Value Lives Upstream of the Jobsite

Higharc raised a $95 million Series C led by Insight Partners, with Suffolk Technologies and MetaProp among the participants. The line worth noting is how the company describes what it builds: Higharc generates homes as spatial databases — capturing code requirements, construction standards, and geometry — to create the structured data foundation that production-grade AI requires. The money didn't follow a prettier rendering. It followed structured data that a machine can reason about.

That round doesn't sit alone. In a recent post, Autodesk CEO Andrew Anagnost described "project intelligence" as a connected brain that preserves context across planning, design, construction, and operations so knowledge doesn't disappear at every handoff. As Engineering News-Record reported, Buildots launched a research lab that turns anonymized project data into industry benchmarks. And one analyst who mapped a hundred funded construction-AI startups counted thirty-seven with Y Combinator backing, reading the portfolio as a bet on preconstruction, the legal layer, and the financial operations of running a firm — not on the jobsite. Suffolk, which spent a decade and more than $100 million building a data foundation of roughly 293 terabytes before scaling AI on its jobsites, put the sequencing plainly: data first, everything else follows.

Zoom out and the shape becomes clearer. The visible story in construction AI is still robots and cameras on site. The investment thesis underneath it is that the durable advantage comes from proprietary, structured data and the preconstruction decisions built on top of it. That's not a niche concern: more than half of contractors in BuiltWorlds' latest AI/ML survey named limited data availability or quality as one of the biggest hurdles to adopting AI. A 2025 Stanford Graduate School of Business case study on Autodesk frames the same logic from the incumbent's side — its CEO betting that industry-specific foundation models and a defensible full-lifecycle platform are what hold up as AI-native startups arrive. Even the counterexamples reinforce the point. Any operator who has lost a day of work to a cloud sync outage on a hosted design platform knows the risk — a reminder that centralizing everything is only as strong as the data foundation beneath it.

The place this bites first for most operators is bidding. A recent Procore-sponsored piece described preconstruction bidding as disconnected systems, unclear scopes, and inconsistent trade-partner engagement — the exact conditions that turn into rework and cost uncertainty later. Tools exist now to close that gap. Platforms are being built specifically to connect RFIs, submittals, and documents across systems that don't talk to each other. Bid leveling — normalizing scopes across trade partners so you're comparing the same work — is one of the clearest cases where structured data upstream changes the decision downstream.

If you run a construction firm, the move this quarter is to treat preconstruction data as an asset you own rather than exhaust you discard. Three places to start. First, structure your bid and scope data so it's comparable across projects, not trapped in one estimator's spreadsheet. Second, standardize how scopes get defined before they go out, so the leveling later is apples to apples. Third, when you evaluate an AI tool, ask what data it needs to be useful and whether that data is yours to keep — because the firms pulling ahead aren't the ones with the newest software. They're the ones who own the structured record the software runs on.