RESEARCH2026
Introduction to Relational Architecture for Agentic AI Systems
A framing paper introducing the human-agent relationship as a first-class object of architecture and evaluation.
Abstract. This paper introduces relational architecture: an approach that treats the evolving relationship between humans and agents as a designed structure rather than as the ambient context around a model.
It defines the core objects of that architecture, principal, mandate, agent and action, and describes how authority moves between them over time. It then outlines the properties such an architecture must preserve: human-originated authority, bounded delegation, coherence between represented and actual behavior, and evidence adequate to justify what was done.