Trunk Tools CEO Dr. Sarah Buchner Spotlighted in Construction Today
Article Title: “Trunk Tools’ Founder and CEO, Dr. Sarah Buchner, on turning construction’s data overload into actionable, AI-driven decisions”
Excerpt:
To begin with, could you introduce yourself and share an overview of your career history and how you came to be in your current position?
I grew up in a pretty poor environment on a farm in Austria. Working on the farm and with my dad in his carpentry work was part of what you did to put food on the table. I started helping my dad around the age of 12, and eventually he took me out on construction sites. I liked the people. They were authentic and there was no BS. And I loved building, so I got stuck in in this industry. I was a blue-collar worker for about eight years and eventually became a foreman and superintendent.
Eventually I switched to the GC side because I thought that’s where the money was. But that was a tough landing. I worked my way up at Strabag (a $20 billion construction company in Europe) working on bigger projects. On one high-rise job in Berlin, I had a fatality on my job. That was the moment when I started thinking about whether I wanted to continue to manage jobs or to try and make more systemic change.
After 15 years in the field, I fell into software by building a health and safety app. At the same time, I had a mentor at Strabag who suggested that if I wanted to become something as a woman in construction, I needed a PhD. The company paid for it under the condition that I used the research time to analyze the company’s data and figure out what we could do with data in construction. In doing the research, I realized how challenging it was to even access our data because it is very unstructured, siloed, and constantly changing. As such, part of my PhD was early data extraction using machine learning before AI was all the rage.
I always loved the United States and thought it was the best place to build a company, so I came to Stanford for business school in 2019. This opened up fundraising and talent doors that could only happen in the US.
When I started Trunk Tools in 2021, the goal was really to leverage advancements in AI to make construction data accessible and actionable to construction firms so they could focus on actually building rather than digging through millions of pages of documents to find answers to their questions.
What specific problems on large-scale construction projects motivated you to build Trunk Tools, and why were existing solutions falling short?
The biggest problem that motivated me to start this company was the sheer amount of documentation on big construction projects. And the amount of documentation has actually increased since the field moved to email and project management systems.
For example, one of our customers was building a 400-foot tower in Manhattan that had 3.6 million pages of project documentation. If you stacked up all that paperwork, it would be three times the height of the building! Unfortunately, this is very common. It is humanly impossible to process that amount of data with a handful of project managers and superintendents. It would take a human 50 years to read through all of this. It takes AI about 30 seconds.
Before AI-native platforms like ours were available to project teams, people would spend around 20 percent of their time just looking up information. Accounting for our Q&A agent, TrunkText, alone, our customers average around 20-to-30 minutes of time savings for every question they ask the agent.
Construction teams deal with massive volumes of unstructured data. What are the biggest risks of not managing this data well?
We think of unstructured data as causing two main problems: (1) the difficulty and slow speed in accessing the information you need and (2) the increased likelihood of discrepancies or missed details that lead to mistakes on the project.
The first of these problems tends to result in RFIs. In our own research, we’ve found that on a typical $100 million project, around 1000 RFIs are submitted. These cost our customers on average about $3000 per RFI. Eighty percent of these are addressable by better access to unstructured data buried in their project documentation, meaning we can help them avoid up to $2 million in costs. In a business where GCs work off of one-to-three-percent profit margins, these savings make a big difference.
Regarding the second problem, we’ve seen lots of opportunity to provide value to our customers in the submittal and drawing review processes. There are so many details to account for in both of these review processes, and making a mistake can easily lead to procurement delays, acceleration costs, or rework. AI has advanced so quickly and can automate most of these processes — and complete them much faster — and allow humans to do the final checks.
How does Trunk Tools turn fragmented project documentation into actionable insights for general contractors on a day-to-day basis?
We’ve spent a lot of time and cash building what I call the “brain” of construction. Essentially, we’re leveraging the latest and greatest advancements in AI but training models specifically on construction data and processes. In the process, we’re building a knowledge graph that connects the fragmented and unstructured data buried in documents (sometimes millions of pages per project) and helps make explicit the implicit relationships between different documents. For instance, everyone in construction knows that submittals are meant to meet the requirements of certain specifications, and that RFIs might impact that relationship. By training our submittal agent to understand those connections, we can automatically run a submittal review in two-to-three minutes with a very high level of accuracy, often catching discrepancies that humans miss because there is simply too much information to process. Equipped with this compliance recommendation from our agent, our customers can avoid multiple approval cycles with their architects, run submittal reviews on time so as to avoid a backlog, and eliminate both rework and acceleration costs that happen when late or incorrect submittal reviews turn into delays on a project.
