AI Is Dividing the Workforce —
And Age Is the Fault Line
New research from the Federal Reserve reveals that AI isn't simply eliminating jobs — it's reshuffling who wins and who loses, with Gen Z workers bearing the highest cost.
The headlines about AI wiping out jobs tell only half the story. Yes, white-collar roles are disappearing at an alarming rate — but new research from the Federal Reserve Bank of Dallas shows the disruption is far more targeted than a broad "job apocalypse." The real story is a generational divide, and younger workers are on the wrong side of it.
J. Scott Davis, an assistant vice president at the Dallas Fed, spent months analyzing wage and employment data stretching back to fall 2022 — right around the time large language models began reshaping the workplace in earnest. His conclusion challenges the simplest narrative about AI and jobs.
Instead, what's emerging is a tale of two workforces: experienced workers who are being amplified by AI, and entry-level workers whose skills AI can now largely replicate.
Experience You Can't Prompt Away
The key variable, Davis argues, is the type of knowledge a job requires. Entry-level workers tend to hold what economists call "codifiable knowledge" — the kind learned from textbooks, training programs, and structured curricula. It's the stuff you can write down, teach, and, as it turns out, automate.
Experienced workers, by contrast, possess something harder to pin down: judgment built through years of navigating ambiguous, real-world situations. That kind of tacit, experiential knowledge is far more difficult for AI to replicate — at least for now.
The result? In AI-exposed industries, experienced workers aren't being replaced. They're being elevated — offloading routine tasks to AI and redirecting their energy toward higher-value work. Meanwhile, the on-ramp for younger workers is narrowing.
A Global Pattern, Not a Local Glitch
The numbers bear out across borders. A February report from Ireland's Department of Finance found youth employment fell by 20% between 2023 and 2025, while employment for workers in their thirties, forties, and fifties grew by 12%. In the United States, Stanford University researchers found employment has declined 1% in the top AI-exposed sectors — law, finance, and education — since 2021, with workers aged 22 to 25 absorbing the sharpest losses while their older colleagues have seen job growth.
The tech industry is already showing what this reorganization looks like in practice. Boris Cherny, creator of Claude Code at Anthropic, recently suggested the title "software engineer" — once a reliable entry point into Big Tech — could be functionally extinct by the end of 2026. Cherny himself hasn't written code manually since November, delegating those tasks entirely to AI.
Wages Are Holding — for Now
One potentially surprising finding: wages have largely held up despite the job losses. Davis attributes this to the fact that the most AI-exposed fields also tend to have steep experience-based wage premiums. Wages in computer systems design are up 16.7% since fall 2022 — more than double the national average of 7.5%. In the top-decile of AI-exposed industries overall, wages have grown 8.5%.
The picture is bleak in roles where AI threatens both ends of the experience spectrum — fast-food cooks, ticket agents, dry cleaners. In those occupations, AI doesn't just replace the novice; it undercuts the veteran, too. Those workers are already seeing negative wage growth.
The Broken On-Ramp
Perhaps the most consequential concern in Davis's research isn't about today's workforce at all — it's about tomorrow's. The traditional model of professional development has always relied on entry-level roles as a proving ground: young workers spend years absorbing institutional knowledge, making mistakes in low-stakes situations, and gradually earning their way to more complex responsibilities.
AI is making that model economically unattractive for companies in the short term. Why staff a team of junior analysts when one senior analyst with AI tools can match — or exceed — their output?
"Firms are going to find that AI is making this method of employee development cost-ineffective," Davis wrote. But he's also clear about the longer-term problem: you can't simply skip a generation of talent development and expect experienced workers to materialize on demand in five years.
"The companies three to five years from now that are going to be the most successful are those companies that doubled down on entry-level hiring in this environment," said Nickle LaMoreaux, IBM's chief human resources officer.
Rethinking the Job Ladder
The challenge ahead isn't just economic — it's structural. Businesses, educators, and policymakers will need to rethink how people gain professional experience in a world where AI handles the tasks that once served as a training ground. Apprenticeship models, rotational programs, and deliberate mentorship may need to replace the slow accumulation of responsibility that characterized career development for prior generations.
The workers winning right now aren't winning because they're older. They're winning because they accumulated experience during a window that's now closing for those just starting out. Whether that window can be reopened — and how — may be one of the defining workforce questions of the decade.
Sources & Further Reading