Erik Brynjolfsson is not a doomer.
He is one of the most respected labor economists in the world. Stanford Digital Economy Lab. The author of The Second Machine Age. The person economists call when they need real numbers on technology and work.
In August 2025, his team published a study that quietly broke the AI jobs debate. Not with predictions. Not with surveys. With actual paycheck data from millions of US workers, tracked monthly, in real time.
The finding was specific and undeniable. AI is not eliminating jobs across the board. It is eliminating one specific thing: the entry point.
The 22-year-old vanished. The 35-year-old got a raise.
Brynjolfsson partnered with ADP, the largest payroll provider in the United States, processing checks for tens of thousands of firms. They tracked employment month by month from late 2022 (when ChatGPT launched) through mid-2025.
They split workers into two groups: those in jobs highly exposed to AI (software development, customer service, copywriting, paralegal work) and those in jobs less exposed (skilled trades, healthcare, manual work).
Then they split each group by age.
Age 22–25, AI-exposed jobs
-13%
Employment decline since late 2022
Software devs 22–25
-20%
From 2022 peak
Age 30+ same fields
+6 to 12%
Employement grew
Age 22-25 Non AI Jobs
Stable
No statistically significant decline
Employment change by age & AI exposure, late 2022 to mid-2025
Why this matters
The study controlled for firm-level economic shocks. It excluded technology companies (in case the tech downturn was driving the numbers). It excluded remote work occupations. The 13% decline held up against every alternative explanation. The only variable that consistently explained the pattern was AI exposure of the occupation.
AI is not eliminating software engineering. It is eliminating the entry point into software engineering.
Presented by Atoms.dev
Tell Atoms your idea. A team of AI employees - PM, engineer, SEO specialist, data analyst that builds it, deploys it, and helps you acquire customers. 165,000 GitHub stars. #1 Product Hunt of the Week. Trusted by Amazon, Microsoft, and NVIDIA.
Why the bottom rung disappeared first
The why is more interesting than the what.
Junior workers depend on what economists call codified knowledge. The kind of knowledge you get from school, textbooks, online courses, formal training. The kind of knowledge that can be written down, structured, and learned from documentation.
Senior workers depend on tacit knowledge. The intuition that comes from years of mistakes. Knowing which client is bluffing. Sensing when a project is about to go sideways. Reading the room.
The Brynjolfsson observation
"It appears what younger workers know overlaps with what LLMs can replace. The codified knowledge that a junior developer brings to their first job - the data structures, the algorithms, the framework documentation - is exactly what large language models were trained on. Tacit knowledge, by contrast, has never been written down anywhere. So AI cannot replicate it."
When companies adopt generative AI tools, the data shows junior employment drops 9 to 10 percent within six quarters. Senior employment barely moves. The mechanism is mechanical:
Step 1: Company adopts AI tools (Cursor, Claude Code, Copilot, Intercom AI).
Step 2: Tasks that juniors used to do - boilerplate code, first-draft documentation, simple ticket triage - get absorbed by AI.
Step 3: The remaining tasks are too complex or too high-stakes for juniors. Companies redistribute them upward to senior staff.
Step 4: Junior openings stop being posted. Existing juniors don't get replaced when they leave.
Step 5: Senior workers handle more output per person, often with AI augmentation. Their wages rise. Headcount at the bottom shrinks.
The result is a labor market that looks fine in aggregate. Unemployment stays steady. Total employment grows. But the career ladder loses its bottom rung.
Junior vs senior employment change after AI tool adoption (within 6 quarters)
Section 03: What this means for you
Three honest questions worth sitting with this week
The Stanford data does not tell anyone what to do. It just shows what is happening. But three questions are worth sitting with - whether you are a founder, a manager, a worker, or a parent.
1. If you hire - are you skipping the entry level?
The companies that drove this data did not announce mass layoffs. They simply stopped hiring juniors. Quietly. If you are running a team, audit your last 12 months of hires. What percentage are entry-level vs experienced? In 2022 that ratio was healthy across most industries. In 2025 it shifted dramatically. Where does yours sit?
2. If you are 22–28 - where does your tacit knowledge come from?
Klarna replaced humans entirely. The hybrid model - AI for volume, humans for complexity - consistently outperforms full automation on both cost and satisfaction metrics. Map every AI-powered function in your business and ask: what breaks if you remove the human entirely?
3. If you are 30+, your value just went up. Are you pricing accordingly?
The same data showing junior employment fell shows senior employment in the exact same fields grew 6 to 12 percent. Your tacit knowledge, the things you know from doing the work, not learning it, is now economically more valuable than it was three years ago. If your compensation, rate, or role has not changed to reflect this, you may be the one undervaluing what you have.
Presented by Semrush
Struggling to understand why your competitors rank higher or where your next customers will come from?
Use Semrush - the powerful platform that brings clarity to your entire digital marketing strategy.
Keyword & topic research to target the right audience
Comprehensive site audits to identify and fix SEO issues
Backlink analysis with valuable link-building opportunities
Content optimization tools to boost rankings and engagement
Advertising & PPC insights to maximize campaign performance
🔮 The Bottom Line
The Brynjolfsson team called this paper "Canaries in the Coal Mine." The phrase is deliberate.
Coal miners brought canaries underground because the birds were more sensitive to invisible gas than humans. When the canary died, the miners had time to leave before the air killed them too.
The 22-year-olds in AI-exposed jobs are the canaries. The 13% employment drop is the warning signal. It does not mean the entire labor market is collapsing. It means the early-warning signal is now flashing - and how the rest of us respond will determine what happens next.
Whether you read this as a threat or an opportunity depends entirely on where you are standing right now.
📧 Forward this to 3 entrepreneur friends who need to see this opportunity
