Facilities management: how one missed preventive maintenance date becomes an emergency call

That is what “nothing owns the work order end to end” looks like when a federal auditor turns up with a flashlight, and it is happening inside contracts averaging $3.4 million apiece.

Why this is worse in facilities than anywhere else

Every industry loses money to coordination failures. Facilities services loses it against the thinnest margin structure in the set.

ABM Industries is the largest US-listed facilities services company. Its fiscal 2025 filing reports $8.75 billion in revenue, a 12.3% gross margin, and a 3.6% operating margin. The filing also states plainly that direct labor costs represented 68% of total revenue.

Run the conversion. At a 3.6% operating margin, one dollar of avoidable cost requires $27.80 of new revenue to replace.

That is the number to hold while reading the rest of this. A facilities services owner does not fix a coordination problem by selling more. Selling more at 3.6% is the most expensive way to solve it available.

What is actually in a technician’s day

The maintenance industry has a term for the share of a shift a technician spends actually performing maintenance: wrench time. Figures of 25% to 35% circulate everywhere, almost always without a source. Chase them and they lead to CMMS vendors citing other CMMS vendors, or to textbooks citing textbooks.

One study has a published method. Mohammad Rahman’s 2024 work-sampling study in the Journal of Industrial Intelligence observed 50 maintenance technicians at a US chemical plant across three months, recording activity every fifteen minutes, with a calculated sample size and a stated margin of error. It found 28% average wrench time, ranging from 13.3% to 45.5% by craft. HVAC technicians came in at 20%, second lowest of any trade measured.

The breakdown of the other 72% is the part worth reading.

One plant, one industry, fifty technicians. Take the shape rather than the decimals.

Two things fall out of it. The first is that pure paperwork is only 4%, so anyone telling a facilities owner his technicians are drowning in forms is selling something. The second is that the categories which are genuinely information and coordination failures, planning, tool hunting, parts hunting, closeout and reassignment, come to 23% of the shift on their own, against 28% spent on maintenance.

For every hour a technician spends turning a wrench, he loses close to an hour to finding out what he is allowed to do, where the parts are, who has to sign, and being pulled off the job by somebody else’s emergency.

And note that last row. Reassignment, 4%, defined in the study as emergency work and priority changes. That is the missed preventive maintenance date, measured, inside the day of the technician who was supposed to be somewhere else.

What that costs on your crew

The arithmetic, with every input labelled so you can substitute your own.

The Bureau of Labor Statistics puts the median wage for general maintenance and repair workers at $49,590 as of May 2025. Load that at roughly 1.35 for payroll taxes, insurance, vehicle, tools and PTO, which is an assumption rather than a sourced figure, and a technician costs about $67,000 a year, or $32.21 per paid hour across 2,080 hours.

At 28% wrench time, only 582 of those hours are spent performing maintenance. Which means the true cost of a productive hour is $115.12, and every non-productive hour costs exactly the same as a productive one.

Now take the recoverable slice, conservatively. Of the 23% lost to information failures, permits and process waits are physical constraints and stay where they are. Recover one third of the planning, tool-hunting, parts-hunting and closeout time, which is 7.7% of the shift, and you get 160 hours per technician per year, worth $5,154.

On a sixty-technician firm: $309,000 a year.

Then convert it, which is the part that lands. At ABM’s 3.6% operating margin, recovering $309,000 of cost is worth the same to the business as $8.6 million of new revenue.

For a firm turning $25 million, that is a third of the top line, without winning a single new contract.

Now the number that decides whether to act on any of this. A first step on one team starts at $15,000, priced once the scope is clear. Against $309,000 of recoverable cost, or against the $8.6 million of new revenue it would take to replace that at your margin, the question is no longer whether it is worth doing. It is whether the $309,000 is real, and six weeks of measurement answers that.

The missed date is the expensive part

Everything above is the cost when nothing goes wrong. The function of a coordination gap is to convert cheap scheduled work into expensive unscheduled work.

“Emergency repairs can cost three to more than ten times that of planned maintenance.”

— Klaus Blache, Director, Reliability and Maintainability Center, University of Tennessee

He adds that top-quartile reactive maintenance in North America currently sits at 9%.

The federal government’s own O&M guidance says the same thing from the other side.

It puts the typical US facility at more than 55% reactive, against top performers at under 10% reactive, and prices reactive maintenance at $18 per horsepower per year against $6 for a reliability-centered programme.

