Why the Labor Shortage Is Really a Knowledge Shortage

The real scarcity in the trades isn’t headcount. It’s the undocumented judgment that takes decades to build — and can’t be hired. 

I got a call at half past midnight. A barrier gate at a busy venue site had stopped responding to every command — manual switch included. The client said they had checked all the terminations. There was a late event letting out and they needed it fixed immediately. 

When I arrived and looked at the control panel, the answer was visible from across the room. The Phoenix connectors — the plug-in terminals that carry communication to the barrier — were not fully seated. Not unplugged, but not home either. I pushed them in, tested the gate, and was back in my vehicle in under ten minutes. The gate worked perfectly. It had been working perfectly the whole time, waiting for someone to see what was in front of them. 

The emergency minimums still applied, so that ten-minute fix was not cheap. But the expense was not really about the late hour or the drive. It was about the gap between what was visible and what the people on site were equipped to recognize. That gap is not a staffing problem. It is a knowledge problem — and it is far more common than the industry wants to admit. 

Two problems wearing the same name 

When someone in operations talks about a labor shortage, they usually mean they cannot find enough bodies. Not enough applicants, not enough people who can pass a background check and show up reliably. That is a real problem. But it is not the same problem as not having enough people who can look at a half-seated connector at midnight and know what it means. 

These two problems — too few workers and too little transferable expertise — get conflated constantly, and the conflation is expensive. Hiring more people does not solve the second problem. It just distributes it across a larger workforce. You end up with more technicians who do not know what they do not know, dispatched to jobs they cannot fully handle, running up service costs and downtime that nobody connects back to the real cause. 

The thing that cannot be hired 

Research from field-service analytics firm Emerj found that somewhere between 20 and 30 percent of the real solutions to real problems in complex equipment service environments exist only in experts’ heads — never written down, never transferred, never findable by the person who needs them. That estimate tracks with what I see. The Phoenix connector story is a mild version. The severe version is a facility that cycles through the same failure for months because the one person who understood the underlying cause retired two years ago. 

This is what makes the knowledge version of the shortage so much harder than the headcount version. A body can be hired. Judgment cannot. Pattern recognition built across decades of real failures cannot. The ability to walk into an unfamiliar site and read it — to notice what is slightly off before it becomes completely wrong — that comes from accumulated experience and does not transfer by job listing. 

Why this is an AI story 

For most of industrial history, there was no mechanism for capturing tacit knowledge at scale. You could document procedures. You could write manuals. But the judgment behind the procedures, the contextual reasoning a veteran applies before they even reach for a tool — that never had a vessel. It lived in people, and it left with them. 

That is changing. Large language models trained on the right data, combined with structured knowledge capture from working experts, can begin to hold and transmit something that approximates that judgment. Not perfectly. Not without verification. But meaningfully. The question is whether the industry moves fast enough to capture what its most experienced people know before those people are gone. The window is not wide. The workforce retiring right now built its expertise in an era when you learned by doing, on job sites, for decades. The generation coming in is walking into a world where that kind of learning time may not exist. Something has to bridge the gap — and for the first time, something actually can. 

What comes next in this series 

The next installment goes inside the problem: what is actually in an expert’s head, why it is so hard to get out, and why previous attempts at knowledge capture mostly failed. Part 3: What’s Actually in an Expert’s Head.