AICOS is a governance-first artificial intelligence architecture designed to integrate decision intelligence, risk governance, probabilistic reasoning, multi-objective optimization, robust decision-making, provenance, cryptographic integrity, and verifiable execution into a unified computational framework. The architecture distinguishes prediction from decision, explicitly represents uncertainty and constraints, evaluates decisions under alternative and adverse scenarios, and preserves decision provenance for auditability and replayability. This research presents the conceptual and engineering foundations of AICOS, including its decision intelligence fabric, governance mechanisms, risk-aware optimization framework, provenance architecture, and Veritas-based verifiable execution model. The work is intended as a research and engineering reference for the development of trustworthy, auditable, and governance-first AI systems capable of supporting complex financial, enterprise, and autonomous decision environments.