Why AI Needs Eyes: Understanding The Physical World

Jeevan Kalanithi is co-founder and CEO of OpenSpace, a Visual Intelligence Platform helping builders turn site reality into action.

Artificial intelligence has become remarkably good at navigating the digital world. It can read documents, analyze spreadsheets, write code and reason across enormous amounts of information. But much of the economy does not operate in a digital world.

Construction sites, factories, warehouses and energy facilities are constantly changing. Equipment moves, materials arrive, work gets installed, conditions change and much of what happens never makes its way into a database or report. That creates a fundamental limitation for AI, as an agent is only as useful as the data it can access. When some of the most important information exists in the physical world, AI needs a way to perceive that world.

In other words, AI needs eyes.​

Why Agents Need Reality, Not Just Intent

A lot of the excitement around agentic AI focuses on what agents can do: analyze information, generate reports, recommend actions and increasingly execute tasks on behalf of users. But every one of those capabilities depends on input.​

In construction, schedules, contracts, BIM models, drawings and specifications tell us what is supposed to happen. They are incredibly valuable, but they largely describe intent. The physical site tells us what actually happened. Has a wall been installed? Has work progressed according to schedule? Was an issue fixed?​

For an AI system to reason effectively about real-world operations, it needs access to both sides of that equation: the plan and the reality.​

That applies far beyond construction. Manufacturers have production plans. Utilities have engineering models. Logistics companies have routing systems. In every case, digital systems describe an expected state of the world, while physical operations determine the actual state. The more effectively AI can connect the two, the more useful it can become.​​