We spent two days on the ground in Chicago at BuiltWorlds Construction Tech Conference with design-build leaders, GCs, and technology executives shipping software in construction. Construction is rapidly accelerating, with data-center buildouts as the dominant forcing function. Contractors expect them to outpace every other category in 2026. That capital intensity is changing how firms think about technology and who are no longer just buying, but actively building.

Here is what I heard and where we see the industry moving:

  • Generative design is real, but requires trustworthy downstream integration. Parametric, code-compliant layout generation is advancing quickly and moving from CAD to computer-generated design. The gap forming is reliability in a high-stakes environment where a confident hallucination costs time, capital, and eventually safety. On top of that, a generated floor plan only creates value if it becomes a single source of truth for purchasing, manufacturing, framing, and sequencing. One BIM integration lead put it that design’s value is measured downstream. We’re watching for the layer that connects generated models to execution across procurement, sequencing, and field coordination, which goes beyond just the design tool itself.
  • Subcontractor selection is a judgement problem. Current tooling is optimized for background checks: financial health, safety records, compliance. Those are necessary but not enough. The harder problem is matching a specific subcontractor to a specific job, site, crew, and moment, which is an assessment that lives partly in rubric and partly in experiential judgment that hasn’t been formalized. On top of this, the relevant data is fragmented across systems, email, text messages, and personal devices. Solving sub management requires decision intelligence infrastructure that can synthesize a multitude of structured and unstructured signals into contextual recommendations.
  • Jobsite monitoring is abundant yet timeline intelligence isn't there. Cameras, drones, wearables, smart glasses, exosuits, instrumented hard hats, and quadruped robots scanning hundreds of thousands of square feet are all deployed now. What remains open is the last mile of stitching that data into a coherent, real-time view of schedule risk. A scan tells you what the site looked like at one moment. What comes next is automated updates on whether the project is behind, at risk, or ahead. The opportunity we see is tying observational data back to schedule, cost, and sequencing logic.
  • The build-vs-buy inflection is accelerating, except where proprietary lock-in holds. One technology leader described an internal engineering team using agentic development, running their own GPUs, and replacing legacy platforms (including Procore) with cheaper, more capable in-house tools. The switching costs in the ecosystem have dropped meaningfully. The one exception was Oracle Primavera P6 because owners contractually require schedules in P6’s proprietary format, making it essentially irreplaceable regardless of internal engineering capacity.

Wherever contract or regulatory structure creates a proprietary format lock, legacy vendors retain pricing power. Everywhere else, the barrier to custom software is falling, which threatens incumbent SaaS and opens space for these vertical infrastructure plays.

The opportunity

The demand pulling construction forward, data centers especially, has created more urgency and capital behind these problems than the industry has seen in years. The consistent pattern is that AI can generate outputs, yet construction still runs on fragmented data, unformalized judgment, and proprietary formats. 

There's clearly much opportunity ahead in the built world. We’ve been fortunate to back incredible companies at the forefront of this movement, including Nox Metals and Congrue. We’re always on the lookout for founders with a deep earned perspective in construction tech. If this is you, please reach out!