How Artificial Intelligence is Reshaping Facilities Management

Building Operating Management Articles

Artificial intelligence, or simply AI, is the new business buzzword. It’s hard to go even a few hours without hearing the term — from commercials on TV, to billboards, and of course throughout the workday. But what does AI really mean? And what impact does this have on operating buildings today and in the future?

The Brookings Institution calls AI one of the most misunderstood terms amongst business leaders. Quoting researchers who have studied AI, the Brookings article says that “AI generally is thought to refer to ‘machines that respond to stimulation consistent with traditional responses from humans, given the human capacity for contemplation, judgment, and intention.’”  

Many of the AI products in facility management are software-based, and, as the Brookings article puts it: “‘make decisions which normally require [a] human level of expertise’ and help people anticipate problems or deal with issues as they come up.” This idea reflects the goal of commercially available products that make buildings more efficient and operationally sound.

Adding to the confusion around AI, many of these solutions for commercial buildings could instead be termed “machine learning” (ML) offerings, as they take substantial amounts of data and use it to predict or calculate scenarios. The building operator can then act on the insights. 

Despite confusion around this nascent technology, it’s important for facility leaders to understand AI and how it can impact their operations. The McKinsey Global Institute estimates that by 2030, AI will deliver $13 trillion in additional economic output worldwide. To put this in perspective, the value of U.S. commercial real estate is estimated at $15 trillion

Moreover, adoption of AI is on the rise. The Wall Street Journal recently citedreport from Deloitte, which found that 25 percent of firms surveyed already have implemented AI or ML solutions. And, within two years, Deloitte expects 75 percent of firms to have implemented (or developed plans to implement) these solutions. 

AI in FM today

There have been some high-profile examples of AI in building management. Google announced in 2018 that it was using AI to manage cooling at some of its data centers. The software had been operating for a few years, adjusting cooling in real time, without human intervention. Google reported that it had saved 40 percent on energy use in these cooling systems. Data centers are a particularly good candidate for AI because the cooling demands are high and the risks of not providing enough conditioned air have an extremely detrimental impact on computing performance. And these performance impacts are quantifiable and immediately clear. 

The Edge in Amsterdam is another example. Currently considered one of the most advanced and greenest buildings in the world, the Edge has deployed 30,000 sensors, collecting data about the building’s operations and how occupants interact with it. 

A common use case of AI is to take significant amounts of data and distill the key inputs that a user can act on. At the Edge, commuters are directed to open parking spaces. Additionally, if the occupancy is less than expected, certain areas of the building are closed to reduce resource consumption. With all the data being collected and the use of AI, it’s possible to develop estimates of occupancy, and use that to change operations in real time.

At this point, only some buildings are investing heavily in AI. But more will follow. 

Advanced analytics like artificial intelligence and machine learning present significant opportunities to reduce operating costs and improve outcomes for occupants. Most AI applications first collect substantial amounts of data and normalize it. 

In a building, energy management is a common use case. 

Actual energy consumption data can be compared to weather, occupancy, and other factors. AI solutions can then determine the dependencies between weather and energy. When temperatures rise by X degrees, it’s likely that energy demand will rise by Y kilowatts, for example. A model is built to represent these dependencies. Then, moving forward, actual or future weather data or predictions can be used to forecast energy use. More data means that a more accurate model is built, which improves the accuracy of the predictions.