BIMINAM AI AND DATA ENGINEERING WORKSTREAM MEETING, 30-04-2026
Date: 30 April 2026
Purpose of the Session
- Explore practical use of AI within the built environment
- Discuss AI for asset information, handover validation and compliance
- Share real-world experiences, challenges and emerging workflows
- Examine technologies such as AI agents, MCPs, vector databases and RAG systems
Key Themes & Discussion Points
1. AI adoption is accelerating
Survey results showed:
- Most participants are already using AI regularly
- Main tools used:
- ChatGPT
- Claude
- Gemini
- Copilot
- Common uses include:
- document creation
- report writing
- information searching
- spreadsheet analysis
- summarisation
- Biggest concerns: security, reliability and accuracy
Participants highlighted concerns around:
- hallucinations and incorrect outputs
- sharing sensitive project data
- lack of governance
- over-trusting AI responses
- uncontrolled AI agents
Strong consensus that:
AI outputs must still be checked by humans.
- AI works best for repetitive and time-consuming tasks
Key use cases discussed included:
- reviewing O&M manuals
- validating COBie data
- checking handover information
- extracting structured information from PDFs
- creating drawing registers
- summarising documentation
- querying IFC models and BIM data
- AI should support experts — not replace them
Participants agreed AI is valuable for:
- automation
- data processing
- pattern recognition
- reducing manual effort
However, compliance and safety-critical decisions still require:
- human oversight
- expert judgement
- validation and accountability
- Vector databases and RAG systems are becoming important
The session explained how vector databases:
- store information semantically (by meaning/context)
- improve document searching and retrieval
- enable scalable AI workflows
Benefits discussed:
- faster document searches
- improved audit trails
- semantic understanding of project information
- better retrieval accuracy
Retrieval-Augmented Generation (RAG) was highlighted as a way to:
- reduce hallucinations
- improve trust
- provide citations and evidence trails
- MCPs (Model Context Protocols) enable AI integration
A major technical discussion focused on MCPs.
These were described as:
- the “connector layer” between AI and systems/data
- allowing AI to interact with:
- APIs
- databases
- BIM models
- Revit
- IFC files
- IoT systems
Examples demonstrated included:
- querying COBie datasets
- interacting with Revit models
- navigating IFC files
- accessing structured asset data
- Governance and permissions are critical
Key warnings included:
- never allow unrestricted delete/write permissions
- always sandbox AI environments
- avoid working directly on live data
- use secure enterprise environments for sensitive data
Participants stressed the importance of:
- governance frameworks
- standards
- access control
- organisational AI policies
- AI maturity across the industry remains uneven
Discussion suggested:
- advanced AI use remains limited to “pockets” of the industry
- many organisations are still experimenting at chatbot level
- few have moved into agentic AI or structured AI pipelines
Barriers include:
- security concerns
- lack of skills
- governance uncertainty
- limited internal resources
- Emerging use cases in the built environment
Examples discussed included:
- AI-assisted competency assessment
- object recognition from 360 imagery
- fire stopping validation
- design optimisation
- virtual walkthrough data aggregation
- automated compliance checking
- AI-assisted BIM and digital twin workflows
- Need for standards, guidance and collaboration
The group agreed there is now a strong need for:
- AI standards for the built environment
- governance guidance
- best practice frameworks
- competency pathways for AI use
New BIMinAM workstreams were proposed around:
- AI standards
- validation
- insurance
- risk/HAZOP processes
- task management
- water quality workflows
Overall Conclusion
The session reinforced that AI is rapidly becoming a major enabler for:
- information management
- compliance validation
- asset data structuring
- operational efficiency
However, the discussion consistently returned to one principle:
AI must operate within controlled, validated and governed processes.
The future opportunity lies not in replacing professionals, but in combining:
- structured data
- domain expertise
- AI automation
- secure workflows
- human validation
to create safer, more reliable and more scalable asset information systems for the built environment.