Featured Product: AI Agents move into the building: Infrastructure AI bets that commercial real estate is ready to run itself

Somerset, NJ The commercial real estate industry has spent years talking about smart buildings. A New Jersey company now says it is ready to deliver one that effectively manages itself, and it is aiming that pitch squarely at the owners and operators who control the country’s office and commercial portfolios.

The case for change starts with waste. According to the U.S. Department of Energy, roughly 30% of the energy used in commercial buildings is wasted, and commercial buildings account for $190 billion in annual energy expenditures. U.S. Energy Information Administration data shows heating and cooling drive much of that load, with space heating alone making up about 32% of energy use across U.S. commercial buildings. For an asset class where net operating income drives valuation, that wasted spend is margin lost every month.

For New York owners, the stakes are sharper still. Under Local Law 97, roughly 50,000 buildings over 25,000 s/f now face annual carbon caps, and nearly 70% of the city’s carbon emissions come from the energy used to heat, cool, and power its buildings. Owners who exceed their limit face a penalty of $268 for every metric ton of carbon dioxide equivalent over the cap. The pressure intensifies at the end of the decade, when limits for most building types drop by roughly half in the 2030 compliance period, pushing tens of thousands of currently compliant buildings toward fines. That is the backdrop against which any new operating technology will be judged.

Infrastructure AI, headquartered in Somerset, has launched its Galaxy Agentic Operating System, branded GAOS, a platform designed to attack both the waste and the carbon exposure by handing day to day building operations over to artificial intelligence. The system deploys thousands of AI agents that monitor building systems continuously, predict inefficiencies before they compound, and execute corrective actions on their own. The company estimates the platform can cut total operating costs by 35 to 40%.

For property teams, the practical question is whether a new platform can work with the equipment already installed. GAOS includes a data normalization engine that translates information from systems built by Siemens, Honeywell, Johnson Controls, and others into a common language, allowing equipment to communicate regardless of protocol. The AI agents then construct a live digital twin of the building and optimize against it. The company offers a simple illustration. If a thunderstorm cools a building unexpectedly at noon, an agent recognizes that the chiller no longer needs to run at full capacity and dials it back, without waiting for an engineer to notice.

That autonomy is where the industry’s skepticism surfaces. Operators have long kept a human in the loop because mechanical systems carry tenant comfort and safety consequences. Co-founder Glen Allmendinger, who has tracked smart systems since founding the research firm Harbor Research in 1983, argues the deeper problem is architectural. He contends that most of today’s smart infrastructure platforms are built on yesterday’s technology, systems that were never designed to govern autonomous agents in the first place. Infrastructure AI’s answer is to treat trust as a design requirement. “AI can only transform infrastructure at scale if customers trust the GAOS platform behind it,” Allmendinger said. “That means designing for security, visibility, and accountability from the start, so owners and operators know their data is protected and every system decision can be understood and verified.”

That principle drives the platform’s most distinctive feature. GAOS includes a blockchain infrastructure ledger that records every operational event, maintenance action, and performance metric for each connected asset, supported by a native cryptocurrency designed for infrastructure transactions. Co-founder Dilip Rahulan describes the broader model as investment grade autonomy, an approach in which agents earn the right to act and every action is provable and auditable, aimed directly at the liability and insurability concerns that slow adoption. For owners managing large portfolios, a permanent and verifiable record of asset performance could carry weight in due diligence and financing. Whether the market embraces the cryptocurrency component is a separate matter the company will need to prove building by building.

The two founders come at the problem from opposite ends. Rahulan is the builder. He delivered the Middle East’s first building to earn LEED Platinum certification in 2007 and later opened what the company describes as the world’s first global command and control center, in New Jersey, for real time energy and asset management. Allmendinger is the forecaster. His 2005 Harvard Business Review article anticipated the rise of connected devices and smart buildings years before the term Internet of Things was in common use, and his firm has published research on every major smart systems wave since.

“Buildings today run on aging, outdated systems that rely on manual oversight and static programming,” Rahulan said. “We built the intelligence layer that allows these systems to think, adapt, and manage themselves.”

The company is entering the market with a freemium model, running live building tests, and seeking strategic investment partners. For commercial real estate, the question is no longer whether buildings can be made smarter. It is whether owners, facing both rising energy costs and tightening carbon law, are prepared to let the building make the call.