Artificial Intelligence (AI) is making headlines everywhere—but when it comes to buildings, what does it actually do? I hosted a discussion at ACEEE’s Summer Study last summer and shared my initial learnings in a blog post “Can Artificial Intelligence Get Us to Net-Zero Buildings?”, which outlined the potential applications of AI—especially large language models—and the challenges of integrating them into the building space.
Since then, conversations about AI in business have accelerated, but many in the building industry are still searching for clarity: How can AI really help us tackle complex problems like energy use, peak demand, carbon emissions, and—most importantly—occupant comfort, health, and building functionality?
At a recent roundtable I facilitated at NBI’s Getting to Zero Forum in Los Angeles, we dug into that question with experts from across the building tech world. The big takeaway? AI isn’t some future promise—it’s already at work, cutting costs, optimizing systems, and preventing failures. Here’s what we learned.
AI Is Already Optimizing How Buildings Breathe and Operate
AI for buildings refers to a broad set of technologies that synthesize and analyze data to provide insights, recommendations, and automations. Subsets within AI include machine learning, deep learning, large language models, and generative AI. These range from straight forward algorithms and decision-trees to complicated neural networks and transformers. Many of the “smart” building systems already in use are utilizing AI’s capabilities, but there are numerous other potential applications and innovations being explored for use in buildings and with the electricity grid.
Forget robots. In buildings, AI quietly works in the background—learning energy use patterns, spotting inefficiencies, and recommending smarter ways to run systems like HVAC, lighting, andventilation. Alexi Miller, Director of Building Innovation at NBI, shared the recent example of Southern California’s Sonora Elementary School, where an AI-driven dynamic model predictive control retrofit reduced HVAC peak demand by 30%, reduced total building peak demand by 24%, and shifted 16% of the cooling load from on-peak to off-peak hours – while maintaining tighter indoor temperature control than the previous system. By layering machine learning on top of traditional controls while tightening zone temperature controls, the system reduced compressor runtime, avoided dramatic indoor temperature swings, and prevented all compressors from kicking on simultaneously—flattening peaks, reducing energy costs, avoiding expensive demand charges, reducing maintenance costs, and cutting emissions. All this was done without updating the existing rooftop HVAC units themselves: it was purely a controls retrofit project.
And this isn’t an isolated case. As part of broader efforts from the California Energy Commission (CEC) to improve building-grid integration and enable virtual power plants, CEC and the Lawrence Berkeley National Lab have created a database with time-varying electricity rates, greenhouse gas emissions data, and California Flex Alert signals—called MIDAS (Market Informed Demand Automation Server). Realtime, publicly available data streams like MIDAS are key to smart building controls systems that can cut carbon emissions. NBI is involved in one such project: a CEC-funded pilot project in which two Southern California high schools will roll out a virtual power plant controls retrofit. The AI-powered controls aim to harmonize HVAC, lighting, and electric vehicle charging to co-optimize cost savings and carbon emissions reductions in real time.
Connecting the Dots: Digital Twins and Data Islands
One emerging innovation is the use of digital twins—a virtual duplicate model of real-world buildings—that can be paired with AI to optimize building operations. The digital twin eliminates “data islands” by bringing together isolated information from various building systems including building information models (BIM) used during design and construction, real-time data from the building management systems (BMS), and other systems like low voltage systems, HVAC, appliances, central utility plants. By layering in live inputs such as occupancy, weather, and energy prices, AI tools can now synthesize both static and dynamic data to optimize whole-building operations—going well beyond traditional HVAC controls and building automation.
In our roundtable discussion, Bert Van Hoof, CEO of Willow, shared examples from airports and hospitals where AI manages the operation of ovens and cooking equipment to ensure they don’t all turn on at the same time, reducing peak demand without disrupting meal service for travelers or patients. Thanks to new large language models capabilities, AI can also quickly “read” and learn from building manuals, diagrams, and drawings. This enables continuous commissioning, predicting and preventing system failures like leaks or equipment malfunctions—avoiding costly and disruptive emergencies. The ability to rapidly diagnose and address issues—reducing what typically takes weeks down to minutes—makes AI particularly valuable for mission-critical facilities. As a result, airports, hospitals, and similar buildings are likely to be early adopters, showcasing AI’s real-world value in building operations.
From Design to Renovation: Where AI Meets Building Lifecycle
AI isn’t just for smart building controls—it’s reshaping the entire building lifecycle, from design to operations. The data visualization capabilities of AI can provide powerful tools for answering questions and presenting analyses in a more intuitive and accessible manner. Ramya Shivkumar, Cove’s Director of Sustainability, shared how AI can be helpful for architects early in design. Her team uses AI tools to help designers stay current with land use and building regulations, while also analyzing site-specific climate risks that go far beyond basic code requirements. With Cove’s AI tools, designers can visualize which resilience design measures are critical for their building site. They are also developing tools using AI to calculate weighted embodied carbon scores for different materials, which would reduce the amount of time design teams need to evaluate sustainable and climate-resilient material choices.
