In a new piece for The AI Journal, OpenSpace CEO and co-founder Jeevan Kalanithi argues that today’s most capable AI systems still cannot understand what is happening in the physical world. As agentic AI takes on more responsibility, that gap matters: agents are only as good as the data they can access, and the most valuable data is reality data, not just documents. Visual intelligence, Kalanithi writes, is the layer that will let agents move from reasoning about intent to grounding decisions in evidence.
The premise behind Kalanithi’s article reflects what we see every day with OpenSpace customers. Plans and documents describe what should happen on a jobsite. They do not capture what actually got built, in what condition, and whether it matches the spec. That gap between intent and reality is where construction loses time, money, and certainty. For AI to be genuinely useful in the built environment, it has to be grounded in what is happening in the field, not just what was reported afterward. Reality data, captured continuously through smartphones, 360° cameras, and drones, gives teams and the AI systems they rely on a shared source of truth tied to specific locations and points in time.
As agentic systems begin to plan and act across construction workflows, the question is no longer whether AI can reason. It is whether AI can verify. An agent that recommends a next step based on a schedule, but cannot see the site, will always be limited to inference. Visual intelligence closes that loop. It allows agents to check progress, validate quality, and flag risk against the actual state of the work. This is why we built OpenSpace as a Visual Intelligence Platform for builders, with 380,000 users across 130 countries capturing reality data on real projects. The next wave of agentic AI in construction will run on that foundation. Learn more about the Visual Intelligence Platform.
