The term “agentic” has rapidly permeated the tech industry; most software vendors now tout AI agents with autonomous capabilities. We thoroughly reviewed current AI agents in production, analyzing their goals, their operational mechanisms, and the extent of their autonomous behavior. Close examination shows that these agents exhibit widely varying degrees of autonomy. We found five distinct categories of employee support AI agents, characterized by their level of model-driven decision-making and the actions they undertake to achieve their objectives. Tech leaders should review these categories and examples to avoid unwelcome surprises when selecting AI agents.