Labor Day evening, I worked through a list I had made the night before.
I told myself I would stop at noon. I nearly did.
Later, I checked the X account I had been trying to bring back to life. I had posted several times, replied to others, and tried to write something worth reading.
Not much happened.
A few people saw the posts. Nobody new followed.
That is not exactly a hardship. It doesn’t compare to working outside in bad weather, getting called out on a weekend, or trying to bring a neglected system back from the dead.
Still, it got me thinking about two kinds of systems I am learning to work around: the new technology changing the trade and the online machinery changing how people find the work—and the people who understand it.
In the physical world, I can usually follow the path between effort and result. A contactor closes. A motor starts. A blower moves air. If the air never reaches the room, there is a reason. Maybe the duct is undersized. Maybe a damper is closed. Maybe somebody buried a filter behind twelve years of dust.
Online, the path is harder to see.
You can write something decent, push the button, and watch it disappear.
Another invisible system
X says its recommendations depend on signals such as the accounts and topics people follow, the posts they like, what their networks like, and whom those networks follow. In other words, publishing something does not mean it will be distributed. The platform decides where it might fit and whether anyone is likely to care.
That sounds a little like ductwork.
Creating the post is only the beginning. It still has to travel through a system, pass several restrictions, and arrive in front of somebody who wants what it carries. If it never reaches the room, it doesn’t matter how good the air was when it left the unit.
The difference is that I can open a panel, follow a duct run, and usually find the restriction. I cannot open the algorithm and see where the post went.
That isn’t a complaint. It is a reminder that there are important systems around us that most people never see.
I happen to work on some of them.
Now I am trying to understand the others.
The trade is not disappearing.
There is plenty of talk about artificial intelligence and robots replacing people. Some of it is serious. Some of it sounds like it was written by someone who has never opened a mechanical room door.
The numbers do not show HVAC work disappearing.
The Bureau of Labor Statistics projects employment for HVAC and refrigeration mechanics and installers to grow 11 percent between 2025 and 2035, compared with 3 percent for all occupations. It expects roughly 40,600 openings each year over that period.
BLS points to sophisticated climate-control systems and commercial construction, including data centers, as demand drivers. In practice, that means technicians will encounter more computerized, connected equipment—and will need to understand more than the mechanical side of the system.
That last part deserves a minute.
Artificial intelligence is often described as if it lives somewhere above us in a clean, weightless “cloud.” It doesn’t. It runs on physical equipment inside real buildings. Those computers consume electricity and produce heat. The facilities need cooling equipment, pumps, fans, controls, piping, maintenance, and people who understand how it all works.
Even the cloud needs supply and return.
AI may change parts of the trade. At the same time, the infrastructure supporting AI is creating more mechanical work.
Both things can be true.
The equipment is learning to watch itself
The change is already happening inside buildings.
The Department of Energy has been working on automated fault detection and diagnostics for HVAC systems. These tools use operational data to identify faults and, in some cases, narrow down their causes. DOE has also supported work that goes beyond finding problems toward automatically correcting certain control, scheduling, sensing, and operating faults.
NIST now has an AI-Optimized Building Controls project developing and testing AI-based control algorithms for HVAC operation.
This is not a robot walking into a basement with a recovery machine. It is software watching temperatures, pressures, commands, setpoints, and equipment behavior all day long.
That could eliminate some diagnostic work. It could also expose problems that have gone unnoticed for years.
A building may eventually recognize that a valve is leaking, a sensor is drifting, or a control sequence is wasting energy before anyone complains about comfort. The service call could arrive with a stack of trend data instead of the usual description:
It just doesn’t seem right in here.
That does not make the technician useless. It changes where the technician enters the problem.
Software may identify an abnormal pattern. Somebody still has to decide whether the sensor is wrong, the programming is wrong, the actuator is stuck, the wiring is damaged, or the mechanical system simply cannot do what the screen is asking it to do.