Worth knowing that those dollar figures originate in a 2001 pump-industry trade article that the Department of Energy republished in 2010 and nobody has updated since, which is itself a reasonable summary of how well this is measured. The reactive-versus-top-performer split is the more useful half, and it is corroborated independently by Blache’s 9%.

At the portfolio scale, the compounding is documented. GAO added building condition to its High Risk List in 2025, reporting that Department of Defense and federal civilian building deferred maintenance backlogs more than doubled, from $171 billion to $370 billion, between fiscal years 2017 and 2024. David Marroni, GAO’s Director of Physical Infrastructure Issues, told the House Appropriations Committee that without a reversal, federal assets “will continue to deteriorate and need premature replacement, which can be significantly more expensive than the cost of repairs had they not been delayed.”

That is the federal government stating this article’s argument in its own words, under oath.

And the crew turns over while you read this

There is a second cost that reprices every year.

BLS reports 1,621,800 general maintenance and repair workers employed, with about 148,700 openings projected each year, most of them replacing people who transfer out or retire rather than adding capacity. That is 9.2% of the workforce replaced annually. HVAC and refrigeration runs at exactly the same rate.

On a sixty-technician firm, that is five and a half technicians to hire and train every year, each arriving knowing nothing about which client site has the old chiller, which building manager wants photographs on every ticket, or which cooling tower needs looking at twice.

JLL’s skilled trades research puts 39% of US facilities managers over the age of 55, against 28% across all occupations. The knowledge is walking out on a schedule, and the only place it survives is in the record.

“Labor shortages often show up as delays such as inspections being pushed out and maintenance cycles stretching far beyond expectations.”

— Kyle Spencer, Director of NFPA LiNK, National Fire Protection Association

The tools are in, and they are not connected

Johnson Controls, which sells building management software, surveyed 760 US business leaders and 260 facility managers at organisations of 200 or more employees in December 2025.

Look at where. Of those using it, 42% of business leaders and 47% of facility managers apply it to predictive maintenance. Nothing in the survey shows AI applied to work order intake, dispatch, closeout, SLA reporting or invoicing.

And the barrier they name is not budget or capability. 33% of business leaders cite ease of integration as their top complaint, with data quality and integration identified as the biggest obstacles for facility managers. Nearly one in four business leaders still tracks attendance and utilisation manually or not at all.

Two thirds have AI. It sits in one tool, predicting failures, while the work order that follows the prediction is still coordinated by a person with a phone and a spreadsheet.

Spencer, again, names where the return actually is: “The most practical application of AI is speeding up administrative tasks like paperwork, scheduling and inventory tracking.”

This is the same pattern across every industry that runs the physical world. The technology lands where it demos well, and the money stays where nobody is counting.

Where the first step goes

If the loss is in coordination, the first step belongs in the layer between the request and the invoice.

Work order intake and scheduling. First-pass responses to client work order requests drafted, technician route notes turned into consistent SLA compliance reports, preventive maintenance dates flagged before they slip into an emergency call. This runs alongside Corrigo, UpKeep, Fexa, whatever you already dispatch through. It is one team, measured, from $15,000.

The six weeks, concretely: a week pulling your own PM completion and emergency call-out history and timing what coordination currently costs, two weeks configuring agents against your live CMMS and your own past SLA reports, three weeks of coordinators running real work orders through it, and a written comparison at the end. Your CMMS stays where it is, and the technicians’ day changes before the software does.

Service delivery and SLA compliance. The heavier version: an embedded pod rebuilds work order to invoice end to end, verifying punch-list items and technician photographs against each client’s SLA before a ticket closes, and generating a ready-to-send service report and invoice. Which is also the answer to the GSA audit finding: a ticket that cannot close without evidence does not get signed off on a tank full of murky water.

For an operator running forty or more client sites, there is a third layer that puts one operating layer across every site you service. That is a longer engagement and a later one.

The full picture of what AI-native facilities services looks like runs from the work order up. No operator should start with all of it.

The number to go and find

Everything above is somebody else’s data. Here is yours, and your CMMS already holds it.

Pull last quarter. What share of preventive maintenance work orders closed on time. Then pull the emergency call-outs at the same sites in the following month, and see how many trace back to a PM that slipped.

Then ask the harder question, the one the GSA audit asks: how do you know the ones marked complete were actually done?