OmniBank Agents Architecture (v1.1): An AI Agent Operating System for Regulated Banking

This white paper presents OmniBank Agents, an AI Agent Operating System (AOS) for regulated banking. The architecture separates probabilistic reasoning from deterministic execution by introducing a centralized Orchestrator (control plane), departmental Manager Agents (policy-grounded reasoning with constrained retrieval), and Tool Agents (permissioned API execution). The system is designed for bank-grade governance: approval gating, policy-as-code checks, PII/PCI handling, scoped retrieval, immutable audit logs, and model risk management. The paper details the component responsibilities, work-order boundary, event-driven substrate (Kafka), data systems (PostgreSQL, vector store, object storage), threat model, workflow case studies (KYC, fraud, AML, lending), and an evaluation approach focused on both model quality and control effectiveness.Brand: AIinovateHub, (LinkedIn newsletter: AI Vanguard).Related video: https://www.youtube.com/watch?v=nMdX96puPxM