The Electricity We Waste Because Buildings Cannot Predict an Empty Room

---

Millions of rooms are illuminated, heated, cooled, and ventilated according to schedules that do not always reflect whether anyone is actually there. Smarter building controls could reduce this invisible energy waste but predicting occupancy is more complicated than simply detecting movement.

An office meeting ends early, but the lights remain on. A conference room sits unused while its air-conditioning system continues to maintain a comfortable temperature. Across a university campus, classrooms receive ventilation even when scheduled activities have been cancelled.

These situations reveal a persistent weakness in how many buildings operate: their energy systems are designed to serve expected occupancy, not necessarily actual occupancy.

The problem is not simply that buildings consume too much electricity. It is that they often cannot distinguish between a room that needs energy now, one that will need it shortly, and one that may remain empty for hours.

As buildings become more connected, occupancy-aware controls and predictive software offer an opportunity to reduce this mismatch. Yet the most effective solution is not necessarily a building filled with cameras and artificial intelligence. It may be a carefully designed system that combines basic sensors, reliable scheduling, appropriate automation, and an understanding of how people use space.

Why Empty Rooms Continue Consuming Energy

Most commercial buildings are operated through a combination of fixed schedules, thermostats, lighting controls, and building management systems.

A conventional office might begin heating or cooling before employees arrive, maintain indoor conditions throughout working hours, and switch to reduced operation in the evening.

This approach is practical. Building operators cannot wait until hundreds of employees arrive before making the workplace comfortable.

But schedules have limitations.

A meeting room reserved from 10 a.m. until noon may be empty by 10:45. An employee may work remotely without updating the office booking system. A classroom may be unused because a lecture was cancelled.

Unless the building receives information about those changes, its systems may continue operating as planned.

Three major energy-consuming systems are particularly relevant:

• Lighting: Artificial lighting may remain active in unoccupied rooms, especially where automatic controls are absent or poorly configured.

• Heating and cooling: HVAC equipment may continue maintaining occupied-space temperature settings when a room could safely operate with wider temperature limits.

• Ventilation: Some systems supply outdoor air based on design occupancy or fixed schedules rather than adjusting to actual demand.

The energy consequences differ substantially between buildings. Lighting waste may be relatively easy to eliminate with occupancy sensors, while heating, cooling, and ventilation require more careful control.

Importantly, an empty room does not necessarily mean its energy consumption should fall to zero. Buildings still need protection against moisture, freezing, overheating, and other conditions that can damage equipment or compromise indoor environments.

The opportunity is to reduce unnecessary conditioning without sacrificing safety, comfort, or building performance.

The Difference Between Detecting an Empty Room and Predicting One

A motion sensor can tell a lighting system that movement has not been detected for a certain period.

That is useful, but it is not the same as understanding occupancy.

Someone reading quietly may remain almost motionless. A sensor may miss a person sitting behind furniture. Conversely, movement near a doorway may activate lighting in a room that nobody intends to use.

Predictive occupancy systems attempt to solve a different problem.

Instead of responding only to current conditions, they estimate how a space is likely to be used in the near future.

For example, a building might combine room reservations, historical usage patterns, and sensor observations to estimate when a meeting room is likely to become occupied.

This matters because different building systems respond at different speeds.

Lights can turn on almost immediately. Heating a large room to a comfortable temperature may take considerably longer.

A building that waits until occupants arrive before starting its heating system could save energy but create an uncomfortable experience.

A predictive system can potentially begin conditioning shortly before expected arrival, rather than maintaining the same conditions throughout the day.

The central distinction is straightforward:

Occupancy detection asks whether somebody is present. Occupancy prediction asks when somebody is likely to be present and how the building should prepare.

That second question creates more opportunities for optimization, but also introduces uncertainty.

How Much Electricity Could Smarter Buildings Save?

The potential is meaningful, but it is important not to confuse savings from all advanced building controls with savings specifically attributable to occupancy prediction.

According to the International Energy Agency (IEA), buildings accounted for approximately 30% of global final energy demand in 2024. This includes electricity and other energy sources, such as fuels used for heating.

Evidence from actual buildings also demonstrates the value of simpler occupancy-responsive technologies.

A U.S. General Services Administration evaluation of lighting controls across five federal buildings reported lighting energy savings ranging from 27% in regularly occupied spaces to 63% in a call center with extended operating hours, compared with the evaluated baseline conditions.

