The White-Collar Moat Is Leaking: Entry-Level Hiring Down 49% Across Five Industries

Think about the deal your parents probably sold you. Study hard. Get into a good school. Land a stable job. Work your way up. That was the contract.

In May 2026, that contract started looking a lot more fragile.

An Investment Bank Analyst’s Job Just Became a Free App

Anthropic open-sourced a toolkit that does the entire workflow of a Wall Street analyst. Comparable company analysis. Industry reports. Reconciliation. Document review. It connects to eleven financial data platforms and pulls numbers automatically.

Twenty-one thousand people starred the repo on GitHub. Twenty-eight hundred forked it¹.

A job that used to require an elite degree and three years of training can now be installed by anyone with a laptop and an internet connection.

The Old Rules Worked for Thirty Years

For three decades, society ran on a pretty simple script. Good college. Good diploma. Good job. Retire comfortably.

This worked because of one assumption. Employers believed a diploma meant you could think. They paid a premium for that signal.

That assumption is now cracking on both sides.

First, AI is getting good at exactly what college teaches you. Finding information. Organizing it. Writing reports. Running standardized analyses and calculations. These are the core skills of a knowledge worker. They are also what large language models do best.

Second, employers are starting to question whether the diploma is worth what they pay for it.

McKinsey researchers found something called the “college premium” — the extra money graduates earn compared to non-graduates — has stopped growing since around 2010. Not slowed down. Stopped. Wages for knowledge workers have been flat since mid-2024².

What Five Industries Actually Look Like Right Now

Industry	Key Change	Source
Software Engineers	Entry-level hires down 49%; employment for ages 22-25 down ~20%	34
Banking	Six largest US banks cut 15,000 in Q1; profits up 18% same period	5
Consulting	McKinsey: 40,000 humans + 25,000 AI agents; client-facing +25%, back-office -25%	6
Law	Baker McKenzie cut 600-1,000 (citing AI); Clifford Chance cut 50; Irwin Mitchell cut 56	7
Accounting	KPMG cut ~10% of audit partners	8

KPMG’s head of audit technology put it bluntly. Within two to three years, routine audit work will involve “almost no humans”⁸.

These industries used to represent the gold standard of white-collar success. Prestigious degrees. Respected titles. High pay. Now they are the first places where AI is eating into headcount.

A global survey by Mercer, the human-resources consultancy, found 40% of workers worry AI will take their job in 2026. In 2024, that figure was 28%⁹.

That is a twelve-point jump in just two years.

What the People Building AI Actually Say

Media reports are one thing. Here is what the executives creating these tools say when asked directly.

Dario Amodei is the most alarmed.

The CEO of Anthropic published a twenty-thousand-word essay in January 2026. His core argument: “AI is not replacing one profession. It is replacing human labor in general. It is a general-purpose substitute for labor. Its impact will be faster and broader than any previous technology revolution. It will hit finance, consulting, law, and technology simultaneously, giving displaced workers no time or space to switch industries.”¹⁰

He had said earlier that AI could eliminate half of all entry-level white-collar jobs within one to five years. Unemployment, in his view, could spike to 10% or 20%. He described a future where “cancer is cured, the economy grows 10% a year, and the government budget is balanced — but 20% of people have no job.”¹¹¹²

Sam Altman takes a more complicated position.

The CEO of OpenAI has said two things that do not quite line up. On one hand, he admits the problem is real. “The balance between labor and capital is being broken by AI, and honestly nobody knows what to do about it.” On the other hand, he is skeptical of some claims. A lot of companies were going to cut jobs anyway and are now blaming AI for decisions they already planned to make¹³¹⁴.

Jensen Huang is the optimist.

The CEO of NVIDIA said the line that went viral everywhere: “You will not lose your job to AI. You will lose your job to a human who uses AI.”¹⁵

He believes AI will eventually create more jobs than it destroys. “At the end of this industrial revolution, more people will be working than when it started.”¹⁶ He has sketched a future for his own company. Seventy-five thousand employees working alongside seven and a half million AI agents — one human for every hundred AI workers.

