Which Jobs Are AI Proof? New Federal Projections Point Away From Tech

Which Jobs Are AI Proof? New Federal Projections Point Away From Tech

Federal statisticians released their new ten-year employment projections on 27 August 2026, and the list of fastest-growing American occupations contains almost nothing people expect.

Six of the ten fastest-growing jobs fall under healthcare and social assistance. The single fastest is nurse practitioner, projected to grow 41 percent through 2035. Second place goes to solar photovoltaic installer at 36.5 percent. Wind turbine technician sits fourth at 29.5 percent.

None of these are the jobs that dominate conversations about artificial intelligence. All of them share a characteristic that turns out to matter more than any technology forecast: someone has to physically be there.

That is the real answer to which jobs are AI proof, and it is more useful than the speculation that usually fills this subject. The rest of this piece covers what the projections actually say, why the pattern holds, where it breaks down, and how to judge your own position against it.

What the new federal projections actually show

The headline number is sobering before it is encouraging. The US economy is projected to add 5.9 million jobs between 2025 and 2035, taking total employment from 170.3 million to 176.2 million. That is 3.5 percent growth over a decade, against 10.9 percent in the ten years before it.

So the overall picture is slow. What matters is where the growth concentrates.

Nurse practitioners lead in percentage terms at 41 percent, roughly 137,800 new positions. Medical and health services managers lead in raw numbers, with 155,100 jobs projected. Data scientists rank third at 34.6 percent, or 95,400 jobs.

Behind the individual occupations sits a broader pattern. Healthcare support occupations are projected to grow 13.3 percent as a group, the fastest of all 22 major occupational groups. Healthcare practitioners and technical occupations follow at 8.0 percent. Community and social service occupations also rank in the top five.

One industry figure stands out above everything else. Services for the elderly and persons with disabilities is projected to add 625,400 jobs, more than any other detailed industry in the economy.

Why healthcare dominates the list

This is demographics rather than technology, which is precisely why it is more predictable than most forecasts.

The US population is ageing, and the likelihood of chronic conditions rises with age. That drives demand across clinical care, home-based support, behavioural health and the administrative layer that manages all of it. Emily Krutsch, who leads the Division of Employment Projections at the BLS, has pointed to the combination of an ageing population and high rates of chronic and behavioural health conditions as the underlying driver.

The effect is already visible in monthly data rather than only in ten-year forecasts. Healthcare has been the leading sector for job creation in the recent US expansion while other sectors lagged.

There is a second factor inside healthcare that rarely gets attention. The workforce itself is ageing. A significant share of registered nurses are approaching retirement, which means replacement demand sits on top of growth demand. Those openings do not show up in a growth percentage, but they are real vacancies.

The energy jobs nobody was talking about

Two of the top four fastest-growing occupations have nothing to do with healthcare or software.

Solar photovoltaic installers are projected to grow 36.5 percent, adding about 11,300 jobs. Wind turbine technicians follow at 29.5 percent, or roughly 3,500 jobs.

The percentages are dramatic and the absolute numbers are small, which is worth stating plainly rather than burying. These are not large occupations. But the growth rate reflects something structural: the BLS has noted that demand for electricity is expected to grow significantly over the projection decade, and data centre expansion is part of why.

There is an irony in that worth sitting with. The infrastructure build-out driven partly by artificial intelligence is creating physical jobs that artificial intelligence cannot perform. Someone has to mount the panel and climb the turbine.

These roles also share a useful feature for career changers. Neither typically requires a four-year degree. Certification and apprenticeship pathways are the norm.

Data scientists are on the list too, and that matters

Third place goes to data scientists at 34.6 percent growth. This complicates any simple story about technology jobs disappearing.

The distinction is not between tech and non-tech. It is between building these systems and performing the tasks these systems now handle.

Stanford Digital Economy Lab research found something specific about how this plays out. When AI automates tasks, such as writing routine code or handling customer chats, entry-level hiring declines. When AI augments tasks, such as supporting problem-solving or checking accuracy, employment holds steady or rises. Two companies adopting identical tools can produce opposite hiring outcomes depending on which they do.

Cybersecurity follows the same logic. Analyst roles are projected to grow strongly because more automated systems mean more attack surface, not less. The technology creates the demand.

So the useful question is not whether your field involves computers. It is whether your daily work consists of tasks these systems now perform, or tasks they create more of.

What "AI proof" actually means, and what it does not

This phrase deserves more precision than it usually gets, because the alternative is false comfort.

The Bureau of Labor Statistics is explicit on this point: exposure to AI indicates the potential for these systems to assist with or perform occupational tasks. It is not a prediction that an occupation will disappear. The International Labour Organization reaches the same conclusion in its 2025 update on generative AI and jobs, finding that task transformation is generally more likely than complete occupational redundancy.

The ILO has also cautioned specifically against treating exposure indicators as direct forecasts of job losses, which is exactly what most coverage does.

No occupation is permanently insulated. What the evidence supports is that some roles carry far lower exposure than others, and that the difference is measurable.

A Stanford analysis of payroll data from millions of US workers found employment for 22 to 25 year olds in the most AI-exposed roles fell roughly 13 percent since late 2022, and about 20 percent for software developers in that age band. Older workers in the same occupations grew 6 to 9 percent. Nursing aides and similar low-exposure roles kept growing throughout.

A February 2026 Harvard study covering 62 million workers found junior employment dropped 9 to 10 percent at companies adopting generative AI, while senior employment stayed stable.

The pattern is consistent across both datasets. Exposure is concentrated, measurable and heavily weighted toward early-career work in specific task categories.

The three things that predict low exposure

Reading across the projections and the research, three characteristics separate resistant work from exposed work.

