Blog · Oct 16, 2026 · By Velocity11

AI isn't firing people. It's not hiring them.

Everyone asks the same question about AI and jobs: will it fire us? Stanford researchers looked at actual payroll records - millions of US workers, through June 2026 - and found a stranger answer. AI is not firing people. Companies are simply not hiring young people into the jobs AI can do. [1]

The study, in plain words

Three Stanford economists, led by Erik Brynjolfsson, used payroll data from ADP, the company that processes paychecks for millions of American workers. Not a survey, not opinions - records of who actually has a job, month by month, from before ChatGPT launched through June 2026. They sorted jobs by how exposed they are to AI. The most exposed include software developers and customer service agents. [1]

The six facts, shortened

1. There is no economy-wide AI job loss. Overall employment is fine. [1]
2. But workers aged 22-25 in the most AI-exposed jobs are now 19% below where they would be if they had kept pace with their peers in less-exposed jobs. Experienced workers in the same jobs show no such gap. [1]
3. The gap keeps widening: it was 15% in mid-2025 and 19% by June 2026. [1]
4. It works through hiring, not firing. Companies are not laying off young workers - they are not bringing new ones in. [1]
5. The declines happen where AI substitutes for human work. Where AI complements workers, employment is flat or growing. [1]
6. Pay has not moved. The adjustment is happening in headcount, not wages. [1]

In raw numbers: between November 2022 and June 2026, employment of 22-to-25-year-olds in the most exposed jobs fell about 11%. In the least exposed jobs, the same age group grew about 10%. Workers aged 35-49 in those same exposed jobs grew about 10% too. [1]

Why young workers specifically?

The researchers' explanation: AI is very good at the kind of knowledge you learn in school - codified, written-down knowledge. That is exactly what a fresh graduate brings to their first job. What AI cannot easily copy is the unwritten knowledge experienced workers carry: how this client thinks, which shortcut breaks the system, when to escalate. So the senior developer keeps their job, helped by AI. The junior one never gets hired. [1]

The paper checks the obvious other explanations. The pattern holds when you remove tech companies, when you account for interest rates, and when you account for remote work. It is not "the tech downturn." [1]

Fair warnings

• The authors themselves call these "canaries in the coal mine" - early warning signs, not proof of cause and effect. Some of these trends started before ChatGPT. [1]
• This is US payroll data. Other countries may look different. [1]
• Young workers' total employment only fell about 2% overall, because many moved to less-exposed jobs. The pain is concentrated, not universal. [1]

What to do with this

1. If you are early in your career: be the person AI complements, not the person it replaces. The dividing line is judgment, relationships and context - the things not in any textbook.
2. If you hire: cutting entry-level hiring is a slow-motion problem. Today's missing juniors are 2030's missing seniors.
3. If you are a student: the degree alone was always a bet on codified knowledge. Add the un-codified kind - real projects, real clients, real mistakes.

Try it this week

List the five tasks you do most. For each, ask: could a person with no experience do this with ChatGPT in an hour? The tasks where the answer is no are your career insurance. Grow those.

We use AI to help write these posts, and we say so on every one: this post was written with AI help and reviewed by a person.

Is your industry hiring fewer juniors than three years ago? Tell us on Instagram @velocity11ai.

Source

[1] Brynjolfsson, Chandar, Chen, "Canaries in the Coal Mine? Six Facts about the Recent Employment Effects of Artificial Intelligence", Stanford Digital Economy Lab, August 2026. digitaleconomy.stanford.edu

The hero image was made with AI. The chart was drawn by Velocity11 from the published Stanford numbers. This post was written with AI help and reviewed by a person.