Your BMS Isn’t Enough: Why AI Is Becoming the Building Operating System

Most landlords say they’re “using AI”—very few are actually letting it run the building.

The gap between AI hype and real building performance

If you ask corporate real estate leaders, AI is already everywhere. JLL’s latest global technology survey finds that 92% of companies are piloting AI in corporate real estate, up from just 5% in 2023. Yet only about 5% say they’ve met all their AI objectives, and most are still stuck in experiment mode. Energy management stands out as one of the highest-ROI use cases in that report—but it’s also where the gap between promise and practice is most obvious. Many portfolios still rely on manual setpoints and time clocks underneath “smart” dashboards.

The U.S. Department of Energy has been warning about this for years. A landmark DOE study on commercial building controls found that advanced control measures—like optimized scheduling, sensor-based ventilation and supply-air temperature resets—could deliver average energy savings on the order of 29% in retrofitted buildings. Yet as of the early 2010s, roughly 60% of U.S. commercial floor space had no building automation system at all, and most existing controls were “hand-crafted” rules that were hard to maintain.

Fast-forward to late 2025, and ACEEE’s new series on smart building systems underscores that the basic story hasn’t changed: building energy management and control systems can cut waste for owners and occupants, and AI is starting to make those systems smarter—but uptake and execution still lag the need. In short: lots of pilots and dashboards, not nearly enough buildings where software is actually allowed to drive equipment, tariffs and comfort in real time.

From dashboards to autonomous operations

Market analysts see the same shift. Verdantix’s “Smart Innovators: Energy Management Software 2025” report, which compares 47 platforms, says buyers are increasingly demanding real-time optimization, automation and grid interactivity, not just energy charts and alerts. Suppliers are responding by pitching AI as an operating layer, not a bolt-on feature:

• Schneider Electric just unveiled EcoStruxure Foresight Operation, a unified AI-powered platform for building operations that’s expected to enter wider release in 2026. FacilitiesDive reports that it’s designed to bring together metering, building management, power quality and analytics, with AI agents prioritizing events and recommending or executing actions.
  

• In a recent interview, Schneider’s global building team described how its Energy Control Center cut energy use by 25 GWh and saved about €3 million across 23 Capgemini campuses in India, while helping the sites transition to 100% renewable electricity—essentially operating as an AI-assisted “mission control” across multiple buildings.
  

• Schneider and others claim AI-driven control can reduce building operating costs by up to 40% in some portfolios by dynamically adjusting schedules, setpoints and load sharing based on occupancy and weather.

Smaller pure-play vendors echo the theme. BrainBox AI, for example, markets an overlay that uses machine learning on existing BAS data to continuously tune HVAC, reporting 15–25% reductions in energy use in many deployments without changing core hardware. The direction of travel is clear: from periodic commissioning plus dashboards to 24/7 autonomous optimization.

What “AI as the building OS” actually looks like

From a building owner’s perspective, “AI” isn’t a single product; it’s an orchestration layer that sits on top of:

• Your HVAC, lighting and plug loads

• Onsite assets like solar, batteries and EV chargers

• Occupancy, access control and wireless networks that tell you who’s in the building, when and where

Recent research in Nature Communications estimates that AI applied across design, operations and retrofit decision-making could cut building-sector energy consumption and carbon emissions by 8–19% by 2050, primarily by improving controls, occupancy-based operation and equipment selection. That picture lines up with what forward-leaning owners are already doing:

• Using in-building connectivity (Wi-Fi, private 5G, DAS) as the nervous system that feeds occupancy and sensor data back to central platforms.

• Routing every work order, meter reading, BAS alarm and utility tariff update into a common data model.

• Letting AI agents handle the “pattern recognition and recall”—from spotting inefficiencies to forecasting demand charges—while humans focus on exceptions, capital planning and tenant negotiations.

In other words, moving from “smart building” as a collection of projects to “intelligent building” as an operating system, where energy, comfort and resilience are managed holistically.

Why most portfolios still struggle

If the tools are so promising, why are only 5% of firms hitting their AI goals? JLL’s survey and independent coverage point to a few recurring problems:

• Siloed pilots. Many organizations run AI experiments in energy, leasing and facilities separately, without a shared data backbone or governance.
  

• Data quality and access. Legacy BMS, submetering gaps and fragmented utility data make it hard to train and trust models. DOE’s earlier work on controls noted that many buildings lack even basic sensor coverage.
  

• People and process. ACEEE’s briefs emphasize that smart systems only deliver savings when operators are trained, empowered and resourced to act on insights—or to let the system act on their behalf.
  

For landlords, especially in office, healthcare and campus portfolios, that means the “AI question” is now as much about org design and contracts as it is about tech: Who owns the data? Who signs off on autonomous control? How are savings and risk shared between owners, operators and tenants?

Questions owners should ask now

As you look at proposals from BMS vendors, proptech startups and AI platforms, a few diagnostic questions help separate signal from noise:

1. Control, not just insight - Can this system actually control HVAC, lighting, storage and EV chargers—or is it just another dashboard and alert engine?

   
   

2. Tariff- and carbon-aware - Does it optimize against real tariffs and demand charges, and can it incorporate carbon intensity signals or DR programs from your utility or ISO?

   
   

3. Works with what we have - How does it integrate with your existing BAS, meters, access control and networks? What are the data prerequisites and retrofit costs?

   
   

4. Proven at portfolio scale - Ask for case studies where the vendor has deployed across multiple buildings or campuses, with quantified savings and lessons learned—not just single-site pilots.
   

5. Governance and override - How are safety, comfort and overrides handled? What’s the escalation path when the AI is wrong—or when a tenant’s needs trump an optimization?

The building sector is still early in its AI journey. But the contours of the future are coming into focus: energy, comfort, resilience and even tenant experience will increasingly be orchestrated by software that spans systems, not by isolated projects. For owners and property managers, the opportunity is straightforward: treat AI not as a gadget, but as the operating system for your assets—and build the data, contracts and teams to make it real.