Y Combinator’s Winter 2026 batch was, by almost every measure, its most AI-dense in history. Of the roughly 190 companies that presented at Demo Day on March 24, 2026, the overwhelming majority were building on top of AI — a proportion that reflects the near-total shift in early-stage software development toward AI-native architectures rather than any particular change in what problems founders are choosing to solve.
Five of those companies were building for construction, trades, and property technology.
That is a small number out of roughly 190. It is also a meaningful one. Construction represents roughly 13 percent of global GDP and remains one of the most technologically underleveraged industries in the world. The fact that YC — which runs on pattern recognition about where durable software businesses can be built — is producing construction-focused companies at all is a signal worth paying attention to.
What the W26 construction batch reveals, looked at together, is a specific thesis about where AI can enter the construction value chain right now: the document-heavy, rule-bound, information-asymmetric workflows that are slow not because the work is inherently complex but because human capacity has always been the bottleneck. Permitting. Estimating. Cost intelligence. Field-to-office operations.
Here is every one of them.
Bidflow — AI Electrical Takeoff at Three Cents Per Device
Website: usebidflow.com · Category: Specialty Trade Estimating
Bidflow does one thing: it counts electrical devices in construction drawings. Every receptacle, every junction box, every light fixture, every occupancy sensor — automatically, from a PDF upload, with 95 to 99 percent accuracy, in under ten minutes.
The pricing model is as noteworthy as the technology: $0.03 per correctly identified device, pay-as-you-go, no subscription. That structure eliminates the barrier to trial (run a small project, spend a few dollars), aligns cost to value (large complex projects pay more, but those are also the projects with the highest estimating stakes), and embeds a performance guarantee — you only pay for devices the AI correctly identifies.
The founders do not have backgrounds in electrical contracting, which is worth acknowledging. What they have instead is execution proof: Jesse Choe was already a YC founder (S25, Ghostship, QA testing AI) who reached $200,000 in annual recurring revenue before pivoting to Bidflow. Gautham Ramachandran, his co-founder, is a UC Berkeley computer scientist with published work on multi-agent NLP systems. The Ghostship ARR is the signal that matters most here — it demonstrates that Choe can build an AI product and sell it to businesses, regardless of domain.
Bidflow’s current scope — device counting only, not full electrical estimating — is honest about what AI can do reliably in this domain right now. The harder parts of electrical estimating (conduit routing, panel schedules, labour complexity) require contextual reasoning that current models do not perform well enough for commercial use. Owning one high-accuracy slice of the workflow is the right product strategy at this stage.
→ Read our full analysis of Bidflow
AutoSitu — AI Plan Review for Municipalities
Website: autositu.com · Category: Permitting & Compliance
AutoSitu has built an AI plan review system that checks architectural and engineering documents against local zoning, fire, and building codes — automating one of the most resource-constrained steps in the development process.
AutoSitu’s technical architecture is built around three components: a Precedent Graph that encodes local zoning and building codes in a structured, machine-queryable form; a Document Intelligence layer that extracts structured data from architectural drawings; and a Multi-Agent Engine that routes review tasks to specialised AI reviewers and escalates edge cases to humans. The human-in-the-loop design is deliberate — AutoSitu is positioned as AI-assisted plan review, not autonomous approval.
The platform is explicitly two-sided: it serves municipalities (faster reviews, reduced staffing pressure) and developers and architects (pre-submission compliance checks that catch rejections before they happen). Both sides of the permit transaction have clear incentives to use it, and the policy environment — cities across the US under pressure to streamline permitting for housing production — creates a tailwind that most enterprise software companies would envy.
Named clients include the city of Sterling Heights, Michigan, architecture firms Dahlin and SWA, contractor Shafer Construction, and developer Crow Holdings. Investors: YC, Detroit Venture Partners, Pioneer Fund. 10,000+ plans reviewed.
Foreman — Construction Management Software Built for the Contractors Procore Ignores
Website: foreman.co · Category: Construction Management (SMB)
Foreman is attacking a market that Procore has never meaningfully served: the roughly 760,000 US construction businesses with fewer than 20 employees. These contractors — residential remodelers, home builders, roofers, small commercial builders — currently manage their projects on spreadsheets, text threads, and QuickBooks. Not because they are unsophisticated, but because the software built for them has not been worth changing how they work.