Satya Nadella lands in the middle.

The CEO of Microsoft acknowledges the disruption is real. He also believes human adaptability is real. He draws a sharp distinction between knowledge work and knowledge workers. The tasks can go to AI. The people still need to supervise, guide, and own the outcomes¹⁷.

Jamie Dimon is a pragmatist.

The chairman and CEO of JPMorgan Chase says “AI will eliminate jobs. Stop burying your head in the sand.” But he also notes that banks naturally lose about 20% of their staff every year to turnover. Retraining and redeployment can absorb a lot of the shock. “Your children are going to live to one hundred and work three and a half days a week.”¹⁸¹⁹

The economist David Autor offers a framing worth sitting with.

“The real danger of AI to the labor market is the devaluation of expertise.”²⁰

That distinction matters. People will still work. But the four years you spent in college, the three years in graduate school, and the five years building experience may not command the premium they used to. The skills themselves, not the worker, are what lose value.

Larry Fink, the CEO of BlackRock, said something uncomfortable at Davos.

“If AI does to white-collar workers what globalization did to blue-collar workers, we need to face that reality head-on.”²¹

Globalization shipped millions of American factory jobs overseas. The blue-collar middle class shrank for decades. Fink’s point is that the same pattern is about to repeat for office workers. And it will happen much faster this time.

If AI Can Do What You Can Do, What Should You Learn?

All of this data points to one question.

AI can already do what a “good student” does. Find sources. Summarize them. Write reports. Run standardized analyses and calculations. So what should school actually teach?

That question might matter more than whether AI will replace your job. It is about whether the things you spend your days learning right now will still matter in ten years.

Here is what is getting cheaper.

Any task that can be written as a step-by-step manual is at risk. Gathering and organizing information. Drafting documents. Running routine analyses. If the steps can be documented, AI can execute them.

So what is left for humans?

Four things, mainly.

Judgment.

When information is incomplete and the rules are unclear, someone has to decide. AI can generate ten options. It cannot choose. It cannot take responsibility for the choice. It cannot explain the trade-offs to a worried client. That requires a human who is willing to be wrong and to pay the price for being wrong.

Trust.

People pay for advice not just because of the analysis but because they trust the person giving it. Trust takes time. It takes face-to-face conversation. It takes showing up consistently enough that someone believes you have their interests in mind. AI cannot fake that.

Accountability.

An AI cannot testify in court. It cannot sign an audit opinion. It cannot put its name on a decision that carries legal liability. As long as regulations require a real person to stand behind a decision, humans remain essential.

Asking the Right Questions.

When AI can answer almost every question that has already been asked, the scarce skill becomes knowing what to ask. School trains you to answer. It does not train you to interrogate the question itself.

The people who will be hardest to replace are the ones who can make calls when nobody knows the right answer, earn trust without relying on a credential, own the consequences of their choices, and ask questions that no one else thought to ask.

These are the skills standardized tests do not measure. They are also the ones AI handles worst.

What This Means for You

If you are a student right now — grinding through homework, cramming for exams, wondering if any of it matters — here are three things to keep in mind.

First, the path from “study hard” to “good college” to “stable job” is not as reliable as it was for your parents’ generation. The foundation is shifting. That does not mean education stops mattering. It means the kind of education matters more than the credential itself.

Second, being a “standard good student” — high test scores, following instructions, never questioning the assignment — is exactly the profile most vulnerable to AI replacement. The abilities exams measure are the ones AI already has.

Third, learning to do what AI cannot do — exercising judgment under uncertainty, building trust with real people, taking responsibility for outcomes, and asking better questions — builds a new kind of moat.

The white-collar moat is not gone. It has just moved.

You still have time to get there.

