← Cases
United StatesConstruction and engineering2024 – ongoing

Gilbane

TrunkText document retrieval at the Baird Center

Gilbane's Baird Center team validated 87% of TrunkText answers in a vendor-reported pilot

WorkingVendor case study

Problem

The Baird Center expansion generated around 21,000 documents. Finding requirements across drawings and contracts took supervisors away from site work.

Approach

Gilbane began piloting TrunkText in January 2024. Workers queried the large language model tool from phones or computers after project documents were uploaded. The team also used it to reconcile requirements across documents.

Results

87%answers validated as correct by the project team across 246 questions, according to a Trunk Tools case study reported by Construction DiveVendor case study[1]

Takeaway

Project staff checked the answers rather than treating retrieval as proof of correctness. The reported accuracy came from the supplier's case study.

Categories

Innovation type
Process innovation
AI technology
Generative AI
Evidence
Vendor case study

Sources

  1. How Gilbane used an AI tool to track 21,000 documents
    Construction Dive28 Aug 2024News article

Source summaries

[1]Construction Dive28 Aug 2024

Construction Dive interviewed Gilbane's superintendent about the pilot. It repeated a vendor case study reporting 246 questions and 87% answer accuracy after project-team validation.

Full details are in the original sources.

Added 1 Nov 2026 · Updated 1 Nov 2026
Open case page ↗