Some enterprise vendors are converging on a new priority: controlling how AI agents access data, execute workflows, and operate across business systems.
Oracle NetSuite on March 31 announced the expansion of its AI Connector Service to let customers connect external AI assistants to ERP data in a governed, role-based framework. Similarly on March 27, OpenAI introduced a plugin system for Codex that allows enterprises to package, distribute, and restrict AI agent behaviors across development environments.
Taken together, the announcements point to a shared architectural direction, where AI is being treated as an external layer that must be tightly controlled at the point of integration.
Analysis
What this means: Governance is a control plane for enterprise AI. As organizations deploy multiple AI models and assistants, the ability to define, enforce, and audit how those agents access ERP data and execute workflows will determine enterprise readiness.
NetSuite: AI Access with Governance Layer
NetSuite’s AI Connector Service updates center on making third-party AI tools enterprise-aware.
The new AI Connector Service Companion introduces structured prompt libraries, reusable “skills,” and role-based access controls aligned to finance and operational workflows. These capabilities are designed to standardize outputs and reduce reliance on ad hoc prompting.
NetSuite is also extending its Model Context Protocol (MCP) support with embedded application experiences. MCP Apps bring familiar NetSuite interfaces, such as report and record selectors, directly into AI assistants, enabling more structured and auditable interactions.
Expanded support for NetSuite Analytics Warehouse further broadens the scope, allowing AI to operate across transactional, historical, and third-party data. This positions NetSuite as a governed data layer for AI-driven analysis and execution.
OpenAI: Plugin Governance for AI Agents
OpenAI’s latest Codex update approaches the same challenge from the AI platform side, introducing a plugin system that allows enterprises to standardize and control how AI coding agents operate.
Plugins function as installable bundles that package workflows, integrations, and tool access into reusable units. These can be distributed across teams or restricted through centralized policy controls, giving IT administrators the ability to manage agent behavior at scale.
Organizations can define plugin catalogs with explicit policies, including default installation, optional availability, or full restriction. This creates a governance layer where agent capabilities, integrations, and access can be versioned, audited, and enforced.
The approach reflects growing enterprise demand for consistency and control as AI adoption expands beyond experimentation into production environments.
Analysis
What this means: Agent behavior is being standardized and distributed at scale. OpenAI’s plugin model and NetSuite’s reusable skills both point to a future where AI interactions are packaged, versioned, and deployed across teams, replacing inconsistent, prompt-driven usage.
Converging on a Control Layer for AI
NetSuite is defining how AI accesses and interacts with ERP data, workflows, and roles. OpenAI is defining how AI agents themselves are configured, distributed, and controlled across the enterprise. Both approaches establish a control layer that standardizes behavior and reduces risk.
This is a shift toward treating AI agents as managed infrastructure rather than ad hoc tools. By packaging workflows and access into controlled artifacts, organizations can align AI usage with existing IT governance models, particularly around security, compliance, and operational consistency.
Analysis
What this means: ERP systems are becoming controlled data and workflow hubs. NetSuite’s connector strategy reflects a broader shift in ERP architecture, where value comes from governing how external AI interacts with enterprise processes rather than embedding all intelligence within the application itself.
