The headline number from March 2026 JOLTS was bad enough on its own. For the first time since 2021, there are more unemployed Americans than job openings, with the ratio sitting at 0.95. NextGig’s live Job Openings vs Unemployed dashboard makes it easy to watch the flip happen month by month. But the aggregate number hides the real story. The labor market did not cool evenly. It cooled almost entirely in one place, and AI is the reason.
The 318,000 number nobody is talking about
Inside the March 2026 JOLTS release, one line item does most of the work. Professional and business services shed 318,000 job openings in a single month, by far the largest decline of any sector. Finance and insurance actually added 98,000. Healthcare and trades held flat or grew. If you carved professional services out of the data, the rest of the labor market would look perfectly fine. So why is white collar knowledge work the only thing falling off a cliff?
The data points all line up
Look at what is happening to the people who normally fill those roles. Job listings for entry level corporate roles are down roughly 15% from a year ago, while applications per posting are up 30%. The unemployment rate for recent college graduates hit 9.7% in September 2025, matching the rate for 20 to 24 year olds with only a high school diploma. That has basically never happened in the modern data. The Federal Reserve Bank of New York’s tracker shows underemployment among graduates aged 22 to 27 sitting at 43% as of late 2025, the highest since the pandemic.
And it is concentrated in the roles AI does best. A Stanford Digital Economy Lab paper found early career workers in heavily AI exposed roles saw 16% slower employment growth between mid 2024 and September 2025 compared to the least exposed roles in the same age band. The list of affected jobs reads like a 2019 LinkedIn feed: junior financial analyst, paralegal research, marketing coordinator, customer service rep, junior software developer.
The CEOs are saying the quiet part out loud
Jamie Dimon of JPMorgan said publicly that the bank expects to hire fewer people in the coming years because of AI. Salesforce, Klarna, IBM, Duolingo, and Shopify have all explicitly told investors that they are slowing hiring in functions where AI tools can absorb the work. When the CEO of a $700 billion bank tells an analyst call that AI will reduce headcount, it is no longer a fringe prediction. It is guidance.
Blue collar is having a different year
This is the part the JOLTS aggregate completely buries. Skilled trades, healthcare technicians, and logistics roles are not seeing the same retreat. CNBC reported in May that hiring for electricians, plumbers, HVAC techs, and other trades has actually accelerated since 2024. Wages for those roles are growing faster than knowledge work for the first time in decades. If you were graduating high school in 2026, the data suggests you would have a better shot at a 12 month income trajectory going into an apprenticeship than into a bachelors program in marketing.
The economist’s warning from 1962
Kenneth Arrow won a Nobel partly for explaining that productivity grows through learning by doing. Workers get better at jobs by being given small tasks that teach the bigger ones. The thing AI is automating right now is exactly that. The junior analyst summarizing earnings calls, the new paralegal reviewing contracts, the entry level developer writing CRUD endpoints. If those tasks disappear, the pipeline of mid career workers in five years gets a lot thinner. Some economists think we are setting up a brutal experience curve cliff around 2030.
What this means right now
If you are looking at the FRED job openings series and feeling alarmed about the macro picture, you should be more alarmed about the micro. The aggregate ratio of 0.95 hides a fact that matters more for individual outcomes. For new grads in AI exposed white collar fields, the effective ratio is closer to 0.5. For tradespeople, it is still well above 1.0. The labor market did not weaken. It bifurcated.
Watch the next two JOLTS releases for the same thing. If professional and business services keeps shedding openings while everything else holds, the AI story is no longer a forecast. It is the present tense.