A clean graphic cannot loosen a rusted fastener. An alarm cannot tell you everything the installer did behind the wall, and a diagnostic model only knows what its sensors can see. Field experience may matter even more when a computer is confidently pointing everyone in the wrong direction.
Robots like repetition. Service work rarely cooperates.
Robotics will probably reach the controlled parts of trade work before it handles the messy ones.
A machine can cut the same shape repeatedly. It can load material, move parts, or make the same motion all afternoon without getting tired. NIST has described collaborative robots as tools that can remove monotonous work such as picking, placing, packaging, and loading machinery while allowing skilled workers to concentrate on more valuable tasks.
That makes sense in a shop.
A service call is different.
The equipment may be thirty years old. The access panel is blocked. The previous repair left three wire nuts, two abandoned wires, and no diagram. The homeowner remembers a noise that the unit has refused to make since the truck pulled into the driveway.
The real world is full of variation.
That does not mean robotics will never handle it. It means the difficult part is not merely turning a wrench. It is recognizing what kind of problem is actually in front of you.
The International Labor Organization studied occupational exposure to generative AI and estimated that about one in four jobs worldwide has some degree of exposure. Its conclusion was not that one in four jobs would disappear. The researchers said transformation is more likely than replacement because most occupations still include tasks that require human involvement.
That feels closer to what I see.
Jobs are bundles of different tasks. Technology rarely takes the whole bundle at once. It starts with the portions that are repetitive, measurable, or easy to perform from a screen.
Being useful is no longer the whole job.
But the technology inside the work was only half of what bothered me that Labor Day afternoon.
In the trades, it is tempting to believe that good work speaks for itself. Sometimes it does. A system runs. A customer is comfortable. A building owner stops calling.
But businesses are discovered online now. Customers form opinions before anyone arrives at the house. Young people learn what a career looks like from the people willing to show it to them. Manufacturers collect equipment data. Software schedules calls, prices work, and decides which advertisements appear on which phones.
Meanwhile, an algorithm helps decide whose knowledge gets noticed and whose disappears into the feed.
Knowing the physical work remains valuable. Pretending that the digital systems around it do not matter would be naive.
I don’t need to turn into a technology expert or spend every evening trying to impress an algorithm. I do need to understand enough to recognize where the work is going.
That means learning what the equipment can measure.
It means understanding what automated diagnostics can—and cannot—tell me.
It means paying attention to robotics, controls, and the businesses that build them.
It also means learning how ideas travel online, even when doing so feels less honest than tracing a wire or putting a meter on something.
One more system to learn
Every trade has changed with its tools.
Technicians learned electronic ignition, variable-speed motors, inverter-driven equipment, digital controls, and networked systems. None of those developments made mechanical judgment worthless. They gave experienced people more things to understand—and occasionally more things to blame.
AI and robotics are the next part of that story.
I’m not worried that a robot will arrive tomorrow morning and ask for my tool bag. I am more interested in the technician standing beside me who understands the equipment, the controls, the data, and the customer better than I do.
That person will have an advantage.
So I’ll keep learning the physical systems. I’ll also keep trying to understand the invisible ones deciding how buildings operate, how businesses are found, and which ideas reach another person.
A quiet X account is a small place to start.
But every unfamiliar system looks a little less mysterious once you begin tracing where everything goes.
Research notes and sources
- The U.S. Bureau of Labor Statistics projects 11 percent HVAC employment growth from 2025 through 2035 and approximately 40,600 openings annually. It also identifies sophisticated climate-control systems and data-center construction as demand drivers.
- The International Labor Organization’s 2025 global study estimates that 24 percent of jobs have some exposure to generative AI and concludes that transformation is more likely than wholesale replacement.
- NIST’s AI-Optimized Building Controls project is developing and evaluating AI-based HVAC control algorithms.
- The Department of Energy explains how automated fault detection and diagnostics use building data to identify faults and sometimes isolate their causes. DOE has also researched automated fault correction.
- NIST describes collaborative robots as tools for shifting repetitive work away from skilled workers.
- X describes the signals used in its approach to recommendations.
No comments:
Post a Comment