There’s something deeply ironic happening in the American workforce right now. People are so afraid of AI taking their jobs that they’re going back to school to reskill — into the very fields AI is displacing fastest.
A new survey from Eastern Washington University found that 52% of American adults aged 25 and older are considering returning to school for reskilling. Of those, 21% cite AI job threats as their primary motivator, and another 31% say AI fears have heightened their interest.
The instinct is sound: adapt or die. The execution? That’s where it gets complicated.
The Reskilling Stampede
The data shows a clear pattern. Workers in healthcare, hospitality, administration, and traditional white-collar roles are looking at AI’s progress and deciding they need new skills. Fair enough — some of these roles really are at risk.
But here’s the trap: where are they reskilling to?
Programming. Data science. Digital media. These are the “safe” careers that career counselors and bootcamp ads have been pushing for a decade. They pay well. They’re future-proof. They’re tech.
Except they’re not. Not anymore.
The Programming Squeeze
The 2026 computer science graduate market is brutal. Placement rates at top programs have crashed from 90%+ pre-2025 to as low as 11-16%. Starting salaries that once hit $180,000 have halved to around $90,000. AI handles entry-level coding now, and junior hires are frozen across the industry.
If you’re a 35-year-old accountant who just spent $30,000 on a coding bootcamp to “future-proof” your career, you’ve walked into a shrinking market competing against 22-year-olds who’ll work for less and AI tools that’ll work for nothing.
This isn’t speculation. It’s the current reality. Entry-level tech jobs are evaporating while mid-career and specialized roles remain relatively stable. The reskilling pipeline is pushing people into the part of the job market that’s contracting fastest.
Data Science: Same Story, Different Syntax
Data science has been sold as the ultimate AI-proof career — after all, someone has to build and tune the models, right? But the field is bifurcating sharply.
The “data science” that’s growing — ML engineering, data infrastructure, model deployment, AI safety — requires deep mathematical foundations and systems thinking. It’s not something you pick up in a six-month bootcamp.
The “data science” that’s accessible to career-changers — basic analysis, visualization, dashboarding — is exactly what AI tools are automating. When ChatGPT can generate a Tableau dashboard from a natural language prompt, the person who only knows how to use Tableau is in trouble.
The Real Skills Gap
So if programming and data science are traps, where should people actually be reskilling? The answer is less about specific technical tools and more about capabilities AI struggles with:
-
Physical trades — Electricians, plumbers, HVAC technicians. Workers are already pivoting here, and demand is growing as AI data centers need maintenance and infrastructure.
-
AI infrastructure — Not building models, but deploying and maintaining the physical and digital infrastructure they run on. Think cloud engineering, DevOps, data center operations.
-
Human-facing roles — Therapy, social work, nursing, education. Jobs that require empathy, physical presence, and complex human judgment are among the most AI-resistant.
-
Specialized domain expertise — People who understand a field deeply enough to evaluate AI outputs, not just generate them. A lawyer who can spot the hallucination. A doctor who knows when the AI is wrong. An engineer who can verify the model’s design.
-
AI itself — Not “learning to code” in the generic sense, but understanding how AI systems work, their limitations, and how to work alongside them productively.
The NZ Angle
New Zealand has its own version of this problem. Our universities are still pushing traditional CS and data science programs as the path to the future, while our tech sector contracts and entry-level roles dry up.
The government’s AI workforce initiatives tend to focus on basic digital literacy rather than the deeper skills that actually provide a moat against automation. Teaching someone to use ChatGPT is not the same as teaching them to think critically about when ChatGPT is wrong.
For NZ workers specifically, the trades route may offer the best risk-adjusted return. Our construction and infrastructure sectors are growing, physical work can’t be offshored or automated yet, and the training infrastructure already exists through apprenticeships and polytechnics.
The Bigger Question
The reskilling panic reveals something uncomfortable about how we think about career adaptation. The assumption is that the solution to technological displacement is more technology training. Learn to code. Learn data. Learn AI.
But what if the real adaptation is learning to do things that technology can’t do? What if the most future-proof skill isn’t Python fluency, but the judgment to know when a human needs to be in the loop?
The Eastern Washington University study shows that workers know they need to change. That awareness is valuable. But awareness without direction is just anxiety with a tuition bill.
🔍 THE BOTTOM LINE
Over half of American adults want to reskill because of AI fears, and many are sprinting toward the very fields AI is devouring first. The cruelest joke of the AI era isn’t that machines are taking our jobs — it’s that we’re reskilling ourselves straight into their mouths. The real future-proof skill isn’t learning to code. It’s learning what code can’t do.
