Finnish Universities Test Agentic AI for MEP BOM and Measure Construction Productivity Impacts
AI assistants · MEP · BIM collaboration
Researchers from Tampere and Aalto universities presented preliminary findings on AI in construction. One project used agentic AI on an IFC model and local LLM to generate procurement-ready MEP Bills of Materials in hours. Another project evaluated AI's impact across construction, concluding that while knowledge-intensive tasks see significant productivity gains, overall project-level savings remain limited because roughly 80% of work takes place on-site.
Preliminary results from two Finnish construction digitalization research initiatives were presented by researchers from Tampere University and Aalto University. Under the Talotekniikka 2030 consortium, the Tampere research group developed a proof of concept demonstrating that agentic AI can automate the Bill of Materials (BOM) creation process using unstructured and non-ideal project data. The workflow started with a mechanical, electrical, and plumbing (MEP) design model in IFC format, utilized an agentic workflow powered by a locally hosted large language model (LLM) without API token expenses, enriched the model with external product information, and produced procurement-ready material lists in hours rather than the days required by human estimators. Concurrently, the Building 2030 consortium led by Aalto University examined the broader productivity effects of data-driven construction and AI. Researchers reported that automation yields sharp efficiency improvements in upstream, knowledge-intensive phases such as design, cost estimation, procurement, and management. However, because approximately 80% of construction work occurs on-site where AI automation currently exerts less direct impact, whole-project productivity gains remain harder to quantify. Industry discussions noted that AI tools may also expand design services and improve overall project collaboration rather than merely cutting hours.
Key takeaways
- Tampere University researchers demonstrated an agentic AI proof of concept that converts an MEP IFC model and external data into a procurement-ready BOM using a local LLM.
- Preliminary tests showed the agentic BOM process completed in hours what typically takes a human estimator days, while capturing details human estimators might overlook.
- Aalto University's Building 2030 project found AI delivers substantial efficiency in design, estimation, and procurement, but whole-project gains are constrained because 80% of construction work is on-site.
- Industry startup approaches focus either downstream on using agents to clean up siloed data or upstream on generating standardized, machine-readable design outputs.
Source：AEC Business Read the original ↗
Prepared, checked and automatically published by AI; not individually reviewed by a human.