly improved accuracy in complex workflows.
Rather than having a single AI try to handle all tasks, AedisPro uses specialized agents that excel in their specific domains. The Primary Agent coordinates these specialists to deliver the best possible response to your queries.
Why Multiple Agents? Different tasks require different expertise. Just as you wouldn't ask a structural engineer to design mechanical systems, each AI agent specializes in specific aspects of construction documentation, measurement, and analysis. This specialization leads to better accuracy and more reliable results.
The Seven AI Agents
1. Primary Agent
Role: Coordinator and Orchestrator
The central controller that manages all other agents and coordinates responses:
- • Understands and interprets your request
- • Determines which specialist agent(s) should handle the task
- • Combines results from multiple agents when needed
- • Ensures responses stay relevant to your project context
- • Manages workflow between different specialist agents
Think of this as the "brain" of the system—it decides how to best answer your question by delegating to the right specialists.
2. Reader Agent
Role: Drawing Interpreter
Reads and interprets the currently opened drawing or markup page, understanding visual elements and context.
Capabilities:
- • Identifies elements visible on the current sheet (doors, windows, walls, rooms, etc.)
- • Understands spatial context and relationships between elements
- • Recognizes symbols, annotations, and labels
- • Assists with takeoff preparation by identifying measurable elements
- • Interprets drawing notes and references
Example Questions:
- • "What rooms are shown on this page?"
- • "Where is the door schedule referenced?"
- • "How many windows are visible on this floor plan?"
- • "What grid lines are shown on this structural sheet?"
3. Document Agent
Role: Project Document Specialist
Searches and answers questions from all preprocessed documents uploaded to your project.
Searches across:
- • PDF drawings and construction documents
- • Specifications and technical documentation
- • Reports and analysis documents
- • Schedules (door, window, finish, etc.)
- • Meeting notes and project correspondence
- • Any other uploaded project files
Example Questions:
- • "Find fire rating requirements in the project specs"
- • "What does the structural note say about slab reinforcement?"
- • "Show me all Type A windows from the schedule"
- • "What waterproofing is specified for the basement?"
The Document Agent uses semantic search to find relevant information even when exact keywords don't match.
4. Estimator Agent
Role: Measurement and Takeoff Specialist
Handles measurement-based tasks and performs quantification from drawings.
Capabilities:
- • Measures lengths, areas, counts, and volumes
- • Performs takeoffs on the currently opened markup
- • Understands measurement rules and conventions
- • Generates structured output suitable for BOQs
- • Applies measurement standards and best practices
Example Instructions:
- • "Take off the doors on this sheet"
- • "Measure the floor finish area of Room 1.05"
- • "Count all wall-mounted light fixtures"
- • "Calculate the concrete volume for this foundation plan"
The Estimator Agent acts as your digital QS for repetitive and time-consuming measurement tasks.
5. Assembler Agent
Role: BOQ Creation Specialist
Specializes in creating and structuring Bills of Quantities with proper organization and formatting.
Capabilities:
- • Generates BOQ structures from measurements and takeoffs
- • Organizes items hierarchically with proper work breakdown
- • Applies industry-standard BOQ formatting
- • Creates item descriptions following standard conventions
- • Groups related items logically
- • Applies proper item numbering and referencing
Example Tasks:
- • "Create a BOQ structure for the concrete works"
- • "Organize these measurements into BOQ format"
- • "Generate item descriptions for the measured elements"
The Assembler Agent ensures your BOQ is well-organized and follows industry conventions.
6. Analyser Agent
Role: BOQ Quality Reviewer
Reviews and validates BOQ structure, content, and completeness.
Capabilities:
- • Identifies gaps or inconsistencies in BOQ items
- • Suggests improvements to item descriptions
- • Validates measurement units and quantity calculations
- • Checks for completeness and missing items
- • Identifies potential duplications
- • Reviews adherence to measurement standards
Example Questions:
- • "Review this BOQ for completeness"
- • "Are there any missing items in the concrete section?"
- • "Check if the item descriptions follow standard conventions"
This agent acts as a quality control reviewer, ensuring professional BOQ standards.
7. Research Agent
Role: Construction Knowledge Expert
Provides general construction knowledge, industry standards, and best practices.
Knowledge Areas:
- • Material descriptions and specifications
- • Standard methods of measurement (SMM, CESMM, etc.)
- • Construction methods and best practices
- • Terminology and definitions
- • Industry standards and code references
- • General QS and estimating principles
Example Questions:
- • "Explain ceiling coffers in simple terms"
- • "What is formwork striking time for concrete?"
