What happened?

Nolan Lovett, a researcher at the NATO Special Operations University, explains in an article published in Human Resource Development Review how individually rational corporate AI decisions can erode the shared expertise of entire professions. Lovett adapts the concept of the "tragedy of the commons," defined by ecologist Garrett Hardin in 1968, to AI employment: when each company replaces entry-level positions with AI, it captures the full efficiency gain for itself, but the cost of eroding expertise spreads across all organizations drawing from the same talent pool.

According to Lovett, this erosion happens in two ways: AI directly eliminating entry-level jobs, and AI-assisted junior employees reaching productivity levels that once required experience without accumulating deep domain knowledge. The researcher links this to the "validation tether" problem: catching domain-specific errors in AI output requires exactly the deep expertise that AI use itself erodes.

Why does it matter?

Lovett argues that experienced professionals in today's market were trained 5 to 20 years ago, meaning the full impact of disruption to entry-level positions may not surface until between 2030 and 2045. He calls this delayed effect the "human reserve paradox": organizations need a deep reserve of expertise for validation, crisis management, and situations where AI falls short, but no single organization has the economic incentive to sustain it alone.

According to the research, the risk is not equal across all professions. Fields with high task substitutability, light regulation, and strong modularity — such as software engineering, financial analysis, and legal research — fall into the most vulnerable category. Medicine and engineering are partially protected by stricter regulatory requirements and strong professional bodies, but they are not entirely immune.

What we know

  • A study published in summer 2025 found declining employment among young workers in AI-exposed occupations, while employment among experienced workers remained stable or increased.
  • A Federal Reserve study found that growth in programming jobs has nearly halved since ChatGPT's launch.
  • A study from January 2026 found that the employment crisis in AI-exposed occupations began before ChatGPT.
  • Anthropic's March 2026 study found no measurable overall impact of AI on the labor market, but noted a half-point drop in the employment rate among 22-25 year-olds.
  • An MIT study based on EEG measurements found that even brief AI use weakened neural connectivity, and more than 80% of participants struggled to recall their own AI-assisted writing.

What's next?

Lovett does not propose banning or restricting AI use. Instead, he recommends preserving AI-free learning environments, gradual AI integration, and establishing a human performance baseline before AI is introduced. He also advises professional organizations to certify domain competence alongside AI skills, and calls on policymakers to make education and training more attractive.

In one Anthropic study, software developers with access to AI performed 17% worse on knowledge tests, with the biggest losses seen among those who used AI purely as an answer machine. In contrast, those who used AI as an explanation tool learned significantly better — a finding suggesting that how AI is used matters more than how much.