Trunk Tools, an AI platform automating construction workflows, has closed a $40 million Series B round led by Insight Partners, bringing total funding to $70 million. The startup was founded by Dr. Sarah Buchner, who began her career as a carpenter at age 12 in Austria, worked her way up to superintendence and contracting, witnessed job-site inefficiencies and fatalities, and returned to build construction software. She holds a PhD in civil engineering and data science.
The platform uses custom LLMs trained on construction data (drawings, plans, RFIs, schedules, submittals) to answer natural-language queries on job sites and automate workflows like document review, conflict detection, and site communication. A user asks 'Does this door require electricity?' and the system returns sourced answers from project documents. Autonomous agents now handle scheduling, project tracking, and submittal-spec comparisons. Customers include Suffolk Construction, Gilbane, and DPR Construction. The 60-person company operates across the U.S. and Canada.
The Series B includes participation from Redpoint Ventures, Innovation Endeavors, Stepstone, Liberty Mutual Strategic Ventures, and Prudence. When Trunk Tools released its natural-language documentation assistant in 2023, it was among the first to bring conversational AI to job sites and was quickly adopted by major contractors. Buchner has modeled the deployment of AI specialists into customer organizations after Palantir's forward-deployed engineering approach, managing behavioral adoption from within.
For architects: The $13 trillion construction industry has historically resisted digitization, making it ripe for vertical AI. The key insight: general-purpose LLMs fail on construction data because of domain-specific vocabularies, siloed document systems, and complex interdependencies. Custom models trained on proprietary drawings and workflow data unlock value that off-the-shelf models cannot. This pattern—deep domain expertise + AI agents + outcome-based pricing (some agents are billed per business outcome)—mirrors patterns in legal AI (Legora) and other enterprise verticals where unstructured legacy data is the moat.