Using AI to maximize resiliency in infrastructure

Editor’s note: This is a narrative version of Kevin Scarborough’s comments on the Energy Beat Podcast. You can listen to the complete podcast interviews here.

At Siemens, we recognize that there is a global energy transition happening. We’re confident in this observation because we’re basing our long-term strategy on five global megatrends: demographic change, urbanization, glocalization, environmental change and resource efficiency, and digitalization. Each of these will have significant impact on energy—how we generate it, distribute it, and use it. And as we all know, companies are prioritizing renewable energy, energy efficiency, and electrification of their assets.

We also know this transition is straining existing energy and building infrastructure. It requires rapid transformation to maintain the reliability and sustainability of these complex systems.

The building sector represents about 40 percent of global energy demand and the amount of usable space in buildings across the globe is expected to double by 2060. That's a lot of square footage that's about to be built, putting demand on our resources. This is one reason why electricity demand is expected to triple by 2050. Thus, as our CEO of Smart Infrastructure Matthias Rebellius recently stated in his post for Reuters Plus, “a resilient energy supply has gained in importance to become the number-one infrastructure priority.”

Siemens views this as a major opportunity for building and energy infrastructure to become more intelligent with AI, to create that sort of future-proof technology that we aspire to.

We are leveraging and researching AI within what we call our Building X ecosystem to improve delivery of data reporting that in turn brings greater energy efficiency and operational outcomes for customers. One example of this would be using generative AI to help identify the history of work orders for an asset to help prioritize maintenance and respond more quickly to maintenance needs. We also use machine learning to optimize computer-room air handlers and air conditioners.

Put simply, we’re asking AI to think about energy use within buildings and the buildings themselves. But like everything in this new landscape, there are challenges and opportunities—underscored by a huge need for the right data.

The challenges of integrating AI into the energy and utility industries

The three biggest challenges for integrating AI into the energy and utility industries are data security, data governance, and over-reliance on AI.

Data security related to AI is largely unregulated, so this is an area of both opportunity and risk. Opportunity could arise by industry associations like the American Society of Heating, Refrigerating and Air-Conditioning Engineers (ASHRAE) coming together to create best practices for using AI in the industry. The potential risks within data security include bad actors using AI in a malignant way, such as hacking. The industries must evolve their approach to data security quickly to address these challenges. Data and decisions must be auditable, and there must be role-based access control to the algorithms to prevent bad actors from impacting the brains of the AI engine in use.

In data governance, for the industry to be efficient and effective in driving results, AI must have a uniform output, especially with generative AI. Quality data governance includes standardized data categorizations, reporting mechanisms, and communications, especially if the AI is communicating with a building automation system. Ease of communications helps prevent the creation of data silos that hinder the sharing of key data.

There is, however, big risk around over-reliance on AI. If we rely on AI too much, we lose that key spark of creativity; everything will be based on—and limited to—what's fed to the AI engine. This can lead to bias and complacency, especially with generative AI. Companies that use AI as a creativity tool do so successfully only with the input and oversight of human minds.

Ultimately, an AI engine will only be as smart as the data that's fed into it.

Using the best data to leverage AI in the building and energy industries

In any sort of AI application in buildings and energy, the most important data is accurate, applicable historical data, because to predict the future, we need to know how things reacted or operated in the past to various stimuli to that system. Historical data can include images—you feed images of a central plant, for example, into an AI engine to help you scour the internet for design information for a device or a motor.

With generative AI, having work-order reports and a full history of maintenance on a specific system could give a specialist—someone who is an expert in optimizing a system—the ability to move more quickly to address a complex problem that a customer may have. With machine learning, having a building automation system that uses the correct sensors and formatted with the correct trend data and measurements is absolutely critical.

Both data accuracy and a really reliable amount of data are essential for optimizing AI. For example, weather data can be used to make near real-time decisions for a central operations plan, or for an air handler using machine learning, to calculate potential savings. This is a nod toward the use of an application programming interface (API), a digital intermediary between two applications that enables one program to request data or functionality from another without having to know how the other system works. System managers can create APIs that can be integrated into the in-use AI, resulting in greater scalability, enhanced functionality, and flexibility with the ability to customize outputs to meet the specific needs of users.

Figuring out the right amount of historical data for AI

How do you determine what extent of historical data you need? That depends on what actually happened in, say, the past three years. Is that enough or do you really need more?

In my background as an energy engineer, most managers would want two to three years’ worth of utility data. But if you told me during the COVID pandemic that you wanted three years of utility data, I would tell you 2020 and 2021 were not really relevant to forecasting what the future is going to look like because buildings were unoccupied at that time.

How much historic data you need also depends on the building system, the building itself, and what you’re trying to achieve operationally. An example of this is when you've got a conference-room air handler that needs data. If there’s one meeting a day in that conference room, I could probably get by with about two weeks’ worth of data. If the temperature outside was representative of a greater quantity of hours at a certain temperature, that should be sufficient data.

A lot of considerations about historical data depend on the application, but energy managers must seriously look at the past and how global events might have shaped the actual quality of the data that they have.

Using AI to address load management and demand flexibility

The energy industry is learning how to leverage machine learning and AI for the very key matters of optimizing and right-sizing the load on the grid and the systems. Digital twin technology also plays a major role in this. A digital twin can help increase the lifetime of electrical assets by detecting faults in these systems before they become a problem. AI can power this digital twin by processing the large data sets necessary for creating twins of such assets.

Another technology that can come into play for energy distribution is an offering within the Siemens Xcelerator platform: Electrification X. This offering, built on cloud services, is designed to manage, optimize, and automate electrification infrastructure by providing a holistic view of substations and other assets. A feature set within Electrification X, called Electrification X Asset Management, utilizes sensor data analytics to enable asset owners and operators to increase uptime and reliability, reduce operational expense, and bolster cybersecurity.

Siemens has another offering called Gridscale X that provides easy-to-deploy modular software for end-to-end grid management and leverages AI to analyze vast amounts of grid data. Layered within that offering is Gridscale X DER Insights, which uses AI to uncover behind-the-meter distributed energy resources (DERs) and assess their behavior and their impact on grid equipment. This helps with forecasting, analyzing, and extracting useful insights and turning them into actionable next steps. This is valuable because it helps customers optimize DER operations, understand DER performance and health, and maximize business and grid resilience.

Staying ahead of the major AI trends

Industry managers must keep an eye on the AI regulations that are coming in the U.S. and globally because they will surely impact how AI develops in the future. This means that what we’re doing now with AI will change. We also need to be on the watch for new AI capabilities and technologies as they come online. Industry managers need to learn quickly how these new elements can apply to building and energy operations so that their value can make a difference for business and for customers.

Ultimately, we must all have an open mind about what AI can do. Let’s embrace the changes that are coming and assert the human role in AI. We can be cautious, but we shouldn’t be afraid. Everyone can lean on AI by learning all that you can about it and try to grow your business, using it where it adds value, improves operations, and benefits customers.

Published: December 30, 2025