The most important AI story in the trades isn’t the one about robots taking jobs. It’s the race to save what a retiring generation knows before it walks out the door for good.
The technician who can’t be replaced
There’s a technician in your market who has been fixing the same equipment since before the manufacturer’s current support team was born. He can tell you which circuit boards run hot in July, which sensors go quiet a couple of weeks before they fail outright, and what a machine sounds like right before something lets go. Nobody wrote any of that down for him. He learned it over thirty years, on job sites that never came with a manual.
He’s retiring soon, and there’s no one behind him who can do what he does.
I work in parking. Revenue-control systems, gates, the machinery a city leans on without noticing until the morning it stops working. But the technician I just described isn’t really a parking story.
Swap a few details and he’s in a manufacturing plant, a hospital’s biomedical shop, a utility substation, an HVAC bay. Any industry that runs on physical equipment runs, underneath everything, on the people who understand it. And that group is leaving faster than anyone is replacing it. I’ve come to think this is one of the bigger AI stories of the decade, even though almost nobody frames it that way.
The numbers behind the “silent army”
The scale is easy to miss until you sit with the numbers. A JLL analysis published this year, reported by Fortune, estimates that about 2.1 million skilled-trades jobs will go unfilled by 2030, with as much as a trillion dollars in annual output at risk as a result.
Almost four in ten facilities managers are already past 55. In manufacturing and construction, roughly five workers retire for every two who come in behind them. JLL calls this workforce the “silent army,” and it is getting quieter.
Backwards from the usual AI story
The usual AI-and-work conversation points the other way, toward machines taking jobs. In the trades, that gets the problem almost exactly backward. We don’t have a surplus of workers to automate away. We have a shortage, and the people who know the most are the ones walking out first.
So the most useful thing AI might do here, at least for now, isn’t standing in for human expertise. It’s holding onto it.
What a veteran actually knows
That is harder than it sounds, and the reason isn’t obvious until you’ve watched it happen. Most of what a veteran knows was never written down, and a good chunk of it can’t be. The chemist and philosopher Michael Polanyi had a phrase for this kind of tacit knowledge: we know more than we can tell.
The real value in a senior tech isn’t the procedure in the binder. It’s knowing that one particular garage has had a grounding fault nagging the same controller since 2019, or that a certain firmware build drops its connection once the system gets busy, or that the unit in bay three has always needed coaxing rather than commanding.
One field-service study found that between a fifth and a third of the problems and fixes a service team actually relies on live only in people’s heads, recorded nowhere. Decades of manuals never caught it, because it isn’t procedure. It’s judgment.
How the loss actually shows up
None of this shows up as a headline. It shows up as service calls that run long, as the same equipment getting a repeat visit it shouldn’t need, as a building that keeps having the problem everyone thought was fixed.
A two-hour job turns into six. A workaround nobody documented finally quits, and there’s no one left who remembers why it was there. When institutional memory goes, it doesn’t make a sound. You just notice, a quarter later, that things got worse and you can’t quite say when.
A category takes shape
What’s different now is that the tools might finally be up to the problem, and a real category is forming around it. Companies such as Aquant, Augmentir, and XOi are building systems that pull from service histories, expert judgment, and recorded job footage to hand a newer tech something usable while they’re still standing at the machine.
Large manufacturers like Atlas Copco are turning years of service tickets and hallway know-how into searchable internal tools. Further afield, groups like the Internet Archive and StoryCorps have spent years making the same case in a different domain: capture what people know while you can, give it structure, and it stops being something you lose.
The methods vary a lot. The premise underneath them doesn’t.
The hard part is trust
The catch isn’t storage. It’s trust. A younger worker in front of a dead machine at one in the morning needs an answer that holds up, because a confident wrong one is worse than nothing at all.
Getting that part right — the verification, the accountability, knowing where the system’s knowledge runs out — is the actual work. It’s where the serious efforts will pull away from the demos, and I’ll spend real time on it later in this series.
The window is narrower than we admit
The industry doesn’t get a vote on whether this handoff happens. The math already settled that. The only open question is whether the knowledge a generation spent its career building gets saved in some usable form, or leaves with the people who built it.
Over the next couple of months I’ll follow this where it leads, across the industries living through it right now. It starts from a plain and slightly uncomfortable idea: we are one retirement wave away from forgetting how our own infrastructure works.
