Published on :
September 4, 2026
by
Anisha Bhattacharjee
Views...
The most effective way to introduce AI to facilities teams is to make clear what it is there to change, and what it isn't, before asking people to trust it. Then let them see that distinction in practice. Some of the most difficult pushback leaders encounter isn't about the technology itself. It's about what staff believe the technology means for their work, and that belief changes less through announcements than through what people actually experience once the system is in use.
AI is increasingly moving from experimentation into facilities operations. But the people buying it and the people working alongside it often look at the same rollout through different lenses. For leadership, it usually means better decisions, less reactive work, and greater visibility into what's actually happening across a site. For frontline teams, it can mean something else: changed responsibilities, closer scrutiny, and real uncertainty about where their role stands once the system is live.
That difference isn't specific to FM.But it's worth noting FM's workforce is a mix of blue-collar and white-collar roles, technicians and soft services staff doing hands-on work, managers and directors overseeing and reporting on it, and the same AI finding can read very differently depending on where in that mix someone sits. Randstad's 2026 Workmonitor, based on responses from more than 27,000 workers and 1,225 employers across 35 markets, found that 95% of employers expect business growth this year, compared with only 51% of workers. Nearly half of workers, 47%, said they feared AI would benefit companies more than employees.With a workforce this varied, that gap doesn't land as one uniform difference in FM, it stretches across a wider range of roles and reactions than the average industry.
A gap that wide doesn't close on its own. It closes through what leaders choose to explain, and what they choose to show.
AI's role in FM shifts depending on the function. Maintenance is where it applies most directly, since the work already runs on structured data, BMS trends, CMMS logs, sensor readings, giving AI a clear starting point. Soft services and security carry the same intent, using AI wherever it helps most, but look different in practice, since that work leans more on people and judgment than continuous data. Maintenance offers the clearest picture today, so it's the lens this section uses. Here, AI's role isn't necessarily to automate the work itself. It can automate much of the work before the work, the part that determines whether someone is ready to act by the time they're needed.
Different organisations use different tools for different parts of this, and there's no single standard yet for what "AI in FM" covers. But the underlying problem it's aimed at is a familiar one across maintenance-heavy industries more broadly. McKinsey's maintenance research has found that frontline technicians in many organisations spend less than half their time on hands-on work, with significant time lost to activities such as finding parts and tools, waiting for access and permits, and coordinating work. That pattern still holds: McKinsey's more recent analysis of aviation maintenance technicians found wrench time falling below 20 percent in some operations, with technicians spending a significant share of their time waiting on materials, decisions, or handovers instead of hands-on work. The implication is relevant to FM too: productivity isn't determined only by what happens during the maintenance task itself. It's also shaped by how much work is required just to get someone ready to perform it, work that can mean checking a BMS, pulling CMMS history, and piecing together information that lives in more than one place.
For us, that's the difference between adding another AI-generated signal and building a system of decisions: the goal isn't simply to surface another alert, but to bring enough context together for someone to decide what deserves attention and what should happen next.
AI, like any use of data or analytics, tends to draw a clearer line between people and process. The same visibility leadership values can look like something else to staff, closer scrutiny, or even a step toward replacement. The table below shows how that line plays out in practice, the same finding, read from two different sides of it.
Underneath that table are two distinct fears, and they call for different answers from leadership.
The first is whether findings will be used against them, not just as feedback but as grounds for exposure or eventual removal. A flag from an AI system isn't automatically evidence of individual failure, but if staff believe every finding counts against them, the system stops feeling like support and starts feeling like surveillance.
The second fear can't be waved away as easily: if AI can take on more of the work, does the role itself shrink? The honest answer is that AI will change what a job looks like. The question worth asking isn't whether that happens, but which parts should change, freeing people to spend more time on judgment, exceptions, and hands-on problem-solving, and less on the searching and coordinating around it. Positioned this way, the system works alongside a technician, more like an assistant handling the legwork, not something standing in for them. The goal shouldn't be preserving every existing task. It should be protecting the part of the work that actually needs a person.
When AI flags a performance issue, leaders should investigate whether it reflects a capability gap or an execution gap before deciding on coaching.
The sequence should be finding, then investigation, then context, then intervention. Not finding straight to blame.
This is where decision-support AI can be particularly useful. Rather than leaving someone to track down why a task fell behind, it can already have pulled together the information, dependencies, and coordination history that explain what happened, so the person reviewing it starts from a fuller picture instead of a guess. Any performance difference that remains once those process and resourcing issues are addressed is easier to assess as a potential capability gap, worth coaching directly.
A few things matter more than the rest:
This complements our earlier piece on why digital transformation stalls in facilities management. That article looks at the structural side of adoption, making new processes practical and aligning incentives. This one is about a different part of the same problem: when AI makes operational performance more visible, how leaders respond to what it finds becomes part of the operating model too.
AI will change how facilities work gets done. Pretending otherwise won't build trust with the people doing it. But not every gap AI surfaces is a gap in the person. The organisations that make that distinction, in both language and practice, will be better positioned to make AI feel like support rather than surveillance.
Because AI, like any use of data or analytics, draws a clearer line between people and process. The same visibility leadership values can look like scrutiny or a threat to job security from the staff side, which is why trust has to be built deliberately, not assumed.
AI's role shifts by function. In maintenance, where work already runs on structured data (BMS trends, CMMS logs, sensor readings), AI can reduce the searching, checking, and coordinating that happens before someone acts. In soft services and security, the same intent applies, though it looks different since that work relies more on people and judgment.
Lead with what AI removes, not just what it makes visible. Be explicit about how findings will be used and who sees them, show an early example of the benefit before asking for trust, and keep human judgement central to how findings get acted on.
Investigate whether it's a capability gap (the person lacks the skill) or an execution gap (the process or environment is blocking consistent execution) before deciding on coaching. The sequence should be finding, then investigation, then context, then intervention, not finding straight to blame.
No. Maintenance is used as the example because it already runs on structured data, giving AI a clear starting point. The same underlying intent, using AI wherever it helps most, applies to soft services and security too, even though it looks different in practice.
© Xempla. All rights reserved.