An AI agent will eventually take an unexpected action that is logged in perfect detail yet answerable by no one, and the board will ask who in the organization was responsible. This white paper introduces the Intent Architecture Stack, a three-layer organizational design framework (Context, Intent, Governance) that answers that accountability question before the incident occurs. It shows that five major governance frameworks (NIST AI RMF, the EU AI Act, ISO/IEC 42001, the Cloud Security Alliance's Agentic Trust Framework, and Microsoft's Responsible AI Standard) require agent-level accountability but do not specify the organizational design that makes it operational, and it provides that design, operationalized for Microsoft-first regulated enterprises and grounded in 2024 to 2026 incidents, litigation, and regulatory guidance.
Sougata Roy (Sun,) studied this question.