The jobs least likely to be replaced by AI are ones requiring physical presence in unpredictable environments: skilled trades, emergency services, and hands-on healthcare. A 2026 study analyzing more than 55 manual professions found emergency services averaging 11% automation risk and electricians around 15%, compared to 80 percent or higher for retail, production, and agricultural work. The pattern is consistent: AI automates structured, repeatable tasks well and struggles with constant physical variation, which is exactly what a job site, a patient, or an emergency call provides every time.
What the data actually says
A study released in April 2026 by construction scheduling platform Planera analyzed automation risk across more than 55 physical and manual professions, using U.S. Bureau of Labor Statistics employment data. It found sharp divergence by sector. Emergency services averaged 11 percent automation risk, social services 12 percent, and healthcare 16 percent. Agriculture sat at the opposite extreme, at 89 percent, followed by production at 82 percent, utilities at 81 percent, and retail at 80 percent.
Within skilled trades specifically, separate automation-probability research puts electricians around 15 percent risk and plumbers around 8 percent, figures that get cited repeatedly across labor market analyses because they hold up against how the work actually functions. Every panel, every crawlspace, every fault behind a wall is a slightly different problem, and that variability is precisely what current AI and robotics handle poorly.
Why physical trades resist automation
The pattern isn't about the trades being old-fashioned or protected by regulation. It's structural. AI systems are trained on large volumes of similar, repeatable examples, which makes them strong at tasks like data entry, document review, and routine customer service. A construction site, an HVAC system built in 1987, and a house with three previous unpermitted renovations don't offer that kind of repeatability. A worker diagnosing why a system won't cool evaluates sound, smell, and behavior in combination, in real time, in a space that was never designed to be standardized.
This is also why the disruption sequence has run in the opposite direction from what many expected. Office and knowledge work, long assumed to be the safe, stable category, got hit first: language-based AI reached usable quality for writing, analysis, and code between 2022 and 2025. Physical labor is only now entering its own disruption phase, through embodied AI and early-stage humanoid robotics, and remains years behind office automation in practical deployment.
The trades with the lowest measured risk
Electricians sit near 15 percent automation risk. The BLS projects 9 percent employment growth through 2034 with roughly 81,000 annual openings, driven in part by AI data center construction, EV charging infrastructure, and grid modernization, a demand story covered in more detail in our power lineman guide.
Plumbers measure lower still, around 8 percent risk, reflecting work that is almost entirely site-specific and physically inaccessible to current robotics.
HVAC technicians follow a similar pattern, and demand is compounding as AI infrastructure itself requires industrial-scale cooling, a dynamic explored in the cooling problem behind AI's biggest bottleneck.
Structural ironworkers and power linemen add a further layer of resistance: the work happens at height, outdoors, and often in emergency conditions like storm response, environments where deploying robotics is both technically and economically impractical for the foreseeable future.
Welders, particularly in specialty certification work like pipeline and structural inspection, combine physical unpredictability with a liability requirement, a licensed human has to be accountable for the weld, which keeps a human in the loop regardless of how good welding automation becomes in controlled factory settings.
What doesn't survive
The honest caveat matters here. The field technician role is low risk. The office layer built around trade businesses is not. Dispatching, scheduling, basic bookkeeping, and routine customer service inside trade companies face the same automation pressure as those same functions anywhere else. The protection in this data isn't about working for a blue-collar company. It's specifically about doing the physical work itself.
It's also worth resisting the framing that any category is fully AI-proof forever. The OECD estimates that 27 to 28 percent of roles across the broader economy carry high automation risk today, and McKinsey Global Institute has estimated that up to 30 percent of hours worked could be automated by 2030. Those are economy-wide averages, and skilled trades sit well below them, but "low risk" is a real, measured, current advantage, not a permanent guarantee.
The AI paradox
The clearest irony in this data is that the technology driving automation anxiety is simultaneously one of the biggest sources of new trade demand. AI data centers require enormous electrical capacity, industrial HVAC and liquid cooling systems, and constant physical maintenance, all of which require electricians, HVAC technicians, and data center facility technicians to build and keep running. The industry most associated with job displacement is, at the same time, hiring physical trades workers to make its own infrastructure function.
- What jobs are safest from AI automation?
- Research consistently ranks emergency services, skilled trades, and hands-on healthcare roles as the lowest automation risk categories. A 2026 study of manual professions found emergency services averaging 11% automation risk and electricians around 15%, compared to 80% or higher for retail, production, and agricultural roles.
- Why are skilled trades more resistant to AI than office jobs?
- Skilled trades require physical presence in unpredictable environments, where every job site, building, and repair situation is different. AI and robotics perform well in structured, repeatable settings but struggle with the constant variation of real-world physical work, which is part of why office and administrative tasks have been automated faster than trade work.
- Are trade jobs completely immune to AI?
- No. The field technician role itself is low risk, but the office functions that support trade businesses, including dispatching, scheduling, and basic bookkeeping, are being automated quickly. The safest position is doing the physical work directly, not working in the administrative layer around it.
- Does AI growth actually create demand for trade workers?
- Yes, in a direct and measurable way. AI data centers require enormous electrical capacity and industrial cooling systems, which has increased demand for electricians, HVAC technicians, and power line workers. The technology driving automation fears is, at the same time, a source of new trade work.