Any complex construction project generates thousands of pages of documents and specifications, along with a stream of change orders and RFIs. Finding the right information and communicating it effectively to field personnel has long been the bane of project managers and superintendents. One startup is asking: “Can we text you that?”
Trunk Tools, the tech company that aims to incentivize work crews to meet targets by automating and formalizing the process of extra pay for doing tasks by preset times, has introduced an AI-based system dubbed TrunkText that can help PMs and superintendents find key project information in a few minutes instead of a few hours, all by sending texts on a smartphone right from the jobsite.
“In a $500-million high-rise project, there’s 120,000 project files with an average of 30 pages—a total of 3.6 million pages,” says Sarah Buchner, founder and CEO of Trunk Tools. “A construction superintendent would take 50 years to read all of that information; TrunkText only needs seconds.”
Buchner says she and her development team had to build out a large language model (LLM) trained on construction terms which could quickly reference files such as PDFs and 3D models for relevant information. Over a period of three months, Trunk Tools tested it out with large contractors who were having difficulties using ChatGPT and other mass-market LLMs that have no point of reference for construction terminology.
“It’s actually trained for how construction works,” says Buchner.
After submitting project materials to the system, TrunkText offers PMs, superintendents and other stakeholders a number for texting questions about the project.
Buchner says the AI’s training allows it to answer questions such as “what diameter rods are being proposed for use in this anchoring system?” and other specific queries. TrunkText is designed to respond within 15 seconds, with the relevant documentation attached to its responses.
Video courtesy of TrunkTools
TrunkText at Work
TrunkText is now being used in pilot programs on construction projects with more than $2.4 billion in combined contract value. Contractors who have tried it say it has saved them time.
“We have seen our superintendents emerge as the leading users of TrunkText for quick answers while out in the field,” says Parker Mundt, vice president of platform, at Suffolk Technologies, a venture-capital construction technology company affiliated with Boston-based contractor Suffolk Construction. Suffolk has been using TrunkText on a project in Lowell, Mass., for about two months.
Trunk Tools is also part of the most recent class of tech companies that went through Suffolk Technologies’ BOOST startup accelerator program, allowing Trunk Tools to work closely with Suffolk employees on real projects as they further refine their product offerings.
Given how project data can be relatively disorganized on a good day, Buchner says TrunkText works best with well curated data sets. Mundt agrees. “The engagement levels are highest when TrunkText has access to analyze a project’s entire document set in real time. The time savings per question asked to TrunkText is at a minimum five minutes, depending on how many documents would need to be parsed through to find an answer traditionally,” says Mundt.
“Instead of walking through the trailer and trying to find documents, having it right there on your phone [can] reduce rework up to 65%,” Buchner claims.
TrunkText has also been used in a pilot effort by DPR Construction. “We have seen strong potential from their team and product using large language models for the construction industry,” says Yash Lalwani, a project manager at DPR.
Trunk Tools came out of stealth mode earlier this year, securing a $9.9-million round of seed funding in July. Innovation Endeavors is Trunk Tools’ largest seed round investor with contributions from Fifth Wall, Foundation Capital and others. Suffolk Technologies is one of Trunk Tools’ seed round investors as well. “We see tremendous potential for generative AI in the built world,” says Mundt.