To get the most out of artificial intelligence in facilities management, think of it as a teammate, not a tool, the head of a facilities management company says.
The large language models underneath AI are at the point where they can play a role throughout the facilities management workflow, Shawn Black, founder and CEO of OnyxFM, said last month in a Facilitiesnet webcast.
AI can write standard operating procedures, analyze data from a computerized maintenance management system, review vendor bids, write building maintenance plans and draft emails, freeing up facilities managers’ time to focus on higher-level matters, he said.
“AI has changed so much that it’s to the point where you’re going to start leaning on it like a teammate in your business,” said Black, whose company manages building repairs and maintenance and coordinates vendors for multi-site facilities. “When you do that, it changes everything.”
Use cases
Black sees AI playing three main roles in facilities management: sifting through building system data to monitor what matters, analyzing the data to predict equipment failures before they happen and then automating workflows to maintain or repair the equipment.
The newest evolution of the technology, agentic AI, is central to automated workflows, because it decides on the facility manager’s behalf what steps need to be taken and dispatches the work. “It takes the next three steps instead of recommending one,” he said. “The manager sees the one exception, not the 49 tickets.”
A team member needs to remain part of the process to keep the technology within guardrails that are set, but fundamentally, he said, “the relationship between the tickets and the work orders is going to change.”
Managers can expect to see a return on investment on all of the ways AI is brought in, he said. “If they start where the money’s leaking, the average ROI will be pretty large,” he said, in the range of 10% to 30% over a year or two of adding the technology.
Realizing an ROI on intelligent monitoring and workflow automation can happen quickly, possibly within a year, while predictive analytics takes a few years, since that application is predicated on the accumulation of data over time. “It needs history,” he said.
A national restaurant company Black works with saw an immediate ROI from adding intelligent monitoring to its CMMS in part because it could dispatch technicians during regular business hours to work on problems while the equipment was still functioning. Prior to that, the company was dispatching crews to work on problems after the equipment had failed, which typically required the teams to work after hours.
“If you can lay this out to be able to use those working hours, instead of overtime hours, [you’re] saving a lot of time and money,” he said.
Black recommends adding AI to HVAC systems first, because that’s where much of the savings opportunity is and also where the technology is furthest along.
“There are amazing vendors out there,” he said. “Getting into this is going to be really easy for you. Detection alone is going to pay for itself. One zone is heating while the other one is cooling. It happens in more buildings than anyone wants to admit.”
He recommends automating lighting next. “The value is knowing which space people are actually using,” he said. “So when you’re able to get that information, that changes the whole game. You can get predictive analytics with that and use it for maintenance, turn systems off and on to save money. There’s quite a bit you can do with occupancy data.”
The third area to automate is equipment energy use. He recommends installing sub-meters if they’re not already in place and tracking the performance of equipment. “Sub-meters are going to allow you to find out [if] something’s been running all night that nobody knew about, pieces of equipment that are running out of parameters,” he said. “All of that data is going to allow you to manage your assets at a much higher level.”
Separate from the use of AI and agentic AI for these common applications, he expects to see the use of robots increasing, especially for janitorial services. “Right now, the bots are [cleaning] the open floor while the crews are doing the edges, the detail,” he said.
From an operations standpoint, robots should be treated in the same way as other types of AI, he said. “Faciities managers are subscribing to a service, not buying the hardware,” he said. “They’re programming the software inside the robots just as they would any other AI application…. The mapping, the training, the maintenance — all this stuff is covered. The equipment is just too expensive for people to buy, and no one wants to do that. They don’t want to maintain these.”
Robots are taking over the reporting function as well, he said.
“The sleeper benefit [of robots] is the coverage report,” he said. “It’s proof that the service was done every night without anyone writing it up…. Having that report come in with all that data consistently, every single night, makes all the world of difference.”
Humanoid robots will be the next iteration of the technology, but it will be several more years before they assume facilities work in a meaningful way.
They’re “not that far away from showing up in a store or a restaurant,” he said. “You’re already placing your order [digitally] in a restaurant.”
Facilities managers planning to add AI and agentic AI to their operations should be focusing on data hygiene first, he said. The technology can only work if the underlying data is accurate.
Black said he worked with a facilities manager who had building equipment in more than 120 locations and learned, after running data from the equipment through an AI program, that about 30% of the equipment was mislabeled. “Just knowing what the equipment was and [getting it] labeled was actually a benefit,” he said.
Once the data is accurate, he said, it’s not labor intensive to keep it updated, because the process can be automated. “Let the AI do it,” he said. “It runs. You get a report [and it can] fix the anomalies in your data. But for the first time, spend some time on it.”
