What happened
Nolan Lovett of the NATO Special Operations University published research arguing that AI adoption erodes professional expertise by eliminating entry-level positions where junior workers traditionally learn through hands-on mistakes and gradual skill-building. Even when entry-level jobs survive, AI assistance lets junior workers reach productivity levels that used to require years of experience, preventing the cognitive effort that builds deep domain knowledge.
Why it matters
The erosion of entry-level training creates a "validation tether" problem—organizations lose the ability to catch AI mistakes because spotting domain-specific errors requires exactly the kind of deep expertise that AI use is wearing away. The experienced professionals currently on the job market were trained 5 to 20 years ago, so the full damage may not show until between 2030 and 2045. Research backs the concern: an MIT study found AI use weakened neural connectivity; software developers with AI access scored 17 percent worse on knowledge tests; and Chinese students saw exam performance drop by up to 24 percent despite homework grades improving 18 percent.
What to watch
Software engineering, financial analysis, and legal research face the highest vulnerability because they have high task substitutability and light regulation. Lovett proposes solutions including AI-free learning environments, phased AI introduction, and professional certifications testing domain competence alongside AI skills—but acknowledges the tragedy isn't inevitable if professions act to maintain expertise reserves.
Summaries like this, in your inbox every morning.
Lovett's research identifies a market failure he calls the "tragedy of the cognitive commons." When a company deploys AI to replace entry-level work, it captures 100 percent of the efficiency gains immediately. But the cost of eroding expertise is distributed across every organization that draws from the same talent pool—no single firm bears the full cost of the expertise loss, so none has adequate incentive to preserve the training environment. This creates a collective-action problem: rational individual choices (AI adoption for short-term efficiency) combine to produce a collectively damaging outcome (the destruction of expertise pipelines).
The mechanism works through two channels. The first is direct: AI systems take over entry-level tasks, eliminating the positions where junior workers once learned. The second is subtler but perhaps more dangerous: when entry-level jobs persist, junior workers with AI assistance reach productivity levels that previously took years to attain. The cognitive effort—the mistakes caught, the problems solved independently, the domain intuition built—never happens. Without that foundation, the next generation of professionals will lack the deep expertise needed to validate AI outputs, spot domain-specific errors, or manage crises where AI fails. Research confirms the cognitive cost: participants who use AI as a substitute for thinking lose cognitive abilities faster, while those who use it for explanation learn significantly better.
Pick your industry and the AI tools you use, and get news related to your work every day.
Free · 30 seconds with Google · unsubscribe anytimeWhat is AIToday? →
Ask AI anything about this article. The AI reads this article, earlier AIToday articles, and Wikipedia, and cites its sources. Q&As are published on this page for other readers too.
Vinci4D provides chip-design simulation software that trains on physics equations instead of customer data, an…
SEMI President and CEO Ajit Manocha told DIGITIMES that the semiconductor industry reached the US$1 trillion r…
CoreWeave fell more than 25% this week, its debt yield briefly rose above 16% on Thursday, and SK Hynix droppe…
Nvidia's AI Enterprise lists at $4,500 per GPU annually, and paid subscriptions add supported software, securi…
Anthropic told investors this week that it is making 85% gross margins, a figure the columnist calls a cash co…
Modal Labs, an AI infrastructure provider, raised $750 million in a Series D led by Accel at a $15.75 billion…
