What the Janitor Knows That Your AI Doesn't

There's a workforce crisis coming in facility management, and it's not the one everyone's talking about.

Yes, the numbers are brutal. IFMA estimates that 40% of the FM workforce is approaching retirement age. The Bureau of Labor Statistics puts the maintenance technician shortfall at 53%, with roughly 158,000 unfilled roles in the US alone. Those numbers are real and they're urgent.

But the crisis behind the crisis is worse: every single one of those retiring FMs is walking out the door with decades of knowledge that was never documented. It's not in a system. It's not in a database. It's in their heads. And when they leave, it's gone.

The Knowledge That Doesn't Fit in a Spreadsheet

I've spent twenty years in IT and nearly a decade at Salesforce building the RETECH organization. In all that time, the hardest problem I've encountered isn't a technical one. It's getting what people know into a system that other people can use.

In facility management, this knowledge is incredibly specific. It's physical. It's spatial. It's the kind of thing you learn by walking a building for five years.

Dave knows that the chiller on floor six makes a specific sound two weeks before it fails. Maria knows that the loading dock door sticks in cold weather and needs to be manually adjusted before 7 AM or deliveries get backed up. Tony the coffee service guy knows that the espresso machine in the EBC breaks down every time someone uses the "extra hot" setting with the large cup size.

That knowledge is priceless. It prevents failures. It saves money. It keeps buildings running. And right now, across the entire FM industry, it lives in approximately zero databases.

The Leak Sensor Lesson

I keep coming back to a story from my time at a previous company. We'd installed leak detection sensors in the bathrooms and kitchens — real IoT, connected to the building management system, AI-driven alerting. The works.

The sensors triggered constantly. The response team would sprint to the alert location and find the cleaning crew. Mopping. Water on the floor. Sensor says "leak." AI says "dispatch team." Team finds janitor with mop. Repeat.

Any human in that building could have filtered that alert in two seconds. Jose mops the third floor bathrooms at 6:30 AM on Tuesdays and Thursdays. That's it. That's the missing data point. And no amount of machine learning was going to discover it because cleaning schedules aren't in the training data.

This is what I mean when I say AI augments facility management knowledge — it doesn't replace it. The sensor detects water. The AI classifies the alert. But only a human who knows the building can tell you whether it's a crisis or a guy with a mop.

The Facility Management Workforce We're Losing

The FM workforce crisis isn't just about headcount. It's about knowledge loss at a scale this industry has never experienced.

Think about what happens when an experienced facilities manager retires. They know which vendors are reliable and which ones pad their invoices. They know the quirks of every system in the building — the HVAC unit that needs a manual override in August, the fire panel that throws false alarms when it rains, the elevator that hesitates between floors three and four but isn't actually broken. They know the relationships: who to call at the utility company, which building inspector to talk to, which tenant needs extra attention.

When that person leaves, the replacement gets the org chart and a set of login credentials. Maybe a three-ring binder that was last updated in 2019. They're expected to manage the same building, the same tenants, the same budget — with none of the context that made their predecessor effective.

This is happening right now, across the industry, at a rate of roughly 40% of the workforce. We're about to lose more institutional knowledge in the next decade than we've lost in the previous fifty years combined.

Technology That Captures, Not Replaces

Here's where I think the technology conversation in FM has gone sideways. The pitch from most vendors is: "AI will predict what's going to break and tell your team what to do." The reality should be: "Let's capture what your team already knows so it doesn't disappear."

When Maria inspects a kitchen and notices mineral buildup on the dishwasher, she adds a photo and a voice note: "Needs descaling, third time this quarter, check water softener." That observation — from someone who's done this inspection a hundred times — is worth more than any sensor data. But only if it gets captured, indexed, and made searchable for the next person.

When Tony the vendor finishes repairing the espresso machine, Maria notes: "Replaced dispensing valve, same issue as six months ago, should last twelve months." That repair history, from someone who knows this specific machine, is the foundation of predictive maintenance. Not the prediction part — the data part.

The facility management workforce doesn't need technology that thinks for them. They need technology that listens to them. That captures what they see, hear, and know in the thirty seconds they have between one issue and the next. On a phone. While walking. Without logging into a desktop application that was designed by someone who's never been in a facilities room.

What This Means for Your Team

If you're managing a facilities team right now, you're probably already feeling this. The experienced people are getting closer to retirement. The new hires are competent but don't have the institutional knowledge. And you're trying to maintain the same level of service with a workforce that turns over faster than it used to.

My honest recommendation: start capturing knowledge now. Don't wait for the perfect system. Don't wait for the AI transformation. Start with the simple stuff — every inspection documented, every repair noted, every vendor interaction tracked. Make it stupidly easy to do, because if it's not easy, it won't happen.

At sonpito, that's the core of what we're building. Not because knowledge management sounds exciting in a pitch deck — it doesn't — but because Fawn and I have seen what happens when buildings lose their institutional memory. The new person makes the same mistakes the last person already solved. The vendor charges for work that was already done six months ago. The failure that Dave could have predicted by listening to the chiller becomes an emergency that costs $50K to fix.

The facility management workforce crisis is real. The technology response should be about preservation and amplification of human knowledge — not replacement. The janitor knows things your AI doesn't. Let's make sure those things don't walk out the door.