Buildings as a Lever for Productivity

This article is part 8 of our 12-part series “Where Corporates Can Achieve Productivity Gains,” in which we systematically examine the areas of business where data-driven and AI-supported approaches can lead to demonstrable productivity gains.

‍

In most areas of a business, the biggest hurdle to demonstrating an efficiency gain is the lack of a benchmark: how much would a process have cost without the new technology? With buildings, this question rarely arises. Energy consumption, the volume of service calls for building services, and staff costs for cleaning and security are already recorded before an AI system is even deployed. This makes facility management one of the most directly verifiable productivity levers in our series—with an impact that is already evident in the next accounting period, rather than only within a measurement framework that still needs to be devised.

‍

Buildings account for an estimated 40 percent of global energy consumption. Within a building, the largest single item is typically heating, ventilation, and air conditioning (HVAC): around 45 percent of a commercial building’s annual energy consumption. Historically, an estimated 30 percent of this was lost due to inefficient control systems—such as a rigid thermostat timer running at full power at night in an empty office or a cooling system that never adapts to actual occupancy levels. This combination of scale, systematic waste, and an existing, clear baseline—the energy bill—makes AI-powered building management one of the most consistently proven productivity levers in this series.

‍

Why This Area Is a Measurable Productivity Lever

‍

The mechanism is simple to describe: a fixed timing schedule is set once by a person and rarely reviewed. It does not respond to actual occupancy or weather conditions. An AI system, on the other hand, continuously monitors occupancy, weather data, and thermal load forecasts and adjusts the controls in real time—not even daily, but every few minutes. Because the waste that has occurred to date is systematic and predictable, rather than randomly scattered, this results in a structural advantage for a system that continuously readjusts.

‍

The following diagram illustrates the principle: whilst a fixed timing system operates independently of actual usage, an AI-controlled adjustment follows the real occupancy curve throughout the day—the area in between corresponds to the energy waste that has been avoided.

‍

‍

‍

This structural logic explains why the energy savings documented in practice—typically between 25 and 43 percent, depending on the baseline—rank among the consistently largest productivity gains that have been demonstrated in this series to date.

‍

The Toolbox: Four Technological Building Blocks

‍

AI-supported HVAC control attracts the most public attention because the energy savings are the easiest to quantify. However, the actual productivity gains in this area stem from a broader toolkit that encompasses climate control, physical labor, space utilization, and maintenance:

‍

• Monitoring a building’s vital functions (AI building management systems): Deep learning models adjust HVAC control in real time based on occupancy, weather forecasts, and projected thermal loads—instead of a rigid, preset timer, the system optimizes itself every few minutes.
• Covering more floor space per shift (autonomous cleaning and security robots): AI-controlled robots clean or patrol larger, repetitive areas—such as warehouses, airports, or retail spaces—more consistently and at a lower overall cost than comparable staff.
• Measuring actual space utilization (occupancy analytics): Sensor networks and computer vision systems track which workstations and meeting rooms are actually in use—providing a basis for decision-making on the appropriate size of a property portfolio, rather than relying on assumptions or outdated floor plans.
• Detecting faults before failure (predictive building maintenance): The same physics-based methods used in industrial maintenance can be applied to HVAC compressors, lifts, and other building infrastructure to detect vibration or temperature signatures that precede a failure before a specialist would notice them.

‍

What This Means in Figures—Three Case Studies

‍

The documented results come from different geographical contexts and are in each case substantiated by company data or independent certification—not by an unsubstantiated vendor estimate.

‍

‍

• Schneider Electric – Le Vaudreuil, France: At its smart factory site, Schneider Electric achieved a documented 25 percent reduction in energy consumption and a 17 percent reduction in material waste through its AI and IIoT integration. What is noteworthy here is the company's dual role as both a user and a provider of the same technology: having tested the approach at its site, Schneider Electric now markets the underlying energy management and digital buildings capabilities to industrial and commercial property clients.
• Meera Tower, Dubai: Using BrainBox AI’s deep-learning HVAC optimization—protected by a granted US patent—HVAC energy consumption was reduced by 42.7 percent within four months, one of the strongest documented results for a single building in this field. BrainBox AI now controls the HVAC systems in more than 4,000 buildings worldwide, ranging from small retail premises to airports.
• Treptower Berlin: Property manager Patrizia AG collaborated with DABBEL AI to optimize the HVAC control system at the Treptower Berlin office complex. The system collects real-time data from sensors, meters, occupancy signals, and weather forecasts, and autonomously adjusts the HVAC settings. The result—certified under the Solar Impulse Foundation's Efficient Solutions Programme—was a 36 percent reduction in HVAC energy consumption, equivalent to 1.76 million kilowatt-hours and around 605 tonnes of CO₂ annually.
• ‍

These three results relate to different buildings, baseline conditions, and measurement periods and are therefore not directly comparable with one another; however, they illustrate a consistent order of magnitude of the effect.

‍

How the Provider Landscape Is Consolidating

‍

For investors, this sector has already produced two clear examples of consolidation: BrainBox AI was acquired by Trane Technologies following a US$103 million funding round in January 2025. Nantum AI was acquired by Johnson Controls in April 2026 to strengthen its own OpenBlue ecosystem with AI algorithms, which the company preferred to acquire rather than develop in-house. This pattern—an established provider acquiring tried-and-tested AI technology rather than building it in-house—is consistent with observations from other sectors covered in this series, such as predictive maintenance and demand planning.

‍

By contrast, two technology fields show comparatively little independent venture activity: autonomous cleaning and security robotics, where profitability tends to stem from bundled facility service contracts rather than standalone AI licenses, and predictive building maintenance, which is increasingly offered as a feature within existing building management and IoT sensor platforms. Occupancy analytics and building data infrastructure, on the other hand, remain open: providers such as VergeSense, Infogrid, and Metrikus are pursuing different approaches here—sensor-based occupancy analysis, unified building data integration, and portfolio-wide analytics, respectively—without any single provider having clearly established itself as the market leader to date.

‍

Context for Institutional Investors

‍

For Averdas, this area is particularly relevant in the context of resource productivity: how effectively a company converts energy and other resources into economic output. It also touches on the topic of process productivity, as the shift from rigid to continuous control is ultimately an improvement in operational processes. The individual case results described here are historical results—certified either by the companies themselves or by independent bodies—relating to individual buildings and projects; they cannot be directly extrapolated to a company as a whole, an index, or an investment product. Furthermore, historical results are not a reliable indicator of future performance.

‍

When assessing companies within the Averdas approach, it is relevant to consider the extent to which a company has already realized or is addressing structural efficiency potential within its property and energy sectors—as one component among several within the four Averdas productivity factors, not as the sole selection criterion.

‍

Conclusion and Outlook

‍

Buildings present a rare case where productivity gains can be directly measured against an existing key performance indicator: the energy bill. The documented case studies show savings of between 25 and 43 percent with a comparatively short time to impact, and consolidation among established HVAC and building services providers suggests that the core technology is becoming a standard component of building management platforms.

‍

This article is part 8 of our 12-part series on productivity levers for businesses. The other parts of the series highlight additional areas in which businesses can unlock structural efficiency potential through data-driven approaches—an overview of all 12 areas can be found in the series overview.

‍

‍

‍