- • "How do you measure formwork according to SMM7?"
- • "What's the difference between a lintel and a beam?"
This agent makes AedisPro accessible to new QS users and provides quick reference for experienced professionals.
How Agents Work Together
The Primary Agent orchestrates the other agents based on your query. Here's how they collaborate on different types of requests:
Simple Question
"What is a soffit?"
Agent Used: Research Agent alone provides the definition and explanation.
Document Query
"What concrete grade is specified for the ground floor slab?"
Agents Used: Document Agent searches specifications → Primary Agent presents answer with source references.
Measurement Request
"Take off all doors on this floor plan"
Agents Used: Reader Agent identifies door locations → Estimator Agent performs measurements → Assembler Agent creates BOQ items → Primary Agent presents results.
Complex Analysis
"Review my concrete BOQ for completeness based on the structural drawings"
Agents Used: Reader Agent reviews drawings → Document Agent checks specifications → Analyser Agent reviews BOQ → Primary Agent synthesizes findings and recommendations.
Important Reminder
While the AI agents are highly capable, they are assistants—not replacements for professional judgment. Always verify quantities, BOQ entries, and extracted information before using them for estimates or project deliverables.
Disclaimer: Results Depend on Drawing Quality
The accuracy and reliability of AI-generated results are directly influenced by the quality, clarity, and completeness of your project drawings. Understanding these factors helps you set realistic expectations and optimize your drawings for better AI performance.
Factors Affecting AI Performance:
Drawing Resolution and Clarity
• Low-resolution or blurry scans significantly reduce text recognition accuracy
• Poor image quality affects element detection and measurement precision
• Recommended: Use high-resolution PDF exports (minimum 300 DPI) from CAD software
Drawing Type and Format
• CAD-generated PDFs with vector graphics produce the best results
• Hand-drawn or heavily marked-up drawings may have reduced accuracy
• Scanned paper drawings are more challenging than digital exports
• Raster images (photos of drawings) provide the least reliable results
Drawing Standards and Conventions
• Drawings following standard conventions (AIA, BS, ISO) are easier for AI to interpret
• Non-standard symbols or unconventional notation may confuse AI agents
• Consistent labeling and annotation practices improve accuracy
• Clear, legible text and proper line weights help element identification
Drawing Completeness and Organization
• Missing schedules, legends, or specification notes limit AI capabilities
• Incomplete or partially detailed drawings may produce incomplete takeoffs
• Well-organized drawing sets with clear sheet references improve cross-document analysis
• Revision clouds and outdated information can cause confusion
Drawing Complexity
• Highly complex drawings with dense information may require more verification
• Overlapping elements or cluttered layouts reduce detection accuracy
• Multi-layered composite drawings are more challenging than single-discipline sheets
• 3D views and perspectives are less reliable than standard plan/elevation views
Text and Label Quality
• OCR (Optical Character Recognition) quality depends on font clarity and size
• Handwritten notes are less reliably interpreted than typed text
• Rotated or curved text may not be properly recognized
• Sufficient contrast between text and background is essential
Scale and Dimension Information
• Drawings without clear scale information require manual calibration
• Mixed scales on the same sheet can lead to measurement errors
• Proper dimension lines and measurements improve AI understanding
• NTS (Not To Scale) drawings should be clearly marked to prevent misuse
Best Practices for Optimal Results:
- ✓ Export drawings as high-resolution vector PDFs directly from CAD software
- ✓ Ensure all text is clear, legible, and uses standard fonts
- ✓ Include complete schedules, legends, and specification references
- ✓ Follow industry-standard drawing conventions and symbols
- ✓ Remove or clearly mark superseded information and revisions
- ✓ Provide proper scale information and dimension references
- ✓ Avoid excessive clutter—use separate sheets for different disciplines when possible
- ✓ Always verify AI results, especially for complex or non-standard drawings
Remember: The AI agents work with the information available in your drawings. Higher quality inputs lead to more accurate outputs. When working with challenging drawings, expect to perform additional verification and manual corrections.
Benefits of Multi-Agent Architecture
- ✓ Higher Accuracy – Specialists excel in their domains
- ✓ Better Context Understanding – Each agent understands its specific area deeply
- ✓ Improved Reliability – Purpose-built agents make fewer mistakes than general-purpose AI
- ✓ Scalability – New specialist agents can be added for emerging needs
- ✓ Coordinated Responses – Primary Agent combines insights from multiple specialists
- ✓ Domain Expertise – Each agent trained specifically for its construction/QS role