Generative AI is draining junior workers from the economy—or so the narrative goes. A Harvard study claimed AI adoption shrunk entry-level jobs. Stanford researchers documented “large declines in early-career hiring concentrated in GenAI-exposed occupations.” U.S. new-grad unemployment is at a four-year high, and AI seems like the obvious culprit.
The logic is compelling: AI automates mundane cognitive tasks—coding boilerplate, legal review, customer service—that juniors do. But a new
LSE working paper by Lambert & Schindler says this verdict is premature.
The paper analyzes 243 million new hires and 407 million job postings across the US, UK, Canada, and Australia (2017–2025). Their finding is devastating to the AI narrative:
GenAI exposure is strongly correlated with working from home (WFH). Both hit white-collar, routine-cognitive jobs post-pandemic. Previous studies attributed to AI what was actually caused by remote work.
When estimated separately, both look equally damaging: a two-standard-deviation increase in either predicts roughly a 5 percentage point fall in junior-share of new hires. But when estimated jointly, the results shift dramatically:
WFH effect remains strong (~5pp decline, statistically robust)
GenAI effect attenuates sharply and becomes statistically indistinguishable from zero
Alternative exposure measures, residualization designs, and replacing exposure with actual WFH adoption all confirm the same finding: WFH is the robust predictor, not AI.
The mechanism is clear. Junior workers need supervision, feedback, and on-the-job learning. Remote work makes mentorship expensive:
Informal learning disappears (overhearing conversations, quick Q&A)
Monitoring productivity is harder
Onboarding is broken without physical proximity
Seniors need less hand-holding, so they’re unaffected. Firms prefer “safe bets” when oversight is costly.
For policymakers: Don’t blame AI for labor market failures without evidence. If WFH is the issue, RTO incentives or hybrid-mentorship programs could help.
For economic development: FDI and industrial clusters need human capital pipelines, not just AI adoption. Skill transfer happens through physical mentorship, not Zoom calls. When multinationals adopt remote-first models in India, they may skip junior hiring entirely—undermining long-term industrial development.
For professionals: Juniors should seek roles with in-person mentorship. Seniors should remember their value rose because firms can’t mentor remotely, not because AI hates juniors.
The verdict on AI is premature, not forever wrong. But if remote work is breaking the junior ladder, the fix is structural: co-located training, apprenticeships, hybrid models that preserve learning.
Human capital pipelines are physical infrastructure. You can’t build them remotely.
What’s your experience? Are juniors in your firm struggling more with AI or remote work?
More reading..
Office Chai summary of the paper
Harvard study on AI and entry-level jobs
Forbes: “AI Is Not Killing Entry Level Jobs”
Forbes: “Harvard Got It Wrong: AI Is NOT Killing Entry Level Jobs
No posts
Comments
Nothing yet. Say the first thing.
Sign in to join the conversation.