Many companies are cautious about adopting enterprise AI. What helped you earn trust from traditionally conservative construction teams?
Even though everyone knows the data about construction being the second-slowest industry to adopt technology, what we’re seeing with our customers is that the top firms have been pretty quick to pilot AI solutions and are making AI adoption a core part of their growth strategy.
With the massive influx of supposedly AI startups in our construction vertical, we’ve built trust by listening very closely to our customers and delivering truly agentic solutions to augment their teams. We’re on multiple project sites across the US weekly helping users learn and adopt Trunk Tools. Plus, the people in the field can sniff out marketing fluff pretty quickly. Our standards are high because our customers’ standards are high, so we’ve built trust by delivering meaningful business impact and staying very close to them.
Can you share a concrete example of how Trunk Tools flagged an issue early enough to prevent a major delay or safety risk?
I’ll share two! On one project, a user query uncovered that a particular door required power actuated hardware. However, the TrunkText agent noted that there was not a callout for power with a panel designation at the opening in the electrical drawings. By flagging this discrepancy before close-in, the customer saved more than $17,000 in rework, finish repairs, inspections, and delays.
On a different project, a user asked whether a Stellar fireplace required sealing of the vent pipe and elbow joints. The agent was able to find this requirement in the corresponding specifications, submittals, and installation manuals within 30 seconds. Catching this detail before drywall installation saved our customer around $100,000 in labor, materials, and delays.
How does connecting project data directly to schedules change decision-making on the jobsite?
Connecting project data directly to the schedule turns it from a static plan into a real-time decision-making tool. When teams can see how RFIs, submittals, and drawing revisions impact activities and the critical path, they can address risks earlier instead of reacting after delays occur. That visibility leads to better coordination, fewer surprises, and more confident decisions on the jobsite.
Your career spans hands-on carpentry to data science and a PhD. How has that combination shaped the way you design AI for construction?
I think my 15-plus years of experience as a builder myself helps me easily distinguish between what is marketing fluff and what is a real solution that will help the people in the field. One of our core values at Trunk Tools is “No BS,” and I hold our team to that standard when we’re building solutions for the field. Most of our team comes from the field or from other ConTech firms, so we understand our customers’ needs quite well.
My PhD gave me a robust analytical framework to understand where the technology is and where it could be in five-to-ten years, which helps inform our product roadmap.
As one of the few women CEOs in construction, what leadership perspective do you think the industry is missing more broadly?
I actually don’t think the fact that I’m a female CEO in construction matters all that much. What matters more is that our industry doesn’t have enough leaders who have deep knowledge of the industry, AI, and human behavior. To build technology that will meet the demands of the future, leaders in our industry need to account for all three of these.
What misconceptions do you see about AI in construction, and what should executives better understand before investing in it?
Most technology firms have been slapping “AI” and “agent” labels all over their marketing for several years now, even when the product has barely changed. Additionally, many who have adopted AI — whether legacy businesses with code bases not built for the processing demands of AI, or new startups trying to capitalize on the AI economy — are merely white-labelling foundational models from OpenAI, Anthropic, Google, and others. These foundational models are not trained on construction data for construction-specific workflows.
I always advise our customers and prospects to test truly agentic, industry-specific solutions like our submittal and drawing revision agents against “custom agent builders” from other potential vendors and evaluate what will actually help them accomplish their technology innovation goals. Do you want your teams to spend time building agents that may or may not work? Or would you prefer ready-to-deploy solutions that make an immediate impact?
Integrating AI into legacy systems is slow, complex, and often constrained by architecture that wasn’t built for real-time intelligence. AI moves fast. Construction teams can’t wait. Legacy systems weren’t designed for the intensive, real-time data processing required for real AI. Fusing the two has been known to lead to poor data quality, degraded performance, and technical debt.
Finally, the composition of the team building these solutions matters. Pay attention to the depth and breadth of experience and education of the AI/ML engineers at every potential vendor you evaluate. A bigger company might not necessarily have a heavier-hitting AI/ML team.
Looking ahead, how do you see AI changing the relationship between the physical jobsite and digital project management over the next five years?
Given the conflicting factors of high demand for new construction (data centers, housing, infrastructure) and low supply of skilled white- and blue-collar construction workers, AI agents that can handle the most time-intensive, paperwork-heavy, bureaucratic processes will be a critical resource to augment project teams, making them more productive, faster, and more accurate. I am not at all concerned about AI replacing workers. Everyone who works in construction knows that we need more talented people, not less. However, the efficiency and cost savings that truly agentic AI offers will make construction companies more profitable and help address the labor shortage, at least in part.