Anthropic released research confirming what electricians, plumbers, and HVAC techs already know from daily experience.

Their work requires physical presence, real-time problem-solving, and human judgment AI doesn’t replicate.

The findings show a gap between what AI could theoretically automate and what gets disrupted in the real world. While language models could speed up 94% of all computer and mathematical tasks, only 33% are affected in practice.

Construction shows even sharper numbers: 15% theoretical exposure, 2% observed impact.

The physical trades remain virtually untouched.

The Jobs AI Won’t Touch

About 30% of U.S. workers score zero on AI exposure metrics. Their tasks don’t appear in AI usage data at any meaningful level.

These roles are built around physical presence, sensory judgment, and reading situations in real time. A language model has no body, no hands, and no instincts developed through years in the field.

The least exposed occupations require physical abilities AI doesn’t provide. You don’t fix a broken HVAC system through a chatbot. You don’t diagnose intermittent electrical problems without being on site. You don’t troubleshoot drainage issues without physically evaluating water flow patterns and structural layouts.

According to The Street’s analysis, these positions aren’t low-skill jobs. They’re roles requiring specialized knowledge combined with hands-on execution.

White-Collar Work Gets Hit First

Every previous automation wave targeted lower-wage workers first.

This one is different.

AI is lining up squarely against white-collar professionals who spent years and significant money building credentials for office-based careers. Electricians, plumbers, HVAC techs, and welders stay protected because they still require human hands, human judgment, and human presence.

The pattern is already visible in hiring data. Job postings for entry-level software engineer positions have dropped roughly 40% compared to pre-2022 levels. Software engineer salaries in Europe saw modest growth of 1-2% in 2025, slower than the previous year.

At the same time, the U.S. HVAC services market is projected to grow from $21.16 billion in 2025 to $29.13 billion by 2030, while the workforce ages out and apprenticeship pipelines stay thin.

The Hidden Reality in Blue-Collar Businesses

Here’s what most people miss about blue-collar companies: 70% of positions in these businesses are white-collar adjacent.

The dispatchers, schedulers, and administrative staff supporting field technicians face automation pressure. But the technicians themselves remain protected.

As The Blue Collar Recruiter points out after placing thousands of workers, this creates an interesting dynamic. The support infrastructure around trades work is vulnerable. The hands-on work is not.

AI handles scheduling, predictive diagnostics, and administrative tasks. HVAC, plumbing, and electrical contractors are early adopters getting AI into real business applications. But they’re using it to automate invoicing, call answering, and scheduling.

The technicians doing the work stay essential.

Why Skilled Trades Remain Automation-Resistant

The work requires physical presence, manual dexterity, and adaptation to unique site conditions.

Every job site is different. Every installation has variables. Human workers adapt in real time based on what they encounter.

Diagnostic work requires experience-based judgment AI doesn’t replicate. An HVAC tech troubleshooting why a system isn’t cooling evaluates dozens of variables based on sounds, smells, and system behavior. Plumbers diagnosing drainage issues read water flow patterns and structural layouts. Electricians troubleshoot intermittent problems not showing up on meters.

This judgment develops over years. AI doesn’t replicate it because the work requires physical interaction with systems in unpredictable real-world conditions.

When an electrician walks into a building and immediately identifies a problem based on a burnt smell or an unusual hum, you’re seeing pattern recognition built on thousands of hours in the field.

No language model can simulate that.

Physical Dexterity Is the Ultimate Firewall

Microsoft research confirms field-based trades rank among the safest careers right now precisely because they require fine motor skills in tight spaces and constant on-the-fly decisions.

We are likely decades away from robots replicating the level of touch and adaptability at scale.

The Tony Blair Institute’s 2024 report states it plainly: “Manual jobs in construction and skilled trades are less likely to be exposed to AI-driven time savings.” Skilled roles in plumbing, gas, electrics, and renewables stay among the least exposed to automation, thanks to hands-on expertise, problem-solving, and customer trust AI doesn’t replicate.

The data suggests physical dexterity is the firewall against automation.

Trust and Human Connection Matter

Service work is relationship-based.

Real estate managers need to trust the people in their building. They ask questions. They want to understand what’s happening. AI chatbots don’t build the rapport keeping customers calling the same company for years.

Skilled trades workers who communicate well create loyal customer bases AI doesn’t replicate. The relationship between a property manager and a reliable HVAC contractor who shows up on time, diagnoses problems accurately, and communicates clearly throughout the process creates value beyond the immediate repair.

Trust compounds over time.

Career Experts Validate the Trade Shift

Jobs in the skilled trades “are the underdog and so AI-proof,” said Monster career expert Vicki Salemi. “They require physical presence, and they are less likely to be fully automated or offshored. Many have union membership, so there is job protection.”

