Article: Why Data—Not Compute—Will Shape Physical AI In Construction

Published: December 22, 2025 | Forbes

Henri Lee, Xpanner CEO, contributed a thought leadership article to Forbes Tech Council on why Physical AI in construction will be defined by data—not compute.

The rapid rise of AI has already reshaped the digital world, from content generation to workflow automation. But the next major leap is happening in the physical world—where AI must interact with real environments by building, moving, and shaping things. In construction, this shift matters deeply as the industry continues to face labor shortages, safety risks, and operational inefficiencies.

In the article, Henri explains how Physical AI can help unlock intelligent, unmanned job sites that improve productivity, reduce cost, and enhance safety. Unlike digital AI applications that can be trained using internet-scale datasets, construction-focused Physical AI requires high-fidelity real-world data captured directly from job sites—including machine behavior, terrain conditions, vibration feedback, and constantly changing environments.

A key message is that data is the foundation of intelligence in Physical AI. While many modern AI breakthroughs are driven by increasing compute power, Physical AI depends on continuous data acquisition to learn how real work is executed on real sites. This creates a feedback loop: as machines operate, they collect richer data, which improves models and results in even better performance and better data over time.

The article also highlights that contractors and construction organizations can begin building this foundation by:

• Instrumenting existing equipment with reliable sensing and telemetry
• Standardizing field data collection with GNSS, sensors, and operational tagging
• Treating data as a job-site asset through governance and repeatable workflows
• Creating cross-project feedback loops so one project continuously improves the next

Ultimately, the future of construction automation will be shaped by the organizations that can collect, structure, and apply real-world jobsite data at scale—turning field execution into a measurable, continuously improving system.