The amount saved depends on existing controls, occupancy patterns, equipment efficiency, climate, building layout, and the quality of installation and operation.

A Simple Calculation Shows Why Small Inefficiencies Matter

Consider a hypothetical office building containing 100 meeting rooms.

Assume each room has 120 watts of installed lighting and that the lights remain on unnecessarily for three hours per working day.

The resulting electricity waste would be:

At an illustrative electricity price of $0.20 per kilowatt-hour, that would represent $1,800 in annual electricity costs.

This is not a measured building or an industry average. It is a simplified example showing how repeated small inefficiencies accumulate.

It also excludes the electricity used by HVAC systems, which may represent a larger opportunity in some buildings.

However, the full 9,000 kWh would only be avoidable if all the assumed unnecessary lighting hours could be eliminated. Real-world sensors, operating requirements, and existing controls may reduce the achievable savings.

The Technologies That Help Buildings Understand Occupancy

No single sensor can perfectly determine how a building is being used. Different technologies provide different kinds of information.

The strongest approach is often to combine complementary signals rather than relying on a single source.

For example, a room reservation might indicate that a meeting should begin at 2 p.m. A presence sensor could confirm whether people actually arrive. A building controller could then adjust lighting and temperature according to the observed situation.

If the meeting is cancelled, the system could avoid unnecessary conditioning.

If attendees arrive unexpectedly, the system should respond promptly rather than rigidly following its prediction.

This combination of prediction and real-time correction is important because human schedules are rarely perfect.

Why Ventilation Makes the Problem More Complicated

Lighting is relatively straightforward: an empty room generally does not need the same artificial illumination as an occupied one.

Ventilation is different.

Indoor air quality depends on both the number of occupants and sources of pollutants that may exist even when nobody is present.

Furniture, cleaning products, building materials, and other sources can affect indoor air independently of occupancy.

Demand-controlled ventilation adjusts outdoor airflow in response to occupancy-related needs, often using carbon dioxide measurements or other occupancy estimates.

According to ASHRAE, occupancy-based ventilation can reduce unnecessary outdoor air supply when spaces are used below their design occupancy. However, ventilation requirements include both people-related and building-related components.

Indoor carbon dioxide concentrations depend on factors including occupancy, activity, outdoor concentrations, air mixing, and ventilation rates. A single CO₂ threshold is not a universal indicator of acceptable indoor air quality.

It must preserve applicable ventilation requirements and account for the building’s specific conditions.

This illustrates a broader lesson: energy optimization works best when it respects the physical purpose of each building system.

Why Artificial Intelligence Is Not Automatically the Answer

Predictive software can identify patterns that fixed schedules miss.

An office might consistently experience low occupancy on certain weekdays. A university might have irregular classroom use during examination periods. A hotel conference facility might experience abrupt changes in room demand.

Machine-learning systems can potentially use these patterns to estimate future occupancy and help optimize HVAC operation.

But prediction introduces its own risks.

A model trained on historical office attendance may perform poorly after a company changes its remote-work policy. A building might underestimate occupancy during an unexpected event. Sensors can fail, and data feeds can become unavailable.

There is also a difference between predicting an empty room accurately and using that prediction to save electricity.

If several rooms share one air-handling system, reducing demand in one room may not produce a proportional reduction in equipment energy consumption.

Likewise, a poorly tuned controller might repeatedly heat and cool a space as predictions change, undermining its intended efficiency gains.

Effective automation therefore needs safeguards, reliable equipment integration, and ongoing performance monitoring.

The Department of Energy identifies interoperability as a major obstacle to advanced building controls. Equipment from different manufacturers can use incompatible systems, leaving information fragmented across the building.

The Privacy Problem Hidden Inside Occupancy Tracking

A building that understands occupancy does not necessarily need to identify individual occupants.

That distinction matters.

Cameras, access badges, connected devices, and other technologies can reveal detailed information about people’s movements and working habits.

Although such information may improve occupancy estimates, collecting it introduces privacy, security, and governance concerns.

An energy-management system usually needs to know how many people are using a space or whether the space is occupied. It often does not need to know who those people are.

Privacy-conscious designs can prioritize anonymous counts, local sensor processing, limited data retention, and access restrictions.

A meeting room that reports an estimated occupancy of six people can support many energy-management decisions without storing the identities of those occupants.

The appropriate technology depends on the building and the applicable privacy rules. The important design principle is to collect only information that is necessary for the intended purpose.