Applications to apprenticeship programs have jumped 70% since 2022 as younger workers recognize this.

The shift is already happening. People are making the calculation. Physical skills in unpredictable environments provide more career stability than knowledge work AI handles increasingly well.

What the Anthropic Report Actually Measures

The Anthropic research is valuable, but understanding its limitations is important.

The entire dataset captures only Claude usage, missing every other AI tool on the market. The methodology is built on text-based language model interactions naturally skewing toward computer and mathematical tasks, which account for nearly half of API traffic.

Tradespeople don’t show up in this data. Not because they’re immune to disruption, but because the technologies threatening trades have nothing to do with chatbots.

Robotics, computer vision, prefabrication automation, and IoT diagnostics represent the real potential threats to physical trades work. Drawing comfort about the trades from low Claude exposure is like concluding horses were safe from automobiles because they weren’t threatened by the telegraph.

If you want to understand what’s happening in the trades, you need entirely different data: robotics adoption rates, prefab construction trends, apprenticeship pipelines, and permit-to-completion bottlenecks.

None of which this report covers.

The Real Mechanism Driving Trades Wages

Trades wages are rising faster than inflation right now because multiple forces are converging.

Demographic cliff: A large number of people who work in the trades are retiring. The median age of construction workers is 42. The workforce is aging out faster than new workers are entering.

Immigration policy effects: Immigrant workers leaving the United States adds to labor constraints in construction and trades work.

Inflation creating cost pressure: Rising material and operational costs get passed through to wages as companies compete for limited skilled labor.

The Bureau of Labor Statistics shows job openings in construction and extraction occupations stay persistently high. A high number of unfilled positions means the trades have a structural moat against automation pressure.

You don’t automate roles you don’t even fill with humans.

Total Economic Outcome Matters More Than Salary

The real comparison people should make isn’t salary alone. It’s total economic outcome.

A software engineer starts earning at 22 after a four-year degree and potentially significant student debt. An HVAC tech starts earning at 18-19 through an apprenticeship with no debt.

By the time a CS grad starts their first job, an HVAC tech has four years of earnings and experience in the bank.

Master HVAC technicians in unionized industrial settings earn well over $100,000 annually. Techs specializing in data center cooling, commercial refrigeration, or geothermal systems push higher.

An HVAC tech who starts a business and runs crews reaches owner-operator income levels exceeding most salaried software engineers.

The trajectories are diverging.

What AI Adoption Means for Physical Work

AI disruption doesn’t make physical trades work more valuable in absolute terms. It creates relative scarcity by disrupting alternatives.

The protection is the value story.

If migration into trades accelerates as AI disrupts knowledge work, the combination could flatten wage growth in trades over time. But for now, demographic retirement, labor constraints, and growing demand creates upward pressure on wages.

The trades aren’t immune to market forces. They’re operating under different constraints than knowledge work.

The Gap Between Potential and Reality

The most important finding in the Anthropic research isn’t about trades specifically. It’s about the gap between what AI could theoretically automate and what gets disrupted in practice.

The gap represents future disruption still waiting to break.

Adoption curves for AI are accelerating as leadership pushes initiatives top-down. The areas where AI fully automates job functions will see faster adoption as organizations recognize the economic benefits.

But physical trades work stays protected by fundamental constraints: the need for hands-on presence, real-time adaptation to unpredictable environments, and judgment developed through years of field experience.

Those constraints aren’t going away.

What This Means for Career Decisions

If you’re advising someone on career paths right now, the core question isn’t about specific job titles. It’s about the nature of the work itself.

Ask: What kind of work will stay valuable as AI capabilities expand?

Ask: Are you building skills getting more valuable as AI gets better, or less?

Ask: How much of this career is physical-world or relationship-based?

The financial decision is important too. Entering the trades immediately means starting with zero debt. Choosing post-secondary education typically means delaying income and in some cases taking on debt putting you in the hole before you start earning.

The math is straightforward. One path starts at zero. The other starts negative and delays earnings into your twenties.

Most people in knowledge work don’t hit their stride until their thirties. A tradesperson has a decade of earnings, experience, and business ownership already in the bank.

The Bottom Line

Anthropic’s research confirms what field experience already showed: AI isn’t coming for the trades the way it’s disrupting knowledge work.

The physical nature of the work, the unpredictable environments, the need for real-time adaptation, and the judgment built through years of hands-on experience create fundamental barriers to automation.

Those barriers aren’t temporary. They’re structural.

The trades aren’t a fallback option. They’re a protected category of work in an economy where AI is reshaping what’s valuable from the inside out.

The people who recognize this early position themselves accordingly.