Hardik Shah, Sr. Manager of Technical Program at GTI Energy, described how his team is using composite AI models (LLM and rule-based) alongside LIDAR data to pinpoint opportunities for building upgrades via envelope retrofit risk profile identifications. Also, using open-source tools like OpenAI and Perplexity, they perform deep research and consumption analysis to right-size the equipment and achieve optimal performance. They layer rule-based models on top of this analysis to extract the targeted information needed for specific building system upgrades. Models like these streamline the process of validating energy consumption data and selecting system upgrades based on actual building energy loads.
Another area of exploration is driving behavioral changes in homes by enabling smart thermostats to engage in more meaningful conversations with residents, presenting them with comprehensive information such as their home’s carbon emissions and the price of energy under specific weather conditions.
Common Concerns and Practical Guidance
For building owners, operators, and designers just starting to explore AI, here are some key questions to ask before diving in.
- What systems already collect data? Start by identifying where data already exists—your Building Management System (BMS), utility meters, connected devices, even building drawings and equipment manuals. You likely have more building data than you realize. The key is understanding that AI is only as good as the data it’s grounded in—and the prompts it’s given. However, don’t let data gaps hold you back. Many AI applications can work with imperfect or incomplete datasets, especially if you keep a human in the loop to validate AI recommendations.
- Where are your biggest pain points? Is high peak demand driving up costs due to a lack of system-level coordination? Are comfort complaints increasing because building systems are out of tune and occupancy patterns are unpredictable? Or are maintenance expenses rising as systems age and degrade? Identifying your most pressing challenges is the best place to start. Targeting high-value problems first can demonstrate quick wins— and think about the full range of potential co-benefits like improved occupant satisfaction, lower emissions, and more efficient operations. AI projects are most successful when they are cross-functional, with energy, facilities, sustainability, and IT teams working together to find multipurpose solutions.
- Will AI make it harder to switch platforms later? Many owners are already stuck with building management system platforms that do not meet their needs anymore and are unable to switch due to the cost and hassle of migrating data and retraining staff. In our roundtable discussion at the Getting to Zero Forum, some attendees expressed concern that adding AI on top of rigid systems could risk deepening that lock-in. Fortunately, tech companies are increasingly aware of this risk and are working to improve interoperability between devices and services. Since AI tools can potentially aggregate data from multiple BMS and other systems—including legacy platforms—they can serve as integrators or linchpins for legacy building devices that are otherwise difficult to connect. Still, building owners should always ask critical questions before investing: How open and flexible is this system? What happens to my data if the vendor disappears in 10 years? Raising these demands will also push the industry toward more open and lasting solutions.
Critical building system information is often lost during the handoff between design, construction, and operations. If integrated early, AI could help preserve and transfer that knowledge—improving building performance across the entire lifecycle. To truly realize the potential, we need a holistic approach that connects design and operations. AI can be the thread that ties it all together.
The Road Ahead
The future of AI in buildings isn’t some far-off concept—it’s happening now in schools, hospitals, airports, offices, and design studios. What’s needed next is more systematic thinking to connect these early applications, demonstrate real-world use cases to the broader building community, and verify the energy, carbon, and cost savings they deliver. As Ana Lasso, Director of California’s Department of General Services, noted during the opening plenary at the Getting to Zero Forum, “We don’t invent something and then look for problems it can solve. We already know the challenges we face—and we want tech companies to partner with us to address those challenges.”
The early-stage efforts we discussed at the roundtable were not about creating proprietary tools or services; they were about exploring how AI can meaningfully solve real building problems and inform the next generation of AI capabilities. The concept of smart homes or connected devices themselves aren’t new—but AI brings the speed, scale, and intelligence needed to fully unlock their value.
With over two decades of experience transforming the built environment through innovative technologies, practices, and policies, NBI is committed to leading the charge in harnessing AI to create a carbon-free, healthy, safe, and resilient built environment for all—at speed and scale. As a trusted voice in the industry, we aim to connect pioneers and leaders, showcase real-world applications, and validate the most beneficial uses of AI in buildings. If you’re interested in supporting this mission or joining the network, please contact us here.
Author
by Nora Wang Esram, CEO of Codes and Policy, New Buildings Institute
Bio
Did you enjoy this content? Consider supporting NBI’s work with a donation today.