Scotty Wristen has wired houses in Abilene, Texas, for decades. He knows which neighborhoods sit on expansive clay that shifts enough to crack conduit runs if you don't leave slack at the slab penetrations. He knows that the county inspector who works Tuesdays is particular about junction box fill calculations in a way the Thursday inspector is not, and he knows this because he has been written up by one and waved through by the other on the same code section across a span of fifteen years. None of this is in a manual. It lives in Scotty's head, and in the heads of electricians like him across the country, and it is leaving the industry at a rate of roughly 20,000 workers per year.
He lost five of his trained electricians to data center work, and the replacements are teenagers. "They're new. They don't even have a set of tools," Wristen told the Texas Tribune in April. "It's usually about four or five months of hell where we have little mistakes that cost us time and money."
Those little mistakes are the visible cost. The invisible one is everything Wristen's departing workers took with them that can't be taught in five months or five weeks or, increasingly, at any pace the industry can sustain.
Quantifying the Invisible
Construction is one of the few industries where knowledge resides almost entirely in people rather than systems. A software company losing a senior engineer loses someone whose work product sits in version-controlled repositories, documented APIs, and test suites that persist after they leave. A residential builder losing a 30-year superintendent loses the person who knows that the framing crew from Guadalajara does better work when you sequence the second-floor joists before the plumber finishes the ground-floor rough-in, because the crew's foreman was trained in a building tradition where structural work takes priority over mechanical, and fighting that sequence produces friction that shows up as callbacks six months after closing.
That is tacit knowledge, the kind that cannot be written into a spec and that nobody has attempted to write into a spec, because the person who holds it does not think of it as knowledge. He thinks of it as Tuesday.
Associated Builders and Contractors reported in January that the industry needs 349,000 net new workers in 2026, rising to 456,000 in 2027. ABC's chief economist, Anirban Basu, identified a detail the headlines mostly ignored: "A majority of new worker demand in 2026 will be attributable to retirement rather than increased demand for construction services." The industry is not growing its way into a labor shortage. It is aging its way into one, and the retirements are concentrated exactly where the knowledge is deepest.
BlackRock's 2026 infrastructure report found that 70 percent of supervisors in the electrical industry are baby boomers. Nearly one in five construction workers overall is older than 55, and the median worker is 42.
Read those numbers from the bottom up. Seventy percent of the people who supervise electrical work in this country are within a decade of leaving. When they go, they will take with them not only the ability to bend conduit and pull wire but the judgment about when the blueprint is wrong, when the load calculation is conservative enough to trust and when it is not, when the soil conditions on a particular lot mean you should oversize the ground rod, and which supplier's THHN wire strips cleanly and which one frays under a standard knife pull. Those are decisions that prevent callbacks, prevent rework, prevent the $15,000 change order your homeowner should never have had to sign.
Five Weeks Does Not Replace Thirty Years
Meta launched America's Workforce Academy on June 8, pledging $115 million in its first year to train construction workers in five-week bootcamps covering electrical, mechanical, plumbing, and fiber installation, with a guaranteed job offer upon completion. Google announced $50 million in similar trades training. Microsoft partnered with North America's Building Trades Unions in April to deliver AI literacy training across apprenticeship programs nationwide.
Combined, that is more than $400 million directed at construction workforce training in the first half of 2026 alone. It is genuine money and it will produce thousands of real workers.
What it will not produce is superintendents.
A journeyman electrician in the United States requires four years of apprenticeship, including roughly 8,000 hours of on-the-job training and 576 hours of classroom instruction, according to BLS occupational data. That is the licensing minimum. The judgment that separates a competent journeyman from a superintendent who prevents catastrophic rework comes from another ten to twenty years of field experience accumulated across hundreds of projects, in different soil types, under different inspectors, with different trade crews who each carry their own institutional habits and quality signatures. Meta's five-week bootcamp produces what the company calls a "job-site ready" worker. That designation is accurate and honest, but job-site ready is the starting line, not the finish, and the distance between the two is measured in decades that no amount of capital investment can compress.
| Knowledge Type | Codifiable? | AI Capturable? | Lost at Retirement? |
|---|---|---|---|
| Building codes (IRC/IBC) | Yes | Yes | No (exists in documents) |
| Manufacturer install specs | Yes | Yes | No |
| Local inspector preferences | Partially | Not yet | Yes |
| Soil/site behavior by neighborhood | Partially | Not yet | Yes |
| Subcontractor quality patterns | Partially | Not yet | Yes |
| Trade crew sequencing judgment | No | No | Yes |
| Material brand reliability by use case | Partially | Not yet | Yes |
| When the blueprint is wrong | No | No | Yes |
Of the eight categories of knowledge a veteran superintendent carries, only two are fully codifiable and capturable by current AI systems. Both of those already exist in documents that survive the person who read them. The six categories of knowledge that actually differentiate a good build from a bad one are either partially capturable or not capturable at all, and every one of them evaporates the day their holder cleans out a truck and turns in a hard hat.
