BIM Data Quality Remains Critical in AI-Assisted AEC Workflows

During a panel discussion at Solibri Connect 2026 in Helsinki, industry representatives from Fira, SRV, and AE Partners examined the relationship between BIM data quality and AI automation. Panelists emphasized that while reliable project data enables AI to reduce construction errors, invalid upstream data risks introducing new mistakes. A major operational friction highlighted was contractors recreating bills of materials from 2D drawings due to lack of trust in architectural models, costing Finland's construction sector an estimated 400 million euros annually. To bridge this gap, structured classification standards such as Finland's RAVA3Pro for MEP products provide the definitions needed for downstream automation. The panel noted that current IFC entities and enumerations cover only a minor fraction of MEP objects and systems, necessitating standardized enrichments. AI integrations, including Model Context Protocol (MCP) servers for BIM software and emerging Graph Neural Networks (GNNs), can assist in identifying commercial components and proposing missing attributes. However, panelists concluded that AI requires consistent input models to minimize guesswork and deliver reliable outputs.

Key takeaways

• Panelists at Solibri Connect 2026 emphasized that AI in AEC requires reliable upstream data to avoid introducing new construction errors.
• Contractors recreating bills of materials from 2D drawings costs Finland's construction sector an estimated 400 million euros annually.
• Current IFC standards cover only 7% of MEP objects and 4% of system types, requiring standardized classifications to enable automation.
• Emerging tools like Model Context Protocol servers and Graph Neural Networks can infer properties, but structured input models remain necessary.