Much of today’s AI conversation focuses on chatbots, copilots, and productivity tools. Yet one of the largest untapped AI opportunities may be hiding inside construction and infrastructure projects. Every year, project leaders lose billions of dollars in enterprise value because they can’t quickly access, connect, and interpret information scattered across contracts, schedules, emails, drawings, change orders, and project correspondence.
John Braley, Head of U.S. Operations at Lupa Technology, believes the next wave of AI adoption will focus less on generating content and more on helping organizations make better decisions from existing data.
Lupa is part of a growing category of companies building what many are beginning to describe as a “construction intelligence layer,” technology designed to transform fragmented project records into actionable commercial, legal, operational, and risk intelligence.
The Problem Isn’t a Data Shortage
Ask Braley what’s holding the industry back, and he doesn’t point to a lack of tools or a lack of data. He points to something else entirely. “Construction does not have a data shortage. It has an intelligence shortage,” he says.
The information a project needs almost always exists somewhere. The trouble is that it’s split across different systems, teams, formats, and versions. One person holds the contract. Another has the specifications. A third has the meeting notes that explain what actually happened. “A single project event can therefore have five different versions of the truth, if not more,” Braley explains.
That fragmentation isn’t just an inconvenience. It’s expensive. Notices get missed. Changes get assessed with incomplete information. Weak positions harden before anyone notices they’re weak. “By the time the complete record is reconstructed, the organization is often no longer managing a project issue; it is managing a dispute,” Braley says.
From Storage to Strategy
Braley draws a sharp line between systems that store information and systems that help people use it. “Most project technology is designed to store information. Storage is necessary, but storage does not provide insight,” he says.
Lupa’s approach is to let teams interrogate the project record directly. Instead of asking where a document lives, a team can ask what changed, who knew about it, which contractual obligations apply, and what the potential cost exposure looks like. “A document-management system may give you ten files. Lupa explains why those ten files are relevant, how they relate to one another, where the evidence is strong or incomplete, and which decision they apply to,” Braley says.
Importantly, this doesn’t mean handing decisions over to a machine. “It does not replace professional judgment. It ensures that professional judgment is based on the strongest and most complete available evidence,” he adds.
A Real Project, A Clearer Picture
Braley points to a major infrastructure project where a team was evaluating a substantial claim built on a long, technically complex chain of events. The supporting evidence was buried across thousands of documents, including correspondence, site records, schedule updates, and contractual notices. Looked at through a narrow set of documents, the claim looked solid.
Once Lupa reconstructed the full record chronologically, the picture changed. “Once the full evidence was visible, the position became much clearer. The team could identify where contemporaneous facts supported the claim, where they contradicted the developing narrative, and where critical evidential links were missing,” Braley says.
He sums up the real value: “Sometimes the most commercially valuable answer is not that your claim is stronger than expected; it is discovering early that it is weaker.”
The Skills That Matter Next
As AI takes on more of the heavy lifting around information retrieval, Braley expects human judgment to matter more, not less. “The strongest professionals will know how to ask Lupa the right questions, challenge its outputs, identify weak or missing evidence, and translate the resulting answers into action,” he says.
He also has straightforward advice for organizations wondering where to start. Skip the abstract AI strategy. “Do not begin with an ‘AI strategy.’ Begin with an expensive problem,” Braley says. That could be a variation review, a claims assessment, or a contractual notice analysis, anywhere teams are currently burning hours searching, reconciling, and reconstructing information by hand.
The Next Five Years
Braley expects AI to stop being treated as a side initiative within five years and instead become part of how construction and infrastructure organizations operate day to day. The bigger shift, in his view, is moving from reconstruction to intervention, catching risk and inconsistency while a project is still live rather than months later.
“The laggards will continue to own terabytes of project data and still be unable to answer the most basic questions when it matters: what actually happened?” Braley says.
For an industry sitting on decades of unstructured project history, that question may define who leads the next chapter of construction technology.
