AI Jobs Impact: New Roles Outpace Layoffs as Junior Hiring Lags

The optimistic case has more data behind it now

For three years, talk about artificial intelligence and work has mostly been about what will be lost. Research and data released in September and early October 2026 support a more mixed and somewhat more hopeful view. AI does not yet appear to be shrinking total employment. It may be adding jobs and raising pay in the fields most exposed to it. The clear exception is young people trying to get their first job in those same fields, and it is a real one.

The headline figure for the hopeful side comes from The Economist. It estimated in September that AI has created about 1 million American jobs since mid-2023, against about 200,000 layoffs attributed to AI. That works out to roughly five jobs gained for each one lost.13 Most of that gain sits in technical work. The same analysis found that engineers, software developers, mathematicians and data scientists added about 730,000 jobs above the broader professional trend since 2022, though it warned that AI cannot be credited with all of them.13 LinkedIn separately counted about 640,000 new AI-specific jobs created between 2023 and 2025.13

Some occupations long described as easy to automate have kept growing. Paralegal employment rose about 11% between 2023 and 2025, and market-research analyst jobs rose about 6%, while national employment grew about 2% over the same period.13

Pay is rising fastest where AI exposure is highest

Wage data points the same way. An Indeed Hiring Lab analysis published September 17 found that advertised salaries in the most AI-exposed U.S. occupations have risen 46% since 2021. That compares with 39% for all posted wages and 25% for the least-exposed jobs. After adjusting for changes in which occupations are hiring, AI-exposed jobs carried a 5.7% pay premium since ChatGPT launched. When each job title was compared with its own past pay, the premium was still 4.7%.

The usual explanation is that employers will pay more for people who can supervise AI systems, fit them into workflows and handle the harder problems that remain. Studies of firms say something similar. A Ramp Economics Lab paper linked company AI spending to workforce records for 21,000 U.S. firms. Heavy adopters grew employment by 10%, and entry-level employment by 12%, compared with firms that would later spend similar amounts per worker.11 A study of Chinese listed manufacturers found that more AI use went along with modestly higher employment and a larger share of revenue going to workers.19 Chinese factories are a very different setting from U.S. offices, though.

Firms aren't adopting AI and then cutting staff

The strongest evidence against mass displacement comes from Denmark. Economists Anders Humlum and Emilie Vestergaard surveyed 25,000 workers in 11 highly exposed occupations and matched their answers to monthly administrative records. They found no effect of AI chatbots on earnings or hours, and their estimates rule out average effects larger than 2%.14 Their data also shows work being reorganized: chatbot users take on new tasks, and so do colleagues who never use the tools.14 One effect did show up. Workers who adopted chatbots were more likely to change occupations, and they moved into better-paid roles. The authors read this as AI giving workers access to expertise that used to be scarce.14 They call their findings a more hopeful picture than the "white-collar bloodbath" often described in the media.14

The Danish results also showed that firms adopting generative AI were not the ones driving the decline in early-career jobs.14 That is the most important point in the debate. Surveys of executives agree: around 90% of nearly 6,000 executives in four countries said AI had not yet significantly affected employment or productivity.11 Gallup found that only 1% of laid-off workers say AI caused their job loss.11

The exception: the first rung of the ladder

The evidence on young workers is less reassuring. The Stanford Digital Economy Lab's Canaries Dashboard uses ADP payroll records. Its latest data shows employment of 22-to-25-year-olds in highly exposed jobs at 19% below where it would be if it had kept pace with less-exposed peers, up from 15% in mid-2025.51 In the August update, employment of that age group in high-exposure jobs had fallen year over year for 35 straight months.2 The authors say the gap comes mostly from companies hiring fewer young workers, not from firing the ones they have.5

A new 41-country study by Bharat Chandar and Bouke Klein Teeselink, covering 1.25 billion job postings, may explain how both trends can be true. At companies that adopted AI, senior employment rose 6.7%, while the junior share of the workforce fell about two percentage points.56 The authors say the drop in the junior share comes mainly from growth in senior jobs, not from cuts to junior ones.5 In their words, AI saves labor at the junior level and expands it at the senior level in exposed occupations.6 Indeed's data shows the same tilt: 71% of the past year's rebound in software-development postings came from senior roles.

U.S. Census data adds to the concern. Graduates from the most AI-exposed tenth of college majors, mostly computer science, were 5 percentage points less likely to have a job right after graduation. Their starting earnings fell about 13%, a hit similar to graduating into a recession.11

Not everyone agrees AI is the cause

This is where the research splits most sharply. An NBER paper by Robert Fairlie and Jane Wu found that unemployment among college graduates aged 22 to 25 was 7.3% in summer 2026. That is above last year's rate but within the 6.3%-to-7.8% range seen since 2022.10 They found no statistically significant link between AI exposure and graduate unemployment. They did find a significant link with how easily a job can be done remotely.8 Their explanation is that remote work weakens the informal mentoring junior hires depend on, which makes companies more hesitant to bring them on.8 Other researchers have made a similar argument, noting that remote work and AI exposure overlap heavily because both are concentrated in white-collar computer work.1

The Nordic data also leans toward skepticism. Studies covering the full populations of Finland and Norway find no meaningful negative effect of AI on junior employment. The Norwegian data even shows that junior employment in exposed jobs started falling before ChatGPT was released.111 A Census Bureau study confirmed the U.S. drop in young hires, but its author suggested firms may have paused junior hiring because they were uncertain about AI, not because they were already using it.11

On balance, the evidence that young workers in exposed U.S. fields are losing ground is now consistent and hard to dismiss. The evidence that AI itself is the main cause is weaker. Remote work, interest rates and companies' caution about AI's future likely all play a part.

The layoff numbers point both ways

Public discussion often runs together two different things: layoffs that companies announce and blame on AI, and changes that actually show up in employment data. Challenger, Gray & Christmas counted 116,175 announced U.S. job cuts citing AI through August 2026, about 22% of all announced cuts.13 Block's plan to cut more than 4,000 jobs, nearly half its workforce, in an AI-driven restructuring is the best-known example. Those figures are what employers say, not verified job losses.3 Some analysts note that companies may be cutting based on what they expect AI to do, not what it does today.16 Andy Challenger has also noted that hiring rose 25% over the previous year, and argued that AI is reshaping the labor market rather than dismantling it.13

The wider economy is softer than the hopeful story suggests. Charles Schwab notes that September's preliminary payroll gain of 29,000 jobs fell far short of forecasts. It also argues that AI spending has so far boosted productivity and profits more than hiring, and that much of the hardware is imported, which limits the domestic benefit.12 Another study estimates that AI-driven reshuffling of workers may have raised the natural rate of unemployment by 0.1 to 0.2 percentage points since large language models arrived, with considerable uncertainty.17

What to watch

The optimistic reading is supported for the economy as a whole. Total employment looks stable, pay in exposed fields is rising, and workers who use AI seem to be moving up. The risk is in the career pipeline. If companies keep filling exposed roles with senior people paired with AI, the shortage of junior workers could grow into a shortage of experienced workers later. One model estimates that this loss of on-the-job learning could reduce long-run U.S. GDP growth by 0.05 to 0.35 percentage points.11 The International Labour Organization has also warned that job opportunities for younger workers are eroding, even though large-scale displacement remains limited so far.

Agentic AI is still early in its adoption by businesses. The junior-hiring trends seen now could spread to the rest of the labor market once current bottlenecks ease.11 So far, AI looks more like a technology that changes who gets hired than one that eliminates jobs. The main policy concern for now is getting young workers into their first jobs.