Rimini Street Outlines Enterprise AI Sovereignty Framework and Guardrails for Existing ERP

At an industry forum, Rimini Street Senior Customer Success Manager Chih-Ping Kao addressed enterprise challenges including poor visibility into models and agents, lack of abort controls, and workflow lock-in. He distinguished tool-centric autonomous enterprises from sovereign enterprises that govern action boundaries, outlining critical control dimensions: data permissions, workflow scope, human approval for decisions, technology modularity, and economic cost evaluation. Because core workflows typically run on ERP systems, Kao proposed a layered architecture that avoids forced upgrades: a Core Layer maintaining existing ERP and data platforms, a Control Layer managing policy and human-in-the-loop governance, an Action Layer executing cross-system tasks via AI Agents, and an Experience Layer providing role-based workspaces. To prevent loss of control, he introduced operational guardrails: least privilege, human approval for high-risk actions, comprehensive audit logging, fallback to manual workflows, rate limits with kill switches, and governance dashboards. Deployment should follow a phased progression, evaluating use cases by frequency, error controllability, measurable impact, redundancy, and auditability.

Key takeaways

• Identifies enterprise AI sovereignty controls covering data usage, workflow scope, human decision approval, modular technology, and measurable costs.
• Proposes a layered architecture atop existing ERP systems to avoid costly system replacement and forced migrations.
• Establishes operational guardrails including least privilege, human approvals, audit logs, manual fallbacks, kill switches, and dashboards.
• Recommends a phased rollout evaluating target workflows by frequency, error controllability, measurable outcomes, redundancy, and auditability.