Secure RAG Architectures with Small Language Models for Governance-Aligned LLM Deployment in Enterprise Service Management Platforms

Abstract

Research identifies a previous lack of attention in the literature to the interrelated challenges of enterprise governance and knowledge automation. The discussion demonstrates that efficient use of enterprise knowledge assets is key for meeting governance objectives. Specialized governance-aligned systems help enterprise owners and board members meet their fiduciary accountability obligations. Such specialized systems both govern knowledge assets and operate using them. Seven key components ensure the effectiveness and integrity of knowledge automation in governance contexts. Multi-model retrieval offers significant advantages, especially in governance-related applications, and these advantages can be realized using orchestrated core models that integrate task-specific strengths of diverse retrieval models, including large language models. A high-level architecture satisfies both traditional enterprise security requirements and the additional security and robustness considerations that arise from the required trust calibration. Enterprises are now increasingly subjected to requirements to demonstrate responsible handling of sensitive information. A dedicated data governance and compliance layer supports monitoring of these requirements and map of compliance processes. The consistency and coherence of knowledge automation results can be further strengthened by integrating an internal evidence-based reasoning strategy.