This audio is automatically generated. Please let us know if you have any comments.
Keeping track of every detail of every contract a builder signs before entering a job site can be a Sisyphean task. You can roll that metaphorical rock down that mountain, but something, some detail, has a chance to derail progress. Andrew Roy of Gilbane Building Co. knows this well.
Roy, a Senior Superintendent, supervised the Baird Center Expansion Projecta $456 million renovation of the Milwaukee convention site, as part of a joint venture with CD Smith of Fond du Lac, Wis.
Together, Roy and his team added approximately 300,000 square feet to the existing structure, which originally opened in 1998, including 24 new meeting rooms, more than 400 indoor parking spaces, six loading docks and a kitchen executive, as reported. project information page. The group began in 2021 and was quickly completed, resulting in a May 2024 finish.
Thousands of documents, one chatbot
But getting there wasn’t easy. The job specifications and contracts totaled some 21,000 discrete documents, a problem that has become increasingly common in today’s more complex versions.
“It’s humanly impossible to be completely familiar with every project document and every change or conversation there is,” Roy said.
That’s where New York City-based Trunk Tools came into play. The company makes an AI-based tool that contractors can use to track a project’s documents and contracts for immediate answers without leaving the job site.
In order to navigate the sea of data contained in the Baird Center files, Roy’s team uploaded all of their documents to the platform, including drawings, RFIs, contracts and change orders.
Once they did, the large language model behind the tool, a chat-like feature known as TrunkText that builders can access from mobile devices or a computer, was able to answer questions, answer queries and reduce the time it took to find details when working on a problem.
“We stumbled upon it and then started using it for coordination between documents, so the coordination between the door hardware program and the electrical drawings, or the low voltage drawings, or the life safety drawings ” Roy said. “That’s where it became extremely powerful for us.”
TrunkText is just one construction-oriented AI offering in a rapidly expanding field; other examples include DocumentCrunchthat can search, evaluate and flag contracts for users based on workplace questions and risks, and Togal.AIwhich uses deep machine learning to help estimators accurately.
Gilbane began using TrunkText in the back third of the Baird Center project’s life, piloting it in January 2024. Roy’s team leveraged it for interior finishes, interior construction, and much of the exterior work and the outer enclosure, Roy said. Gilbane declined to share the cost of the service.
Avoiding rework
Roy pointed out a question the team had about a fireplace in a feature wall with a large exhaust system covered in acoustical plaster. The material for the finish took six months to acquire from Europe. Roy and the mechanical contractor performed an inspection and noticed an anomaly in the piping.
They had a question: Should the seams be sealed? A mistake would mean costly rework, equipment rental costs and wasted time. Normally, answering these types of questions would involve a time-consuming email chain between the mechanical contractor, the design team, and Roy that would take hours, if not days.
Instead, they asked TrunkText.
“Sure enough, within 20 seconds, I had five or six different documents in front of my face where I was and a text response saying, yes, the seams need to be sealed in this duct to create a proper vacuum.”, Roy said. “If the seams are not sealed, the warranty will not be valid and the system may not work properly.”
This is just one example of the hundreds of queries that workers submitted throughout the project.
In fact, a Trunk Tools case study showed that users asked 246 questions over 37 business days at work at the Baird Center. Of these responses, 87% were correct, as validated by the project team. The case study claimed that using TrunkText allowed workers to save 20 to 40 minutes of search and travel time for each query and avoided more than $100,000 in rework per month, according to the case study.
To underline the value of these savings, Roy points out how he has traditionally had to search for information on projects. Typically, when a question comes up, he has to dig through his iPad and manually find the right documents, or go back to the jobsite trailer to look at drawings, contracts or other documents to get the answer. Along the way, subcontractors may stop you with other questions or you may have to deal with a security issue.
Instead, TrunkText answered their questions “within five to 10 seconds,” Roy said.
Roy said the implementation was relatively painless: it only took a day or two to have Trunk Tools linked to Gilbane’s Procore and Sharepoint deployments.
One of the steepest parts of the learning curve in using the tool was figuring out how to ask questions the right way, including providing context to help TrunkText drill down to the right problem, such as behaviors around the fireplace.
But even then, sometimes the tool didn’t come back with any response. In these cases, Roy and his team quickly deduced that the information they were looking for was not in the documents. He described it as a “happy accident” that provided insight into potential blind spots in the documentation.
Otherwise, his team’s biggest win was simply not having to sift through the vast amount of data that is the norm on a modern construction project. In this case, that meant up to 34 gigabytes of information in those 21,000 files.
“Many of the successes we saw with TrunkText in our project [was] their ability to search the documents,” Roy said.