Energy efficiency should not become a justification for unnecessary workplace surveillance.

Why Better Controls May Matter More Than More Sensors

Installing sensors is only one part of reducing waste.

A building may already have occupancy detectors but still operate inefficiently because the controls were configured incorrectly, overridden, or never properly tested.

The U.S. Department of Energy’s commercial building field research has identified lighting and HVAC controls among important sources of unrealized energy-cost savings associated with energy-code compliance.

This means a building owner should not automatically begin with an expensive artificial-intelligence platform.

A more effective improvement process might begin by examining existing operations.

Are lights turning off when rooms become empty? Do HVAC schedules reflect actual operating hours? Are temperature setbacks functioning? Do room-booking records match real occupancy?

These questions can expose inefficiencies without requiring extensive new technology.

Only after the existing system is understood does it become possible to determine whether predictive controls offer enough additional value to justify their cost and complexity.

What a Smarter Building Would Actually Do

Consider a hypothetical conference room used irregularly throughout the week.

A conventional control system might maintain normal temperature and ventilation settings from 8 a.m. until 6 p.m., regardless of how frequently the room is occupied.

A more responsive system could operate differently.

The objective is not to minimize energy consumption at every moment.

It is to provide the conditions people need while avoiding unnecessary operation when those conditions are not required.

That is a more useful definition of building intelligence than the number of sensors or algorithms installed.

The Economics of Predicting Empty Rooms

The financial value of occupancy-aware controls depends on more than electricity savings.

Building owners must also consider equipment installation, software integration, maintenance, cybersecurity, sensor calibration, and the time needed to configure and monitor the system.

A small office with efficient LED lighting and reliable occupancy sensors may gain little from adding sophisticated predictive software.

A large university campus with irregular classroom schedules, centralized HVAC equipment, and substantial variations in occupancy may offer a stronger opportunity.

The distinction is especially important for heating systems.

Not every building uses electricity for space heating. Some rely on natural gas, district heating, or other energy sources.

Consequently, a reduction in heating demand does not always translate directly into lower electricity consumption, although it may reduce overall energy use and operating costs.

The most credible business case is based on measured consumption before and after improvements, adjusted where practical for changes in weather, occupancy, and operating conditions.

Savings should be evaluated against a realistic baseline not an assumption that every empty room previously consumed energy at full capacity.

What Building Owners Can Do Now

For building owners and facility managers, the most useful starting point is an operational assessment rather than an immediate technology purchase.

Five practical priorities are worth considering:

1. Identify frequently empty spaces. Compare booking schedules, actual room use, and operating hours to locate mismatches.

2. Check existing lighting controls. Confirm that sensors work, cover the intended areas, and use appropriate shutoff delays.

3. Review HVAC schedules and setbacks. Look for unnecessary conditioning outside occupancy periods while preserving building protection and comfort.

4. Evaluate ventilation carefully. Use qualified building professionals to ensure that any occupancy-based changes maintain appropriate indoor air quality.

5. Measure results. Track electricity and other energy consumption before and after changes, accounting for seasonal and operational differences.

Buildings with centralized automation may also benefit from commissioning or recommissioningn systematic testing and adjustment of controls to ensure they operate as intended.

Predictive occupancy models can then be evaluated as an additional improvement rather than treated as a substitute for sound engineering.

The Bigger Lesson: Buildings Need to Respond to How People Actually Live

For decades, buildings have largely operated around predictable routines.

Offices opened in the morning. Classrooms followed timetables. Meeting spaces were expected to be occupied during working hours.

Those patterns still exist, but they are not always reliable indicators of actual demand.

Hybrid working, flexible schedules, changing building uses, and irregular occupancy make the difference between scheduled activity and real activity increasingly important.

The resulting energy waste is often invisible because it is distributed across thousands of ordinary decisions: a light left on, a room cooled unnecessarily, or an air-handling system operating longer than needed.

Predictive controls offer one way to narrow that gap. Yet the evidence suggests that the foundation of efficiency remains simpler: understand actual building use, maintain reliable controls, and match energy services to genuine demand.

The smartest building is not necessarily the one that predicts every movement. It is the one that can recognize when energy is needed, when it is not, and when uncertainty requires a cautious response.

That shift from operating according to assumptions to responding to real conditions could make buildings more efficient without asking the people inside them to sacrifice comfort, safety, or privacy.