labor-marketUpdated: March 21, 2026

AI Is Both Helping and Hurting Workers — But Not for the Reasons You Think

Five independent studies paint a paradox: AI is cutting jobs while raising wages. The real story is about who benefits, who loses, and why corporations are firing for potential, not performance.

The Paradox Nobody Expected

If you follow AI news, you have probably seen two kinds of headlines. The first: AI is destroying jobs. The second: AI is boosting productivity and wages. Both are true — and that is exactly the problem.

Five independent studies published in early 2026, from the Dallas Federal Reserve, Harvard Business Review, and the Economic Innovation Group, converge on a finding that defies simple narratives. AI is simultaneously eliminating positions and raising wages for those who remain. It is both the threat and the opportunity, depending almost entirely on where you sit in the experience ladder. (Dallas Fed, HBR, EIG, all published January–March 2026)

Here is what the data actually says, and why the most popular explanations are probably wrong.

Fewer Jobs, Higher Pay — The Dallas Fed Paradox

The Dallas Federal Reserve published two striking analyses in early 2026. Economist J. Scott Davis found that in computer systems design — one of the sectors most exposed to AI — employment fell by 5% while wages rose by 16.7% since fall 2022. (Dallas Fed, February 24, 2026)

That is not a typo. The same sector is shedding workers and paying the survivors significantly more.

The mechanism is what economists call the experience premium. Across AI-exposed occupations, workers with experience-based tacit knowledge — the kind you cannot learn from a textbook or an online course — saw their wages climb. The median experience premium across these occupations is 40%. (Dallas Fed) AI handles the routine work that junior employees used to do, making experienced workers who can direct and verify AI output dramatically more valuable.

For software developers and financial analysts, this creates a stark divide. Senior developers who can architect systems and review AI-generated code are in higher demand than ever. But the entry-level pipeline — the path that creates future senior developers — is narrowing.

The Youth Employment Signal

A separate Dallas Fed study from January 2026, by Tyler Atkinson and Shane Yamco, quantified the damage to young workers. The share of 22-to-25-year-olds employed in AI-highly-exposed occupations dropped from 16.4% to 15.5% of total employment. (Dallas Fed, January 6, 2026)

Critically, this is not primarily about layoffs. Most of the decline comes from reduced new hiring — companies are simply not bringing in as many junior workers. (Dallas Fed) The overall unemployment rate impact remains small, about 0.1 percentage points. (Dallas Fed) Young workers are not being fired en masse; they are being quietly excluded from the pipeline.

For roles like administrative assistants and customer service representatives, where AI tools now handle a growing share of routine tasks, the junior positions that used to serve as entry points are the first to disappear.

Wait — Is This Actually About AI?

Here is where the story gets complicated. The Economic Innovation Group (EIG), in a January 2026 paper by Google economists Zanna Iscenko and Fabien Curto Millet, provides a sharp counterargument.

Their data shows that entry-level hiring in tech-adjacent fields began declining in March 2022 — a full eight months before ChatGPT launched. (EIG, January 14, 2026)

Only 12% of companies were actually using AI at a meaningful scale by mid-2025. (EIG) The EIG researchers argue that the post-pandemic tech hiring bubble, inflated by near-zero interest rates, was already deflating before AI became the convenient explanation.

This does not mean AI has zero effect. But it does mean separating AI's impact from the interest rate cycle, post-pandemic normalization, and a cooling tech sector is far harder than most headlines suggest. (EIG)

Firing for Potential, Not Performance

If AI's actual deployment is still limited, why are so many companies cutting workers in its name? Thomas Davenport and Laks Srinivasan answered this question in a January 2026 Harvard Business Review article that documented a troubling pattern.

About 60% of large companies have undertaken AI-related workforce reductions. (HBR, January 29, 2026) But only 2% of those reductions were based on actual AI implementation and measured performance gains. The rest were based on expectations — what AI might do, not what it has done.

The most instructive case is Klarna. The Swedish fintech cut 40% of its workforce citing AI capabilities, then saw customer satisfaction scores decline, and began quietly rehiring. (HBR) Gartner's research reinforces the point: of every 50 AI investments companies make, only 1 delivers transformative value. (HBR, citing Gartner)

For marketing managers and other roles targeted by anticipatory layoffs, this creates a peculiar situation: you may lose your job not because AI can do it, but because your CEO believes AI will eventually be able to.

The Occupation Shift Map

The broadest view comes from Harvard Business School researchers Suraj Srinivasan, Wilbur Chen, and Saleh Zakerinia, who analyzed job postings across 900+ occupations and 19,000+ tasks from 2019 to 2025.

Their finding: job postings in occupations vulnerable to AI automation fell by 13% over this period. Meanwhile, occupations positioned for AI augmentation — where AI makes workers more productive rather than replacing them — saw postings grow by 20%. (HBR, March 4, 2026)

This is perhaps the most actionable finding. The labor market is not uniformly shrinking or growing. It is tilting — away from roles where AI substitutes for human work, and toward roles where AI complements it. The question for any worker is not "will AI take my job?" but "does AI substitute for what I do, or does it make what I do more valuable?"

Explore how AI affects your role: Software Developers, Financial Analysts, Customer Service Representatives, Administrative Assistants, Marketing Managers.

What This Means for Your Career

The five studies together suggest three concrete takeaways.

First, experience is your moat. The 40% experience premium the Dallas Fed found is not going away. If you are early in your career, the priority is accumulating the kind of judgment and domain knowledge that AI cannot replicate — and that means finding employers who still invest in developing junior talent, even if there are fewer of them.

Second, watch what companies do, not what they say. The gap between AI rhetoric and AI reality is enormous. A company announcing AI-driven layoffs may be making a strategic bet — or it may be making Klarna's mistake. The 2%-versus-60% gap between actual AI implementation and AI-motivated cuts should make every worker skeptical of "AI transformation" announcements.

Third, position yourself on the augmentation side. The 13% decline in automation-vulnerable postings versus the 20% growth in augmentation-friendly roles is the clearest market signal. Roles that involve directing, evaluating, and improving AI outputs are growing. Roles that involve doing what AI already does adequately are not.

Sources

Update History

  • 2026-03-21: Added source links and ## Sources section
  • 2026-03-18: Initial publication synthesizing five independent sources: Dallas Fed (Feb & Jan 2026), HBR (Mar & Jan 2026), EIG (Jan 2026)

This article was researched and written with AI assistance using Claude (Anthropic). Analysis synthesizes findings from five independent sources: Dallas Federal Reserve (February and January 2026), Harvard Business Review (March and January 2026), and the Economic Innovation Group (January 2026). This is AI-generated analysis of publicly available research and should not be taken as professional career or employment advice. We encourage readers to consult the original sources linked throughout this article.

Related Occupations

software developerscustomer service representativesfinancial analystsadministrative assistantsmarketing managers

Tags

#ai-jobs#labor-market#wages#entry-level#automation
Back to Blog