The Bottom Line Up Front
Anthropic just published one of the most important studies on AI's real-world impact on jobs—and the findings should make every aspiring tradesperson feel very good about their career choice. While computer programmers face 74.5% AI task coverage and customer service reps sit at 70.1%, skilled trades like construction, installation & repair, and grounds maintenance show some of the lowest AI exposure of any occupational category—as low as 3% observed coverage.
This isn't speculation. It's measured from millions of real AI interactions, cross-referenced with Bureau of Labor Statistics employment projections through 2034. The data is clear: the more AI-exposed a job is, the slower its projected growth. And trades? They're projected to keep growing.
What Anthropic Actually Measured
Most studies on AI and jobs ask a theoretical question: "Could an AI do this task?" Anthropic's research goes further. They created a new metric called "Observed Exposure" that combines two things:
- Theoretical capability — Can an LLM (large language model) theoretically speed up this task by at least 2x?
- Actual usage data — Is anyone actually using AI to automate this task in professional settings?
The gap between these two measures turns out to be enormous. AI could theoretically cover 96% of Computer & Math tasks, but in practice, only 32% are actually being automated. For most occupations, real-world AI adoption is a fraction of what's theoretically possible.
Data visualization by Peter Walker, based on Anthropic's research. The blue bars show what AI could theoretically do; the red bars show what it's actually doing. Notice how trades at the bottom have almost no red.
The Radar Chart That Says It All
Anthropic's original research included a radar chart (Figure 2) showing theoretical capability versus observed usage across all occupational categories. The visual is striking—and it tells the story of two very different labor markets emerging.
The blue area represents what AI could theoretically do. The red area shows what it's actually doing. For knowledge work categories like Computer & Math, Legal, and Business & Finance, the blue area is massive—meaning these jobs have enormous theoretical exposure. But look at the bottom of the chart: Construction, Installation & Repair, Agriculture, and Production barely register. These are the trades.
Even where theoretical capability exists in trades-adjacent fields like Architecture & Engineering (82%), the actual observed usage is just 18%. The physical, hands-on nature of trade work creates a natural moat against AI automation.
The Most Exposed Jobs—None of Them Are Trades
Anthropic identified the top 10 most AI-exposed occupations based on their observed coverage metric. Every single one is a desk job:
Computer programmers top the list at 74.5% observed exposure. Their leading automated task? "Write, update, and maintain software programs"—literally the core of their job. Customer service reps follow at 70.1%, with AI increasingly handling customer interactions via API integrations.
Meanwhile, 30% of all workers have zero AI coverage. This group includes cooks, motorcycle mechanics, lifeguards, bartenders, and other hands-on roles. The research explicitly notes that many tasks "remain beyond AI's reach—from physical agricultural work like pruning trees and operating farm machinery to legal tasks like representing clients in court."
What the BLS Growth Projections Confirm
Here's where the data gets even more compelling for trades workers. Anthropic cross-referenced their exposure metrics with Bureau of Labor Statistics employment projections for 2024–2034. The finding?
For every 10 percentage point increase in AI coverage, projected job growth drops by 0.6 percentage points.
The research specifically highlights that electricians appear as one of the highest-growth occupations on their scatterplot, sitting firmly in the low-exposure, high-growth quadrant. Software developers, despite also showing growth, carry significantly higher AI exposure risk.
This validates what we've been saying at CraftPATH: trades aren't just surviving the AI revolution—they're positioned to thrive because of it.
Who's Most at Risk? Not Who You'd Think
The demographics of AI-exposed workers paint a surprising picture. Compared to workers with zero AI exposure, those in the most exposed occupations are:
- 47% higher paid on average ($32.69/hr vs. $22.23/hr)
- More educated — 17.4% hold graduate degrees vs. just 4.5% in unexposed jobs
- More likely to be female — 54.4% vs. 38.8%
- Less likely to be union members — 5.3% vs. 11.7%
In other words, the workers most at risk from AI are highly educated, well-paid knowledge workers—not tradespeople. The union membership gap is particularly notable: trade unions provide an additional layer of job security that many AI-exposed office workers don't have.
Young Workers: An Early Warning Signal
One finding from the research deserves special attention. While overall unemployment hasn't meaningfully changed for AI-exposed workers yet, there's tentative evidence that hiring of young workers (ages 22-25) has slowed in exposed occupations. Job finding rates in AI-exposed occupations dropped roughly 14% compared to pre-ChatGPT levels, while entry into less exposed occupations remained stable.
This is a critical signal for young people choosing career paths right now. If you're 18 and deciding between a computer science degree and an electrical apprenticeship, the data suggests the apprenticeship may offer more reliable employment prospects over the next decade.
What This Means for Your Career Decision
Anthropic's research confirms a pattern we've been tracking across multiple data sources: the AI economy is creating a tale of two labor markets.
Market 1: Knowledge work — High theoretical exposure, growing real-world automation, slowing hiring for young workers, and increasing uncertainty about long-term career stability.
Market 2: Skilled trades — Minimal AI exposure (3-6% observed coverage), strong projected growth through 2034, rising wages driven by labor shortages, and physical work that AI fundamentally cannot perform.
The choice has never been clearer. If you want a career that's resilient to AI disruption, offers strong earning potential, and has a decade of projected growth ahead of it, skilled trades are the strategic choice.
And here's the irony: the very AI revolution that threatens knowledge work is actually fueling demand for trades. Every data center needs electricians. Every server farm needs HVAC technicians. Every piece of AI infrastructure requires physical construction. The machines that automate office work still need human hands to build, install, maintain, and repair them.
Take the Next Step
Ready to explore a career that AI can't automate? Here's how to get started:
Source: "Labor market impacts of AI: A new measure and early evidence" by Maxim Massenkoff and Peter McCrory, Anthropic, March 5, 2026. Read the full paper →