As AI agents become more integrated into applications, traditional authorization methods like JSON Web Tokens (JWTs) face significant limitations due to their static nature, which struggles to handle dynamic agent interactions and delegation complexities. These issues are exacerbated in agentic workflows where agents act autonomously and can delegate tasks, demanding a more adaptable authorization model. Policy Decision Points (PDPs) offer a more dynamic solution by evaluating access permissions in real time, allowing for granular, context-aware control that can adapt to changes in relationships and permissions. This approach, exemplified by tools like Open Policy Agent and Google Zanzibar, supports relationship-based access control, enhancing scalability, maintainability, and accountability by dynamically resolving access based on current contexts. Organizations are increasingly seeking these solutions to ensure robust, visible, and resilient access control, moving away from the rigid and fragile structures of token-based models, as the role of tokens evolves to become pointers to relationships rather than comprehensive permission payloads.