AI Is Replacing Jobs — But Not the Ones You Think

The headlines contradict each other. Goldman Sachs says AI eliminates 16,000 net U.S. jobs every month. The Economist says AI created a million new positions. Fortune reports 66% of CEOs are freezing hiring. CompTIA reports 255% growth in AI engineering roles. Every number is correct. None of them tell the full story.

The actual situation is more complicated and more important than any headline captures. AI is not simply “taking” or “creating” jobs. It is restructuring the entire labor market in real time — suppressing entry-level hiring, accelerating senior roles, destroying routine tasks, and creating entirely new job categories that did not exist two years ago. The pain and the opportunity live in the same statistic.

This is the most comprehensive, data-driven breakdown of what is actually happening with AI and jobs in September 2026.

The numbers that matter

Three data sources anchor this story. Each measures a different thing, which is why they appear to contradict.

Goldman Sachs (April 2026): ~16,000 net U.S. jobs eliminated per month. That breaks down as roughly 25,000 positions automated, partially offset by 9,000 created through augmentation. Annualized, that is ~192,000 net positions — about 0.1% of the 160-million-person U.S. workforce. Goldman also reports that 34% of U.S. work tasks are now technically automatable by current AI tools, up from 25% in 2023.

The Economist (2026): AI has created approximately 1 million net new positions in the U.S. since mid-2023, against roughly 200,000 AI-related layoffs in the same period. That is a 5:1 creation-to-destruction ratio. LinkedIn data supports this: 640,000 new AI-specific jobs posted between 2023 and 2025, with AI job postings rising 14% year-over-year, then 156% the next year.

Census Bureau / CES Working Paper (April 2026): Early career workers ages 22-24 in the most AI-exposed industries saw a 12% employment decline over 10 quarters following ChatGPT’s introduction. Hiring rates recovered by early 2025, but only because the employment base had already shrunk — fewer workers to replace.

These numbers are not contradictory. They measure different slices of the same reality. Goldman measures net displacement across the whole economy. The Economist measures new AI-specific roles. The Census Bureau measures entry-level impact. Together, they tell a coherent story: AI is creating more jobs than it destroys in total, but the destruction is concentrated at entry level while the creation favors experienced workers with AI skills.

The hiring freeze is real — and it matters more than layoffs

The most underreported story in AI employment is not layoffs. It is the freeze.

Corporate America eliminated 1.17 million jobs in 2025 under the logic that excess labor had to be cut to fund AI. But the bigger shift is happening in hiring decisions that never get announced:

66% of CEOs plan to freeze or cut hiring through the rest of 2026, according to a survey of 350+ public-company CEOs managing $19 trillion in assets.
Entry-level job listings dropped 30% and middle management postings dropped 42% since 2022.
SAP froze most hiring and non-essential travel to fund AI, telling employees it will “exclusively focus new hiring on selected profiles only, mainly core AI roles.” SAP has 110,000 employees worldwide.
Microsoft froze hiring in cloud and North American sales divisions. Copilot AI teams still recruiting.
Meta froze hiring and internal transfers in its AI division after adding 50+ researchers and engineers in 2025-2026.
Oracle and Salesforce cut roles to fund AI infrastructure.

This pattern — suppressing new hiring rather than eliminating existing positions — is harder to measure than layoffs but potentially more consequential. A layoff is a one-time event. A hiring freeze is a structural shift. Every quarter that entry-level hiring stays suppressed, the career pipeline gets thinner. Fewer junior roles today means fewer senior workers five years from now.

The Census Bureau data confirms this. Early career employment in AI-exposed industries dropped 12% — not because companies fired existing workers, but because they stopped hiring new ones. The hiring rate “recovered” by 2025, but only because the base had already shrunk. You cannot hire your way out of a smaller workforce.

The occupations most at risk

Goldman Sachs scored occupations on a 0-100 AI exposure index. The highest-risk roles share a common pattern: routine, rule-based, data-processing work.

Occupation	AI Exposure Score	What AI replaces
Telephone operators	91/100	Routine communication, script-based responses
Data entry clerks	87/100	Manual data transcription and formatting
Insurance claims clerks	83/100	Document processing, rule-based decisions
Customer service representatives	76/100	Template responses, FAQ handling
Paralegals / legal support	71/100	Document review, contract analysis
Software developers (ages 22-25)	64/100	Junior coding tasks, boilerplate code

The age disparity is striking. Workers aged 22-30 experience AI-driven displacement at nearly 3x the rate of workers aged 40-55. This is not because older workers are better at AI — it is because younger workers disproportionately hold the routine, entry-level roles that AI handles first.

Stanford HAI data shows software developer employment for ages 22-25 is down approximately 20% since 2024. The pattern extends beyond tech: the Census Bureau found declines across most sectors, not just technology.

The most vulnerable jobs are not the ones that require the most intelligence. They are the ones that follow the most rules. AI excels at pattern matching, document processing, template-based communication, and rule-based decision-making — exactly the tasks that entry-level workers perform.

The jobs being created

The creation side of the ledger is real and growing faster than the destruction.

AI specialist roles are expanding at 8x the rate of the overall job market — 69% growth vs. 9% (PwC 2026 AI Jobs Barometer, covering 1 billion+ job ads across 27 countries). The wage premium for AI skills reached 62%, up from 57% a year ago.

CompTIA reported 280,000+ new tech job postings in June 2026, with AI and data job postings surging 80% year-over-year and AI engineering roles specifically up 255%.

