AI’s double-edged impact on education threatens future workforce readiness

Artificial intelligence (AI) is rapidly reshaping classrooms and workplaces, but a new study warns that without careful planning, schools may be preparing students for skills that will soon be devalued.

Andrew J. Peterson, an assistant professor at the University of Poitiers, has published an urgent analysis of this emerging dilemma in his paper Training for Obsolescence? The AI-Driven Education Trap, submitted on arXiv. The research offers a sharp look at how AI’s dual role, enhancing education while simultaneously automating tasks, could leave education systems trapped in a cycle of misalignment between what students learn and the realities of the future job market.

How AI creates a double-edged sword in education

The study examines the dual channels through which AI interacts with education. On one hand, AI delivers unprecedented productivity gains in the classroom by making teaching more efficient and personalized, especially for routine and codifiable skills such as mathematics, basic coding, and standardized language tasks. On the other hand, these same skills are the most vulnerable to displacement in the labor market as AI systems automate increasingly complex tasks.

The author models the decision-making of a typical education planner who is focused on optimizing present-day learning outcomes without considering future wage effects. This narrow planning horizon, he argues, produces systemic inefficiencies. AI incentivizes schools to double down on the skills that are easiest to teach and test, while long-term returns for those skills erode as automation intensifies in workplaces.

The result is a growing mismatch between human capital development and labor market demand. As AI systems expand their capabilities, the risk of what Peterson terms an “education trap” becomes more pronounced, with curricula skewed toward skills that may no longer command value by the time students graduate.

The economics of the education trap

The dynamic model captures how education systems react as AI becomes more pervasive. It distinguishes between two skill categories: routine, codifiable skills labeled as Skill A, and complex, non-routine skills labeled as Skill B.

As AI capital rises, the model predicts that a naive education planner will allocate more time to Skill A because AI makes it easier to teach. In contrast, an informed planner, one who accounts for declining wages in the labor market, would shift focus toward Skill B, which is harder to teach but more resilient to automation. This divergence grows steadily as AI advances, creating what the study describes as a monotonic increase in misalignment between teaching practices and market needs.

The analysis also highlights the underinvestment in non-cognitive skills such as persistence, adaptability, and emotional regulation. These traits, which are essential for thriving in a rapidly evolving economy, are difficult to measure and less amenable to AI-driven teaching tools. As a result, they are systematically deprioritized, widening the gap between what students need and what they receive.

The study further extends its model to account for endogenous AI adoption within schools. When institutions have the autonomy to adopt AI tools, they tend to overuse them, favoring test score gains in routine competencies over a balanced skill portfolio. This over-adoption accelerates the very mismatch the study warns about, especially as the cost of deploying AI technologies falls.

Evidence of a growing mismatch

To ground the theoretical model in empirical evidence, the research incorporates a pilot study with 20 educators evaluating 90 occupational skills from the O*NET database. The findings reveal a clear and statistically significant correlation between the skills most amenable to AI-assisted teaching and those most exposed to automation in the labor market.

The educators consistently rated AI as less effective for teaching non-cognitive and interpersonal skills. Closed, procedural tasks, those that follow fixed rules or repeatable steps, were seen as ideal for AI-driven instruction but also as the most vulnerable to rapid automation. This feedback loop, Peterson argues, creates conditions where educational systems may unintentionally amplify future labor market vulnerabilities by focusing on the very skills most at risk of obsolescence.

The analysis also identifies a non-linear relationship in labor market returns. Basic levels of certain skills retain value as foundational knowledge, and advanced expertise remains in demand for roles requiring deep specialization. However, intermediate skill levels, the “middle” of the spectrum, are most susceptible to automation, creating what the study refers to as a substitution trap. In this scenario, education systems risk producing large cohorts of workers trained for roles that may vanish or undergo significant restructuring.

Policy directions for AI-resilient education

To avoid the pitfalls of the AI-driven education trap, the study calls for forward-looking interventions at multiple levels of the education system.

First, schools should be equipped with tools to forecast future wage trends, enabling data-driven adjustments to curricula. By integrating labor market foresight into planning processes, education systems can better align teaching priorities with evolving economic realities.

Second, there is a need to recalibrate incentives to emphasize the development of non-cognitive and interpersonal skills. Accreditation systems and funding models should reward programs that cultivate resilience, critical thinking, and adaptability—skills that remain harder for AI to replicate and that complement technological advancements.

Third, the study recommends a cautious approach to AI adoption. Rather than pursuing wholesale integration of AI technologies, institutions should evaluate tools based on their long-term impact on both cognitive and non-cognitive skill development. This includes setting guardrails to ensure that adoption is strategic and balanced, rather than driven solely by short-term performance metrics.

Finally, the research highlights the importance of differentiated pathways. By ensuring universal access to basic digital literacy while creating targeted tracks for advanced expertise, education systems can help students avoid the risks associated with the erosion of mid-level competencies.