The conversation about AI taking jobs is missing the point.

Ask most people where AI is destroying work and you'll hear about truckers, factory workers, warehouse staff. The intuition says: robots replace manual labour. It's a clean story. It's also the wrong story.

The actual damage is concentrated somewhere nobody wants to look directly at: white-collar, entry-level roles. Junior developers. Junior copywriters. Data analysts in their first two years. Customer success reps who are learning the job by doing it. These are the roles being quietly hollowed out right now.

And here's the part that makes it a generational problem, not a cyclical one: you can't automate experience.

The numbers nobody is discussing

Yale School of Management published data in May 2026 showing unemployment among recent graduates has climbed to nearly 6%, rising roughly twice as fast as the broader workforce since 2022. Goldman Sachs data cited in the same period shows unemployment among 20 to 30 year olds in tech-exposed roles has increased by nearly 3 percentage points since early 2025. CBS News reported in late May that AI has reduced monthly payroll growth by approximately 16,000 jobs, raising the headline unemployment rate by 0.1 percentage point.

But the headline rate is 4%. For recent graduates it's 6%. For tech-exposed early-career workers it is worse still.

These aren't displaced factory workers retraining as electricians. These are young professionals who did everything right. They got the degree, got the entry-level role, started building a career, and found the ladder pulled up behind them.

Why this is worse than it looks

The standard reassurance goes: new technology always creates more jobs than it destroys in the long run. It happened with the industrial revolution. It happened with computers. The new roles that AI creates will absorb the displaced workers.

That argument might be true over a thirty-year horizon. It is cold comfort to a 24-year-old who cannot get a first role in 2026 because no company is hiring junior anything right now.

More critically: the path to senior-level expertise runs through junior-level work. You do not learn to evaluate code by taking a course. You learn by working alongside someone who knows more than you, getting feedback on real decisions, building pattern recognition through thousands of small judgements.

If junior roles disappear, the senior talent pipeline empties in five to eight years. The AI companies promising their tools will augment human workers are inadvertently accelerating a scenario where there are no experienced humans left to be augmented.

What this means for business operators

For founders and marketing managers, this has a practical implication that is easy to miss: the human oversight layer of any AI initiative is more fragile than it looks.

Most AI transformations are built around a model and a junior analyst to check its work. The theory is sound. The practice falls apart when the person reviewing AI outputs does not have enough real-world experience to know when something looks wrong.

The companies that will win over the next decade are not the ones who automate the most aggressively. They are the ones who invest in the human judgment layer: the person who can direct the AI, catch its confident errors, and make the call when the model does not have the context to decide.

At Foundry, we see this in how we structure client work. Every engagement pairs experienced strategists with AI tooling. The AI handles volume and pattern. The human handles judgment, context, and the decisions that carry risk. That structure is not a concession to AI's limitations. It is the actual competitive moat.

What you should actually do with this

If you are building a team: stop measuring AI ROI purely in headcount saved. Measure it in what your most experienced people can now focus on versus what they were wasting time on. That reframe changes how you hire.

If you are running marketing: the content AI can produce has been commoditised. The content that requires genuine judgment. The positioning decisions, the voice that reflects a real understanding of your customer, the copy that makes a claim instead of hedging. That is more valuable than it was eighteen months ago, not less.

If you are early in your career: prioritise building judgment, not accumulating prompts. The people who can work effectively with AI systems are the ones who understand the domain deeply enough to know when the system is wrong. That understanding comes from doing the work, not from watching the system do the work.

The AI job crisis is real. The danger is not the jobs being lost today. It is the experience being prevented from forming tomorrow.