AI Agents as Policy Instruments in Education: Reducing School Abandonment through Adaptive Governance Systems

Abstract

This paper analyzes the deployment of AI agents within national education strategies aimed at improving learning outcomes and reducing early school abandonment. It conceptualizes AI agents not only as educational technologies, but as policy instruments capable of reshaping governance, resource allocation, and early-warning mechanisms in public education systems. The study focuses on the use of AI agents for student risk detection, personalized learning pathways, administrative burden reduction, and real-time policy feedback loops. Drawing on policy design and socio-economic analysis, the paper highlights how AI agents can identify structural drivers of dropout, such as poverty, regional inequality, and institutional rigidity, earlier and more accurately than traditional administrative tools. At the same time, it examines key risks related to data bias, algorithmic labeling of vulnerable students, and the potential reinforcement of socio-economic stratification. The paper proposes a governance framework for educational AI agents based on transparency, pedagogical alignment, and social safeguards, ensuring that automation complements rather than replaces human judgment. The central argument is that AI agents, when embedded in inclusive and accountable policy architectures, can function as preventive governance mechanisms, shifting education policy from reactive remediation to proactive social investment.   Keywords: Artificial intelligence agents; education policy; school abandonment; early warning systems; personalized learning; public sector digitalization; social inequality