Most Building Owners Are Waiting for Equipment to Break. I'm Not.

Nexus Labs runs one of the more credible communities in the connected buildings space. Their members are facilities managers, building operations directors, and real estate executives at serious organizations. Not hobbyists. Not people kicking tires. People who have to answer for equipment failures, energy bills, and occupant complaints.

This week they ran NexusCast #2, a half-day event on condition-based maintenance. Three hundred registrants. More than a hundred owner organizations in the room, including JPMorgan, DFW Airport, the University of Iowa, and Weber State University.

They polled the audience on their own program maturity. The results were striking.

51% said Level 1: reactive maintenance and calendar PMs running the show. No sensor data in the workflow. Equipment runs until it breaks, then gets fixed. Another 17% said Level 2: sensors deployed, but maintenance behavior unchanged. The technology is in the building, but the faults pile up unresolved, or flood the work order system without closing.

That's 68% still in setup mode or not started at all. And this is the self-selected group — the people who chose to spend a Wednesday afternoon at a CBM seminar.

"The goal is not CBM," said Travis Criner, Senior Director at CBRE and a former HVAC technician with 18 years in the field. "It's the right maintenance at the right time with the right person for the right reason."

That framing is useful. Condition-based maintenance isn't the destination. Knowing what your equipment is doing before it fails is the destination. CBM is just one path there.

The path I've been walking uses $25 smart plugs and $15 sensors.

What Level 3 Actually Looks Like

The Nexus maturity model has a Level 3: signals, triage, and work order execution connect into a loop that produces outcomes you can report. Fewer emergencies. Verified fixes. Declining PM hours on equipment that doesn't need them.

Here's what Level 3 looked like at my sump pump installation.

A Shelly smart plug monitors the pump's power draw continuously. The float switch started sticking intermittently six months ago — not enough to trip a breaker, not enough to trigger an obvious fault. Just enough to make the motor run under load without actually pumping. The kind of thing you'd never catch on a quarterly inspection. The kind of thing that burns out a motor over six to twelve months, usually on the worst possible night.

The AI model caught the pattern at 3am during a storm that dropped two inches of rainfall in the forecast window. It ran 97 automated recovery cycles. No flood. No emergency call. No $3,000 motor replacement. No water in the basement.

That's not enterprise technology. That's a $25 plug and a local model watching a number. But the outcome is exactly what Level 3 is supposed to produce: a signal that turned into action before the failure.

The Signal-to-Resolution Gap

The enterprise CBM programs at NexusCast have a different version of this problem. At the University of Iowa, the fault detection system generates more than 3,500 alerts per day across campus buildings. Brad Dameron's team runs triage through two experienced engineers who sort faults into three buckets: Asset Optimization for investigation, controls for programming issues, frontline mechanical for physical repair.

That's a human bottleneck. Smart people doing the routing work that a smaller system would handle automatically.

At my 40-device community center, a week of data surfaces two or three actionable signals. The AI handles the routing. The building owner doesn't need a dedicated CBM analyst. The signal-to-resolution workflow fits in a weekly digest.

This isn't a limitation of small-building monitoring. It's an advantage. The enterprise programs are building toward what a well-designed small deployment already has — a manageable signal stream with clear action behind each alert.

What the Numbers Actually Say

Weber State University deployed building monitoring with a $5 million internal revolving loan. They generate $3.1 million per year in utility savings. They refused to weather-normalize their reporting — a deliberate choice to speak the CFO's language rather than the engineer's. "The goal," they said, "is to produce a number that leadership can act on."

$3.1 million per year. From monitoring.

A building one percent that size — a small office, a community center, a church, a local retail strip — captures a proportional share of that value. On $31,000 per year in savings, a $199/month monitoring service pays for itself in under eight months. What's harder to price is the emergency that didn't happen.

JPMorgan deployed OT network monitoring at 270 Park Avenue across roughly 8,000 devices. Within the first cycle, they found equipment that vendors had never disclosed during design and construction — wireless access points, CCTV cameras, VoIP phones, devices running on the building's network without documentation. That's not an edge case. That's what a building looks like when nobody has ever looked at it systematically.

Most small buildings have never been looked at systematically. The upside is the same. Just at a different scale.

Why 68% Are Still Reactive

It's not ignorance. The people at NexusCast know what CBM is. They signed up for a half-day seminar on it. The reasons they're stuck at Level 1 or 2 are different.

Tearle Whitson spent seventeen years inside Microsoft's smart buildings program. His CBM platform once live-calculated $2.5 million in energy savings from a single algorithm. Then a current transducer on one VFD turned out to be misreading. The $2.5 million disappeared. They had to claw back credibility with the CFO.

His rule now: calibration and verification before the first fault hits a work order. Stand up the data integrity layer before you start making promises about the data.

That's a real lesson. And it's one of the reasons enterprise programs stall. The technology stack is expensive. The data validation layer takes time. The organizational change management — getting maintenance techs to trust and act on algorithmic fault flags — takes longer. By the time you've done all that, you've spent a year and haven't prevented a single failure.

The small-building version doesn't have that problem. A power monitoring plug is calibrated at the factory. A temperature sensor either reads correctly or it doesn't. The signal set is small enough to validate quickly. You can be at Level 3 in a week.

What's Coming Next

This week, Google DeepMind released Gemini Robotics ER 1.6 through its developer API. The gauge-reading capabilities are the part I keep thinking about. Point a camera at any analog gauge, any LED indicator, any physical dial, and the model reads it — without a QR code, without a custom integration, without a dedicated sensor.

Every mechanical room has twenty gauges that nobody reads systematically. Every HVAC unit has status lights that only get checked during scheduled maintenance. Every distribution panel has indicators that get glanced at during walkthroughs. With a camera and an API call, all of that becomes continuous monitored data.

This is not available to enterprise building management systems yet. It is available right now to anyone willing to put a $30 camera on a wall and run it through an API. The technology distribution is, again, faster at the small end.

The Argument for Not Waiting

The Nexus data tells you something useful: most of the buildings that should be monitored, aren't. Most of the owners who know they should be doing predictive maintenance, aren't. The awareness gap and the deployment gap are both real.

What they don't tell you is that the tools to close those gaps are already here, at prices that don't require a capital budget or a C-suite approval. You can deploy continuous power monitoring, temperature tracking, and anomaly detection on a small building for under $500 in hardware. The AI runs locally. The data stays yours. The first alert that saves you an emergency call pays for the entire system.

DFW Airport's team made a change that sounds obvious in retrospect: they stopped counting a closed work order as a resolved condition. If the fault persisted after the ticket closed, it reopened. Contractors stayed accountable to the actual problem, not the paperwork.

"A closed work order is not a resolved condition," their team said.

That principle applies at any scale. A sump pump that didn't flood is a resolved condition. A motor that ran 97 recovery cycles without burning out is a resolved condition. A community center HVAC that stopped running empty on weekends is a resolved condition.

The 68% waiting for equipment to break will get their resolved conditions eventually. They'll just pay emergency rates for them.

Read the case studies: How edge AI prevented a basement flood | Smart building on a shoestring | Making a sump pump agentic | The edge AI hardware wave has arrived