Physical presence in unpredictable environments. A nurse practitioner examining a patient, an installer on a roof, a technician inside a turbine nacelle. Robotics handles structured environments well and unstructured ones poorly, and that gap has closed far more slowly than software capability.

Regulated accountability. Someone must carry legal and professional responsibility for the outcome. A system can draft a care plan, but a licensed clinician signs it and answers for it. This is why regulated professions consistently show lower exposure even when significant portions of their work involve documents and analysis.

Relationship and trust. Work where the human relationship is the product rather than a delivery mechanism. Social services, senior care, counselling, complex negotiation. The presence of the person is the service.

Most occupations contain some mix of these alongside automatable tasks. That mix is what determines exposure, not the job title.

How to test your own position

The projections describe occupations. Your exposure depends on your specific tasks, and that is something you can assess directly.

Take a normal working week and divide the hours into two buckets.

Bucket one: repetitive digital work. Moving data between systems, producing standard reports, drafting routine documents, first-pass research, handling predictable inquiries.

Bucket two: everything requiring physical presence, judgment under uncertainty, regulated accountability, or a relationship someone would notice the absence of.

The ratio between those two buckets predicts your exposure better than your job title does. A paralegal spending most of their week on standardised document review sits in a different position from one managing client relationships and case strategy, despite sharing a title.

This exercise also tells you what to change. Shifting hours from bucket one to bucket two is the practical version of career protection.

Where the jobs are shrinking

The projections name declines as clearly as growth, which is useful information rarely given prominence.

Retail trade employment is projected to decline 0.2 percent over the decade. Federal government employment is projected to decline 3.4 percent.

These are modest percentage declines across very large bases, so they still represent substantial ongoing hiring through replacement. Krutsch has made this point directly: an occupation projected to decline does not mean opportunities vanish, because openings still arise as people leave.

The more immediate pressure sits elsewhere. Announced tech layoffs ran roughly 83 percent higher year over year through June 2026, and Challenger, Gray & Christmas counted nearly 50,000 job cuts linked to AI in 2026, about 17 percent of roughly 300,000 total announced cuts.

That 17 percent figure is worth holding onto. Most job cuts in 2026 were not attributed to AI, even in a year when AI led the stated reasons for several consecutive months.

The caveat on what employers actually say

There is a credibility problem in the layoff attribution data that deserves stating.

A Harvard Business Review survey of 1,006 executives published in January 2026 found 39 percent had cut headcount in anticipation of AI capabilities, against only 2 percent who did so based on actual automation results.

Anticipation is not measurement. Some of what gets announced as AI-driven restructuring is ordinary cost cutting with a more modern justification attached. The job losses are real either way, but the causal story shapes how people plan their careers, and it is being overstated.

There is a cost to getting this wrong that employers are discovering. Forrester's 2026 Future of Work report found one-third of employers who reversed AI-driven layoffs spent more on restaffing than they originally saved, once recruiting, onboarding and higher salary expectations for returning staff were counted.

The entry-level problem that cuts across every sector

This is the part of the picture the occupational projections do not capture, and it affects people choosing a path right now more than any growth rate does.

Entry-level professional postings have fallen 29 percent since January 2024. The decline is sharpest in exactly the fields where AI adoption is fastest.

Junior roles are not disappearing so much as changing shape. The entry bar has risen. Companies still hire juniors, but increasingly expect output that previously required two or three years of experience to produce.

MIT's Andrew McAfee has warned that aggressively automating entry-level roles can backfire by cutting the apprenticeship pipeline that produces senior talent. An organisation that stops hiring juniors this year has nobody to promote in four.

This matters for the AI-proof question in a specific way. The occupations on the BLS growth list mostly have structured entry pathways that cannot be compressed. You cannot shortcut a nursing qualification or an electrical apprenticeship with better tooling. That regulatory friction, usually described as a barrier, currently functions as protection.

What this means if you are choosing a path now

Several things follow from the evidence rather than from speculation.

The growth is in physical, licensed and relational work. Six of the ten fastest-growing occupations are in healthcare and social assistance. Two more are in energy installation and maintenance. That is eight of ten in work requiring someone to be somewhere.

Credentials matter more than they did, not less. Where certification gates entry, supply stays constrained. Nurse practitioner, physician assistant, licensed electrician and solar installer all share this feature.

Being AI-adjacent works as well as being AI-distant. Data scientists at 34.6 percent growth and cybersecurity analysts around 29 percent both sit close to the technology rather than away from it. Building and securing these systems is growing work.

Tool fluency is now a differentiator within occupations, not just between them. Developers who work effectively with these tools command a premium, while entry-level pay for those who do not has stayed flat. The same split is emerging in analysis, design and writing roles.

Treat projections as direction, not destination. These are ten-year models built on current assumptions. The World Economic Forum projects 170 million new roles against 92 million displaced by 2030. Boston Consulting Group projects up to 15 percent of US jobs eliminated over five years. Both are projections, and they disagree substantially. The payroll data measuring what has already happened deserves more weight than either.

The honest limits of this picture

Anyone offering certainty about a ten-year horizon is selling something.

The BLS projections assume current technology trajectories continue at current rates. If capability accelerates sharply in physical robotics, the physical-presence advantage narrows. If it stalls in language tasks, some currently exposed roles recover.

What the data supports today is narrower and more reliable than any forecast: the measured employment effects so far are concentrated among early-career workers in specific task categories, they depend heavily on whether a company automates or augments, and the occupations projected to grow fastest are those where a person has to be physically present and professionally accountable.

For anyone weighing a direction, the fastest-growing job in America is nurse practitioner, the second is solar installer, and the fourth is wind turbine technician. That is what the federal data says, and it is a better starting point than the forecasts.