What AI Can Actually Do Right Now
Touchplan, a digital collaboration platform used by firms including Jacobsen Construction, has been trying to bridge the generational knowledge gap by putting project planning into a shared digital environment where older workers' scheduling decisions become visible and reproducible. Layne Hess, Jacobsen's corporate director of planning and scheduling, describes the core tension: "You have the 'old dogs' who still refuse to use email, and the younger ones coming out of school, brimming with technological expertise. Both groups have incredible knowledge, and they just need a better way to share and integrate their expertise."
That sounds reasonable until you watch it in practice. Hess also reported that veteran workers were "yelling at me because they were afraid of the software." Workers with thirty years of field mastery feared that a digital platform would expose a weakness they had never needed to address. The knowledge transfer tool that was supposed to capture what they know first required them to demonstrate what they do not know, and that transaction has a price most people will not pay voluntarily.
Microsoft's partnership with NABTU takes a different approach, embedding AI literacy directly into existing apprenticeship programs. Instructors use AI to generate lesson plans, quizzes, and training materials. Jobsite use cases include code compliance lookups and safety protocol updates. This is useful, incremental, and real, but it is also aimed at the explicit-knowledge half of the table above: the building codes, the manufacturer specs, the safety regulations that already exist in searchable databases. It does not attempt to capture the tacit judgment that distinguishes a veteran from a newcomer, because that judgment is contextual, local, and relational in ways that current AI architectures are not designed to encode.
Computer vision tools can detect when insulation is installed incorrectly, when a framing member is out of plumb, when a concrete pour has surface defects. Those are real capabilities with measurable value, and we have covered them extensively. But detecting a visible defect and predicting an invisible one are different problems. A camera can see that a wall is out of plumb. It cannot see that the soil under the foundation will shift three inches over the next eighteen months because the lot sits on the edge of a filled creek bed that only the superintendent who built twelve houses on this street knows about, because the geotechnical report from 2009 was wrong and nobody updated it.
Strongest Counterargument
Tacit knowledge has always been lost at retirement. Construction has survived waves of retirements before, each time accumulating fresh tacit knowledge through the apprenticeship system that replaced departing workers one generation at a time. What is different now, the counterargument goes, is not the retirement wave but the tools available to manage it: AI-powered documentation that captures superintendent decisions in real time, computer vision that records field conditions across thousands of projects and identifies patterns no individual worker could perceive across a career, and digital collaboration platforms that make implicit scheduling decisions explicit and searchable. Given ten years of adoption, these tools could build a collective institutional memory that transcends any individual, making the industry less vulnerable to retirement waves, not more.
This counterargument is architecturally sound and temporally wrong. The tools that could build a collective construction knowledge base need data from the workers who are leaving now. If adoption lags retirement, and every metric available says it does, the window for capturing what the departing generation knows is closing faster than the capture tools are being deployed. According to the June 2026 Dodge/CMiC contractor survey, two-thirds of contractors are not even aware these tools exist. The retiring workers who carry the knowledge the tools would need to learn from are walking out the door, and the tools are not in the building yet.
What This Means for Your Home
If you are building or renovating in the next five years, the person managing your project is statistically more likely to be younger, less experienced, and working without the informal mentorship network that previous generations relied on, because the mentors have retired or moved to higher-paying data center and commercial work. Your project is not doomed. Younger workers bring genuine strengths: comfort with digital tools, willingness to adopt new methods, and a lack of attachment to procedures that have outlived their usefulness.
But the gap is real, and it is widest in exactly the places where mistakes are most expensive: foundation design on difficult sites, HVAC sizing for unusual floor plans, electrical panel layout for homes with solar and EV charging and battery storage, and trade sequencing on complex renovations where the order of operations determines whether a $300,000 kitchen remodel finishes in twelve weeks or twenty.
Ask your builder how many years their superintendent has been working in your area. Ask who supervised their last ten projects and whether that person is still with the company. Ask what happens to project quality when their most experienced field manager retires. The answers will tell you more about the risk profile of your build than any reference letter or online review.
Limitations
The 48.3-million-person-years figure is a rough estimate built from ABC's 349,000 annual attrition projection, BLS median tenure data, and assumed career spans for retiring workers; it has not been independently validated and likely overstates cumulative risk by treating all attrition as retirement (some is voluntary turnover to other industries). Touchplan and Jacobsen Construction quotes are drawn from a published case study and may not represent broader industry experience with digital collaboration tools. The tacit-vs-explicit knowledge categorization in the table is the author's framework and has not been empirically tested in a construction context. BlackRock's 70-percent figure for baby boomer supervisors is specific to the electrical industry and may not generalize to other construction trades. Meta's America's Workforce Academy had not yet enrolled its first cohort at the time of publication, so its outcomes are projected, not measured. The Dodge/CMiC two-thirds awareness gap is from a survey whose full methodology has not been publicly released.