Is AI taking Entry-Level Jobs? What the canaries are telling us
"AI is reshaping work faster than we can keep up, and a new Stanford study just made that crystal clear."
Last weekend, when I was browsing LinkedIn, I came across a post by Bojan Tunguz. Curious, I clicked. The title - "Canaries in the Coal Mine? Six Facts about the Recent Employment Effects of Artificial Intelligence" - grabbed me instantly. I was reading the abstract when suddenly it felt urgent, almost personal. So I spent part of my weekend digging into it more, reading and understanding the paper. What follows is my attempt to break down the six facts, along with my own reflections.
Before we dive into the six facts, it's best to know about the data. If you would like to read or learn more about the paper, it could be found here.
Primary Data Source
The core dataset for analysis consists of monthly, individual level payroll records from ADP, the largest payroll processing firm in the US.
Scale: 3.5 to 5 million records for each month
Time Period: January 2021 through July 2025
Job Titles: 2010 Standard Occupational Classification (SOC) codes
Employment Type: Full-time used; Part-time employees excluded from main sample
Age Range: 18 - 70 years; early career workers are in the range 22-25
These are just the core level information as knowing where the data comes from matters as it will directly shape how we read figures and interpret results. With this taken care of, let's dive into what the paper actually tells about the facts.
Fact 1: Young workers are losing jobs in AI-exposed occupations
Since generative AI became widely used in jobs like software development and customer services, employment for early-career workers (22-25 years old) have seen a notable drop in the number of people employed since late 2022. In contrast, older workers in the same career fields or fields where there is less AI exposure (like nursing aides) have seen stable or even positive growth.
My take: The paper tells that there is a 13% relative decline in employment for early-career workers. I was appalled when I saw the drop in early-career workers in the graph. However, it was not shocking as AI has started to replace certain "entry-level knowledge" that young workers bring, while experience and tacit skills still protects older cohorts.
Fact 2: Overall jobs are growing, but not for young people
The economy as a whole continues to add jobs. But for younger workers, employment growth has been stagnant since late 2022. This stagnation is largely attributed to the decline in AI-exposed job for early-career workers. In low-exposure jobs, young workers are keeping pace with older ones.
My take: Personally, this fact is in conjunction with the first fact. When looking into the growth decomposition of jobs across age and exposures, there has been a negative growth for early-career workers in high AI-exposed jobs (over -5%). This can also be seen in social media apps like Reddit where there are constant discussions of struggles by recent graduates and early-career workers.
Fact 3: Automation hurts more than augmentation
The decline in entry-level employment is concentrated in occupations where AI is more likely to automate (replace) human labor than augment (enhance) human labor. Occupations with high potential of automation are shrinking for young workers whereas augmentation-heavy jobs are stable or even growing.
My take: Automation and augmentation aren't just simple good/bad opposites. It's like a double-edged sword. The challenge is learning how to hold the sword without hurting ourselves.
Fact 4: It's not just industry downturns
The employment declines observed for young workers are not merely a result of broader result or firm-level shocks (for instance, interest-level changes). The trend remains the same, employment in the most AI-exposed jobs showcase a decline for young workers. The authors estimate a 12 log-point relative decline (the 13% decline authors talked about in the abstract).
My take: Even after adjusting for various firm-level shocks, the pattern still remains. This is not just a bad industry cycle, the disruption by AI is real.
Fact 5: Wages aren't falling, jobs are
The labor market adjustments due to AI are primarily visible in employment levels rather than compensation. Fewer people are being brought in, rather than wage cuts for existing employees.
My take: For those already in a job, the ground feels stable. But for those trying to get in, the door is narrowing.
Fact 6: The pattern is robust across industries
The findings presented by the authors are largely consistent across different sample constructions. Even after excluding tech firms, separating remote vs non-remote jobs, even including part-time workers or by extending the timeline before AI rise, the pattern still holds: young workers in AI-exposed jobs consistently lose out.
My take: This is my favorite section. The authors really stress-tested their findings and their results shows consistency.
Conclusion
To conclude, this study does not claim that AI is eliminating all jobs. Instead, it reveals a sharper, more targeted disruption: entry-level roles in AI-exposed fields are shrinking, even as the broader job market grows.
As I finished reading the paper, I realized the authors' pragmatic approach on choosing the title. It reminded me of the COBOL disaster during the pandemic. For years, experts had warned that a shortage of COBOL programmers could cause trouble, but their warnings were ignored. It was when New Jersey's unemployment system crashed, the fragility and importance of such infrastructure was exposed and understood.
In the same way, today's decline in entry-level employment for AI-exposed jobs might looks small. But they are our canary - an early warning that shouldn't be ignored. The signal is clear: we need to act now to shape a future where AI augments human potential rather than simply automates it, and where young talents can still find a foothold to begin their careers.
