Most facilities leaders I talk to aren’t working on a big AI strategy. They’re just trying to figure out how AI makes a difference—something they can actually see in the work.
Spend enough time with a facilities team and you’ll notice a pattern. A demo gets attention. The technology excites people. Then it meets the day-to-day workflow—work orders, dispatching, vendor coordination—and it doesn’t come together as expected.
This happens when AI is introduced as something separate from how the work already gets done. The teams making real progress take a different approach. Instead of replacing their systems or rebuilding everything around AI, they’re integrating it as an intelligence layer on top of work that’s already there.
Used that way, AI works inside existing processes, helping technicians handle jobs faster, making it easier for vendors to respond to calls, and giving facilities leaders clearer visibility into what’s happening.
That distinction changes everything.
The Shift That Actually Matters
There’s a persistent idea that the endgame for AI is complete autonomy, that the system eventually just runs itself, decisions included. That’s not where this is going, especially in facilities management.
FM is inherently people-driven. Context matters. Judgment matters. The difference between a routine repair and a business-critical failure is something only a human can recognize. AI is good at identifying structure and generating options. It doesn’t understand the environment it’s operating in. It’s a powerful assistant, not a human operator.
We’re not shifting from human to machine. We’re shifting from manual to augmented.
When the balance is right, the role of the facilities leader expands. Less time digging through notes and chasing status updates. More time making decisions that require real expertise and hands-on experience.
What AI “Working” Really Looks Like
It’s easy to think about AI in the abstract. A lot of the conversation lives there. But when you look at what happens inside an actual workflow, things get concrete fast.
Take one national retailer. They weren’t trying to overhaul their systems. They focused on a single problem: work orders getting submitted incorrectly. Store teams were selecting the wrong part types for lighting repairs, which meant providers would arrive unprepared, jobs would get canceled, and the cycle would start over—delays, frustration, and a steady stream of avoidable rework.
When AI was introduced into that workflow, it intervened at the point of strain: using historical data to suggest correct inputs at the moment the request was created.
No big transformation. No system replacement.
But the results were clear. Cancellations dropped. Dispatch became predictable. And the people involved spent less time fixing mistakes that shouldn’t have happened in the first place.
The Underestimated Element: Data
AI has a reputation for being magical. Feed it enough information and it conjures insights out of thin air. But AI isn’t magic. It’s pattern recognition. And patterns require a foundation.
In facilities management, that usually means three things:
- Location and asset data that reflects the real world
- Provider data that’s complete and reliable
- Work orders updated consistently as work progresses
If the data isn’t there, the patterns don’t emerge.
This is where a lot of AI initiatives stall—not because the technology is incapable, but because the foundation isn’t built to support it. The AI can only find what the data can show it. Gaps in structure mean gaps in insight.
Where to Start (Without Overthinking It)
The biggest mistake I see is teams trying to define a full AI strategy before testing a single use case. Long planning cycles. Very little actual progress.
A better approach is smaller and more deliberate. Pick one workflow that has both volume and consistent friction—something where:
- The same task happens over and over
- Errors or delays are common
- The impact of improvement will be easy to measure
Baseline what’s happening today. How long does it take? Where do things break? What’s getting reworked?
Then introduce AI in one specific way—just enough to influence that single process. Give it 90 days. Measure what changed. No AI roadmap. No sweeping commitments. One focused experiment. If it works, expand. If not, move on.
This “one workflow at a time” model tends to outperform big, centralized AI initiatives for a straightforward reason: the work itself isn’t centralized. It happens across hundreds of locations, with different people, different conditions, and different failure points. That’s where AI has leverage—inside the workflows where those decisions actually get made.
Bottom Line: AI Isn’t the System. It’s What Makes the System Work Better.
AI is not a new system you install. It’s a capability you layer into systems that already exist.
When it’s treated like a replacement, it creates more complexity than it removes. When it’s built into the point where the workflow usually breaks, it becomes the kind of improvement facilities teams notice without having to be told where to look.
That’s the version worth building toward.
