The commercial building stock in the UK has an energy problem. Research from the British Property Federation found that 83% of commercial buildings in seven major UK cities have an Energy Performance Certificate below the B rating that will be required for commercial lettings by 2030. Most of those buildings will not be demolished and rebuilt. They will need to be made to perform differently.
That is where AI comes in. Not as a theoretical future solution. As software already running in buildings across London, Manchester, and Bristol today.
What AI Energy Platforms Actually Do
Modern AI building management systems do not just monitor. They act. They pull data from smart meters, HVAC sensors, occupancy counters, weather forecasts, and utility tariff signals. They build a continuously updated model of how the building uses energy. Then they make decisions: adjusting heating setpoints before the morning rush, dimming lighting in unoccupied zones, shifting electricity-intensive loads to off-peak tariff windows.
MRI Software's implementation at its own London HQ combined people-counting sensors, air quality monitors, temperature sensors, and smart meters feeding into an AI-powered dashboard. The result was £20,000 in annual savings. That is a mid-sized office. Scale that across a commercial portfolio and the numbers become material to asset value.
AI firms working in this space include Johnson Controls, with its OpenBlue platform, and Arup, which has used MRI's IoT Hub to combine sensor data into actionable efficiency insights. The AI is not guessing. It is pattern-matching against hundreds of thousands of hourly data points per building.
NABERS UK and Why Real Performance Metrics Matter
The problem with most building sustainability certifications is that they measure design intent, not actual performance. A building with a good EPC may perform very differently in operation. The NABERS UK rating system, launched in 2020 and adapted from Australia's successful model, addresses this directly.
NABERS rates buildings on actual measured energy consumption over 12 months. Assessors use verified meter data, normalised for weather, building size, occupancy hours, and equipment density. The result is a one-to-six star rating that reflects how the building actually operates. Buildings improving from a two-star to a five-star NABERS rating can cut energy costs by up to 50%.
This is where AI and NABERS intersect practically. An AI energy platform generates the continuous metered data trail that a NABERS assessment requires. It also identifies the specific system inefficiencies, often in M&E controls and scheduling, that explain why a building underperforms against its design rating. The gap between EPC prediction and NABERS reality is usually found in how the building is being controlled day to day, not in how it was built.
Commercial Buildings and AI Energy
83%: of UK commercial buildings in major cities have an EPC below Grade B (British Property Federation)
40%: potential reduction in energy waste when AI is combined with existing efficiency measures (Nature study)
50%: potential energy cost reduction for buildings improving from 2-star to 5-star NABERS rating
The Data Privacy Question
Smart building AI works because it collects a lot of data. Occupancy patterns. Individual desk usage. When people arrive, take lunch, work late. In commercial buildings with multiple tenants, this creates immediate GDPR and data minimisation questions that many building managers have not fully worked through.
The principle is straightforward even if the practice is not: AI energy systems should collect the minimum data necessary to achieve their efficiency objectives. Aggregate occupancy counts are sufficient for most HVAC optimisation decisions. Granular individual tracking rarely is. Tenants should be informed of what is collected and how it is used, particularly in smart buildings with shared infrastructure.
Alongside privacy, there is a bias concern that is less discussed. AI energy models trained on building data from the UK's commercial core, primarily large, well-maintained, well-instrumented office blocks in major cities, may not perform well when applied to older, smaller, or atypically occupied buildings. The model assumes the building behaves roughly like the buildings it learned from. If it does not, the optimisation recommendations may be wrong.
"The AI does not know your tenants changed their working pattern. It knows what the meter read last Tuesday."
What Facilities Managers Should Know
AI energy platforms are not set-and-forget systems. They require active oversight. The model needs to know about significant changes: new tenants, extended operating hours, major fit-out works, or changes in equipment use. A Nature study on AI's energy-saving potential in buildings found savings in the range of 40% when AI is combined with existing measures, but that assumes a functioning, well-calibrated system with regular human review.
The 2030 EPC deadline is not far off. AI energy management is not a silver bullet, but for buildings that are already well-constructed and poorly controlled, it is one of the fastest routes to performance improvement available. The key is understanding what the AI sees, what it cannot see, and where the human still needs to be in the loop.
AI and Net Zero in the Built Environment
From NABERS UK to AI building management platforms, the tools for improving commercial building performance are available now. Responsible with AI helps real estate and facilities teams use them with confidence. Visit responsiblewithai.com.
Explore the Programme → Responsible with AI
