Marcus sent 200+ applications before he realized the junior developer positions from 2021 no longer exist. Same degree. Different year.
If you’re struggling to land that first role in software development, customer service, or data analysis, you’re not imagining things.
A new paper, “Canaries in the Coal Mine? Six Facts about the Recent Employment Effects of Artificial Intelligence,” reveals something stark: since late 2022, right when ChatGPT went mainstream, employment for workers aged 22-25 in AI-exposed occupations has dropped 13% relative to their older colleagues.
This is administrative data from millions of American workers. The AI revolution is here already, and it’s eating the bottom of the corporate ladder.
What the Data Shows
Young workers in AI-exposed jobs are getting crushed. Early-career workers in software development, customer service, and data analysis have seen employment decline 13% since late 2022. Meanwhile, older workers in identical roles remain stable. Young workers in AI-resistant fields like nursing are growing normally.
The pattern is surgical. When AI can handle 80% of what a junior employee does on day one, why hire the junior employee? The traditional career ladder assumed you started at the bottom and climbed up. AI removed the bottom rungs.
This is happening within firms, not just across industries. Even at companies that are growing and hiring, they’re specifically not hiring young workers for AI-exposed positions. A software company expands its engineering team by hiring senior developers and using AI for what juniors used to do.
Automation, not augmentation, is the culprit. The researchers distinguished between AI that substitutes for human labor versus AI that complements it. Employment declined only where AI automates tasks. Where AI augments human work, young workers are fine.
If your job is applying textbook knowledge to standardized situations, you’re in the automation zone. If it requires wisdom from years of navigating ambiguity, you’re in augmentation territory. Entry-level workers occupy the first zone.
The adjustment is happening through employment, not wages. Young workers who get hired earn normal salaries. But getting hired is dramatically harder. When everyone can use AI to produce professional-looking work, the bar for “entry-level” rises. You’re competing for fewer jobs at the same pay—against candidates with 3-5 years experience.
This isn’t remote work, outsourcing, or tech volatility. The researchers stress-tested against every alternative explanation. They excluded tech firms, controlled for remote work trends, verified these patterns didn’t exist before late 2022. This is AI’s impact without a doubt.
Why This Is Happening
AI excels at replacing codified knowledge: book learning from formal education. That’s precisely what entry-level workers bring.
Experienced workers possess tacit knowledge: on-the-job wisdom, intuition, contextual understanding.
A junior marketer knows how to write content, track project progress, and coordinate meetings. A senior marketer knows how to plan and execute campaigns on the fly, handle stakeholder objections, educate customers, and not just move the needle but project confidence while doing it.
Here’s the paradox: you acquired tacit knowledge through entry-level positions. If AI eliminates those positions, how does the next generation develop expertise AI can’t replace? Understanding this matters because it reveals where opportunity still exists.
What Actually Works Now
The transition is already happening. History shows these shifts are brutal for individuals caught mid-stream, and this one is moving faster than any before.
Build compound expertise AI can’t replicate
Don’t just learn Python. Learn Python and supply chain operations and how to translate technical constraints to non-technical stakeholders. AI can code. It can’t navigate your company’s org chart or understand why the logistics team will reject your technically perfect solution.
Don’t just learn marketing. Learn marketing and customer psychology and the specific economics of your industry. AI struggles at the intersection of multiple domains.
Demonstrate judgment, not execution
Build a portfolio of decisions you made, not tasks you completed. “I wrote 50 blog posts” is automation territory. “I identified our content ranked but didn’t convert, pivoted to bottom-funnel content, increased qualified leads 40%” is augmentation territory.
A writer I know used AI to handle research, competitive positioning, and style matching. He arrives at first drafts in hours instead of days. What he sells isn’t faster writing: it’s strategic judgment about which angles differentiate his clients and which arguments land with their audience. He’s ghostwriting for five executives now.
If you’re already locked out
Traditional advice no longer applies. Create AI-augmented projects that showcase judgment. Don’t show you can code—show you can use AI to code faster while making architectural decisions that demonstrate you understand trade-offs.
Look for apprenticeship-style programs that value demonstrated ability over credentials. IBM, Google, Accenture have launched programs specifically for people without traditional experience who can learn and apply AI tools effectively.
Consider bridge roles in augmentation-heavy environments. Customer success, account management, implementation consulting require enough human judgment that AI can’t automate them fully, but they teach you the tacit knowledge that makes you valuable.
The Uncomfortable Truth
Many entry-level positions were never that valuable. Junior analysts created slide decks to justify fees. Entry-level developers wrote boilerplate code senior engineers reviewed and rewrote.
These positions existed because they were the least expensive way to get work done. They served as training grounds. But they were never efficient.
AI didn’t just make these positions obsolete. It revealed the work was already low-value.
Where This Goes
The data shows the impact is here for young workers in AI-exposed occupations. Next year, different occupations. The year after, different age groups.
Yes, new roles will emerge (eventually). But transitions take years while displacement happens in months. Most people can’t wait out the gap.
The researchers called their findings a “canary in the coal mine” for good reason. The canaries are dying.
Much depends on whether you’ll be the human using AI to augment your work, or the human AI is augmenting out of existence.
You’re already on one side of that line.
Musings by Madhav 
Leave a comment