Foreman is an all-in-one platform: AI-powered takeoffs from plan uploads, one-click proposal generation, e-signature contracts, critical path scheduling, invoicing with QuickBooks integration, and centralised document management. The AI layer is designed into the product from the start rather than retrofitted onto an existing architecture — a meaningful advantage over incumbents like Buildertrend (founded 2006) and CoConstruct in a market where the daily-use workflows are exactly the ones AI can accelerate most visibly.
The founder, Nolan Rossi, is the right person to build this. He is a fourth-generation member of a construction family who triple-majored in Electrical Engineering and Computer Science, Astrophysics, and Business Administration through UC Berkeley’s M.E.T. program — finishing in three years. He worked as a software development engineer at Amazon and ran a software consultancy to $250,000 in revenue before founding Foreman. The combination of lived construction domain knowledge and serious technical capability is rare enough to be worth noting.
At Demo Day, Foreman was a two-person San Francisco company. The market they are targeting has roughly 800,000 potential customers.
→ Read our full analysis of Foreman
Travo — Commercial Real Estate Data Intelligence
Website: travoai.com · Category: Commercial Real Estate
Travo is building AI-powered commercial real estate market intelligence — aggregating CRE data and surfacing it through an AI interface that gives operators and investors access to market insights faster than traditional research processes allow. The company’s positioning, “real estate data in seconds,” targets the research and underwriting workflows where analysts spend hours pulling comparables, tracking lease expirations, and assembling market reports.
CRE data intelligence is a competitive space — CoStar, CBRE, and a range of AI-powered alternatives including Reonomy and newer entrants have staked out territory here. Travo’s differentiation and go-to-market focus had not been detailed extensively in public materials at the time of writing.
Crow — AI for Commercial Real Estate Operators
Website: usecrow.ai · Category: Commercial Real Estate Operations
Crow describes itself as offering “AI sprint services for CRE operators” — providing AI-powered workflow automation for commercial real estate property teams. The company’s focus is the operational layer of CRE management: the recurring tasks that property teams run across lease administration, tenant communication, reporting, and asset management.
Crow’s positioning as “AI sprint services” — discrete AI-powered service engagements rather than a traditional SaaS subscription — is an interesting structural choice. It implies a model closer to a high-leverage service firm than a pure software platform, with the AI deployed in scoped projects rather than as a continuous subscription. At the time of writing, Crow’s detailed product architecture and go-to-market approach had not been widely published.
What the W26 Construction Batch Tells Us
Five companies across roughly 190 is a small share for construction and property tech in what YC describes as a historically AI-dense batch. That number understates the significance of what these companies are building.
Four patterns are visible when you look at the batch together.
The bottleneck thesis is converging. Most of these companies are attacking workflows where human capacity has always been the limiting factor: plan review (AutoSitu) and estimating (Bidflow, Foreman). None of them are building project management software to compete with Procore in its core market. All of them are building tools to make the information-gathering, document-processing, and analysis workflows that precede and inform construction decisions faster and more accurate.
The data moat is the point. AutoSitu’s Precedent Graph of local building codes, Bidflow’s trained models on electrical drawing symbol libraries, Foreman’s AI trained on SMB contractor workflows — each of these represents a dataset that gets harder to replicate as the company grows. This is the structural pattern of durable construction AI businesses: the product improves as usage grows, creating a feedback loop that eventual competitors cannot shortcut.
Founder-market fit is unusually strong. Nolan Rossi grew up in a construction business and watched his family’s operational problems from the inside. Jesse Choe has already demonstrated he can build and sell AI software to businesses. This is a batch where the founders have earned the right to their thesis through experience rather than research.
The permitting opportunity is larger than it looks. AutoSitu’s municipal focus positions it in a market where the political, economic, and operational case for AI assistance is perhaps stronger than in any other part of the construction stack. The construction industry’s housing delivery problem is, in significant part, a permitting speed problem. A company that can demonstrably accelerate municipal plan review at scale is solving a problem that elected officials, developers, and planners have all decided they want solved.
The construction industry’s technology adoption lag is well documented. The W26 batch is a set of founders who have decided that the lag is ending, and that the right moment to build for it is now.
