BIMINAM AI AND DATA ENGINEERING WORKSTREAM MEETING, 19-03-2026
Date: 19 March 2026
(Primarily based on transcript, with key chat insights included)
Purpose of the Session
The session focused on practical applications of AI and data engineering to improve asset information quality, validation and usability, with a strong emphasis on moving from fragmented, document-based processes to machine-readable, structured and auditable data across the lifecycle.
Key Themes from the Transcript
1. AI as a tool for data validation, not a replacement for expertise
Participants shared real-world use cases of AI, including:
- predicting and filling missing data using historical datasets
- validating asset existence and condition (e.g. red/amber/green confidence models)
- automating checks of O&M and Health & Safety files
However, a consistent message was that AI outputs require expert verification, as quality control remains a key limitation.
2. The core problem: poor quality and fragmented data
The group reinforced that:
- asset data is generated from multiple inconsistent sources (design tools, manufacturers, contractors, CAFM systems)
- information is often incomplete, duplicated or incorrectly classified
- O&M documentation remains unstructured and difficult to validate at scale
This fragmentation is the primary barrier to effective digital asset management.
3. “Digital defects” and auditability of information
A significant concept discussed was that:
- missing or incorrect information should be treated as a “digital defect”
- this can be linked to contractual responsibility and defects liability
AI and data engineering can systematically identify these issues, enabling:
- earlier intervention
- clearer accountability across the supply chain
- improved compliance and assurance.
4. Shift from reactive fixing to proactive data creation
A key strategic shift identified:
- current industry approach = fixing poor data at handover
- required future approach = capturing correct data at source (“coal face”)
This requires: - better workflows for contractors and suppliers
- tools that make data capture easier and less manual
- embedding validation earlier in design and construction stages.
5. Standardisation, classification and interoperability challenges
The discussion highlighted major challenges around:
- inconsistent classification systems (COBie, IFC, Uniclass, etc.)
- lack of alignment across tools and supply chains
- emerging risks around ownership and monetisation of standards
There was strong agreement that common, open and accessible standards are essential for interoperability and collaboration.
6. AI-enabled automation of workflows and decision-making
Demonstrations showed how AI can:
- structure and decompose large document sets (e.g. O&M files)
- automatically generate compliance reports and action lists
- integrate multiple systems (e.g. asset data + sensors + APIs)
- enable natural language interaction with asset information
This represents a shift toward real-time, query-based asset management rather than static reporting.
6. Importance of context: spaces, assets and systems
A recurring principle was that data only becomes useful when contextualised, particularly through:
- spatial hierarchy (site → building → floor → space)
- asset relationships within systems
- linkage to operational and regulatory requirements
This structure is critical for both AI processing and human usability.
7. Collaboration and shared solution development
The workstream continues to position itself as a collaborative ecosystem, aiming to:
- bring together problem owners and solution providers
- share tools, processes and methodologies
- develop tested, scalable solutions rather than isolated innovations
This includes building a task-based framework and shared knowledge base across the community.
Key Insights from the Chat
- Strong emphasis on “common sense” asset information delivery, focused on what operators actually need rather than over-complex classification.
- Recognition that clients prioritise usable outputs (e.g. NRM, SFG20) over abstract data structures.
- Reinforcement that handover quality should be driven by operational needs, not just compliance requirements.
Overall Conclusion
The session highlighted that the future of asset information management lies in combining AI-driven automation with robust data standards, clear accountability and early-stage data capture. While AI provides powerful tools for validation, structuring and insight generation, the industry must address fundamental issues of data quality, standardisation and supply chain behaviour to unlock its full value. The workstream is therefore focused on creating practical, collaborative and scalable solutions that transform fragmented information into reliable, lifecycle-ready intelligence.