Agentifying Agentic AI

ABSTRACT

The rapid popularization of the term 'Agentic AI' has led to semantic confusion, often blurring the lines between simple generative loops and true software agency. This position paper critically examines the current architectures, proposing a formal framework to 'agentify' LLMs. The authors outline the strict requirements for autonomy, persistent state, and goal-directed behavior necessary to qualify a system as a genuine multi-agent architecture.