The PwC data tells the most interesting story. Companies most exposed to AI expanded headcount by 52% since 2018, compared to 36% for least exposed. Productivity growth was 40% higher at the most AI-exposed companies. The top 20% of AI-exposed companies achieved 163% productivity growth. And “professionalized” roles — roles where AI augments human expertise rather than replacing it — are growing 2x faster than “democratized” roles, with 42% faster wage growth.

A September 2026 study of 21,559 U.S. firms found that high-intensity AI adopters (spending approximately $33 per employee per month on AI) saw approximately 10% employment growth over 24 months. Entry-level employment at these firms grew approximately 12%. The gains are concentrated in the Information sector at approximately 13% employment growth.

The pattern: companies that adopt AI aggressively hire more people, not fewer. The jobs they create are different — higher-skill, higher-pay, more strategic — but they are real jobs.

The productivity paradox

Here is the twist that complicates both the doomer and the optimist narratives: the productivity gains are not showing up in the data yet.

An Atlanta Fed / NBER study surveyed approximately 750 corporate executives and found that perceived productivity gains are larger than measured gains. This is the “productivity paradox” — revenue realizations lag capability. Firms see what AI can do, but translating that into output takes time.

50% of firms explicitly report AI will not replace any roles. Only 44% indicate some degree of job replacement. Aggregate employment decline due to AI in 2026 is expected at less than 0.4% — “very small and likely hard to detect in aggregate statistics.”

The BCG data suggests the real timeline: 50-55% of U.S. jobs will be reshaped by AI in 2-3 years, but only 10-15% are at risk of elimination. Reshaped means the tasks change, the tools change, the skills required change — but the job itself continues. A paralegal who used to review documents manually now reviews AI-generated summaries. A developer who used to write boilerplate now reviews AI-generated code. The role persists. The work transforms.

Cognizant’s 2026 analysis puts it at 93% of jobs being impacted in some way by AI — six years ahead of their original 2032 forecast. Impact is not elimination. Most of that 93% involves task restructuring, not job loss.

What the data actually says

The honest synthesis is this:

AI is creating net jobs overall. The Economist’s 1 million figure is directionally correct, and it aligns with PwC, CompTIA, and LinkedIn data. The WEF projects a net gain of 78 million jobs globally by 2030 (170M created minus 92M displaced).

The pain is real and concentrated. Entry-level workers, routine-task roles, and workers in industries slow to adopt AI face genuine displacement. The 12% decline in early career employment is not a rounding error. The 30% drop in entry-level listings is structural.

Hiring suppression is the underreported story. Companies are not mass-firing. They are mass-not-hiring. This suppresses opportunities for new entrants while protecting existing workers. The long-term consequence is a thinner talent pipeline and a harder on-ramp for the next generation.

The transition is the crisis, not the destination. The end state — where AI creates more jobs than it destroys, with higher productivity and higher wages — is supported by current data. The transition period of roughly 2025-2030 is where the pain lives. People are losing jobs now. New jobs are being created now. They require different skills, exist in different industries, and are distributed unevenly.

The most honest answer to “is AI taking jobs or creating them?” is: yes, both, simultaneously, unevenly, and the net effect is positive — but the transition is genuinely painful for the people caught in it.

What workers should actually do

Three evidence-based strategies:

Develop AI skills. The wage premium is 62% and rising. AI specialist jobs are growing 8x faster than the total market. You do not need to become an AI researcher. You need to learn how to use AI tools in your existing role. The workers getting hired are not the ones who know the least about AI — they are the ones who know how to combine AI with domain expertise.

Move up the value chain. AI automates routine tasks. It amplifies strategic, creative, and interpersonal work. The BCG data is clear: 50-55% of jobs will be reshaped, not eliminated. The workers who survive are the ones who move from executing tasks to directing AI systems that execute tasks. A lawyer who reviews AI-generated briefs instead of writing them. A developer who architects systems instead of coding boilerplate. A marketer who designs campaigns instead of writing copy.

Stay in high-adopting organizations. The September 2026 study of 21,559 firms is unambiguous: companies that adopt AI aggressively hire more people, grow faster, and create more entry-level positions. Companies that resist AI shrink. The best career move in 2026 is to work for a company that is spending $33+ per employee per month on AI tools and training.

Sources

Primary research

Lee C. Tucker, “You’re (not) Hired: Artificial Intelligence and Early Career Hiring,” CES Working Paper 26-27, April 2026
Goldman Sachs labor market analysis, economist Elsie Peng, April 2026
PwC 2026 AI Jobs Barometer, 1 billion+ job ads across 27 countries, June 2026
Federal Reserve Bank of New York, NACE, ZipRecruiter — college graduate employment data, 2026
Atlanta Fed / NBER study, approximately 750 corporate executives, March 2026
Study of 21,559 U.S. firms, September 2026

Industry reports

BCG, “AI and the Future of Jobs,” April 2026
WEF Future of Jobs 2025
McKinsey, November 2025
Cognizant, 2026 workforce analysis
CompTIA, June 2026 job postings data
Stanford HAI 2026 AI Index Report

News reporting

Fortune, “66% of CEOs plan to freeze or cut hiring,” 2026
The Economist, AI job creation analysis, 2026
CNBC, Federal Reserve Bank of New York graduate employment data, March 2026
The Information, SAP / Microsoft / Meta hiring freezes, 2026

Previously on Father of AI