---

References

Footnotes

1. GitHub — Anthropic financial-services toolkit: https://github.com/anthropics/financial-services ↩

2. McKinsey Global Institute — Agents, Robots, and Us: Skill Partnerships in the Age of AI, December 2025: https://www.mckinsey.com/mgi/our-research/agents-robots-and-us-skill-partnerships-in-the-age-of-ai ↩

3. Indeed Hiring Lab — Software engineer job posting trends, 2025–2026 ↩

4. Stanford Digital Economy Lab — Canaries in the Coal Mine: Young Workers in the Age of AI, November 2025: https://digitaleconomy.stanford.edu/publications/canaries-in-the-coal-mine/ ↩

5. ResultSense — America’s six largest banks cut 15,000 jobs in Q1 2026, April 2026 ↩

6. Business Insider — McKinsey’s AI transformation: 40,000 humans plus 25,000 AI agents, 2025: https://www.businessinsider.com/mckinsey-ai-agents-workforce-2025 ↩

7. Reuters — Baker McKenzie cuts 600–1,000 jobs citing AI, 2025–2026 ↩

8. KPMG US — Audit partner reduction plan, 2025–2026 ↩ ↩²

9. Mercer — Global Talent Trends 2026: https://www.mercer.com/global-talent-trends-2026/ ↩

10. Investopedia — Anthropic CEO warns of AI threat to jobs, “underclass” looms, January 2026: https://www.investopedia.com/anthropic-ceo-warns-of-ai-threat-to-jobs-unemployed-or-very-low-wage-underclass-looms-11893595 ↩

11. Axios — Anthropic CEO: AI could eliminate half of entry-level white-collar jobs, May 2025: https://www.axios.com/2025/05/28/ai-jobs-white-collar-unemployment-anthropic ↩

12. Axios — Same interview, “cancer cured but 20% unemployed” scenario: https://www.axios.com/2025/05/28/ai-jobs-white-collar-unemployment-anthropic ↩

13. Fast Company — Even Sam Altman thinks CEOs are blaming AI for layoffs, February 2026: https://www.fastcompany.com/91495694/even-sam-altman-thinks-ceos-are-blaming-ai-for-layoffs ↩

14. Fortune — Sam Altman: AI is breaking the labor-capital balance, March 2026: https://fortune.com/2026/03/12/sam-altman-ai-labor-capital-jobs-nobody-knows/ ↩

15. Fortune — Jensen Huang: You will be replaced by a worker who uses AI, April 2026: https://fortune.com/2026/04/22/jensen-huang-nvidia-ai-replace-workers-not-jobs/ ↩

16. HR Chief Magazine — NVIDIA CEO on AI creating jobs, April 2026: https://hrchiefmagazine.com/news/nvidia-ceo-jensen-huang ↩

17. Economic Times — Microsoft CEO Satya Nadella: AI job disruption is real, but so is human adaptability, March 2026: https://economictimes.indiatimes.com/tech/artificial-intelligence/ai-led-job-disruption-is-real-but-so-is-human-adaptability-microsoft-ceo-satya-nadella/articleshow/129037925.cms ↩

18. Fortune — JPMorgan CEO: AI will eliminate jobs, October 2025: https://fortune.com/2025/10/15/jpmorgan-ceo-jamie-dimon-ai-job-losses/ ↩

19. Economic Times — Jamie Dimon: Next generation will work 3.5 days a week, November 2024: https://economictimes.indiatimes.com/news/international/global-trends/jpmorgan-ceo-jamie-dimon-envisions-a-future-where-next-generation-will-work-3-5-days-a-week-and-live-to-100-years/articleshow/115691808.cms ↩

20. MIT — David Autor: The real danger of AI is the devaluation of expertise, February 2024: https://shapingwork.mit.edu/research/applying-ai-to-rebuild-middle-class-jobs/ ↩

21. Investopedia — BlackRock CEO Larry Fink on AI worsening inequality, January 2026: https://www.investopedia.com/blackrock-ceo-larry-fink-what-happens-to-everyone-else-if-ai-fuels-inequality-11889891 ↩