Agentic AI: Designing for Autonomy Without Losing Control
Inès Bedar, Desroches, Clément · HAL (Le Centre pour la Communication Scientifique Directe) · 2025
This article explores AI Agents as a strategic and technical inflection point in the evolution of artificial intelligence. While generative AI has revolutionized content creation and assisted workflows, the rise of AI agents -systems capable of reasoning, planning, acting, and collaborating marks a new paradigm defined by autonomy, orchestration, and continuous adaptation. Drawing from technological breakthroughs, emerging middleware standards, and geopolitical shifts, we analyze how Agentic AI is reshaping competitive advantage, enterprise architecture, and global regulatory priorities. To navigate this shift, we propose a dedicated risk assessment framework designed to address the unique challenges posed by agentic systems. This framework moves beyond static regulatory compliance, offering organizations a practical tool to evaluate and manage the risks tied to autonomy, from system design to sustainability.