CData has expanded its Connect AI platform with native Model Context Protocol (MCP) connectivity inside Microsoft Copilot Studio and Microsoft Agent 365.
While not all organizations have adopted MCPs, they are emerging as a standard. Many enterprises now use Microsoft 365 Copilot to query and update ERP, CRM, and service platforms directly, far beyond its original role as a productivity assistant.
As this usage grows, existing integration infrastructure struggles to bridge fragmented SaaS and ERP data. MCP thus becomes a prerequisite for enterprisewide production use. CData Connect AI gives teams building agents in Microsoft environments the foundational infrastructure needed to operationalize AI at scale.
The integration lets Microsoft-built AI agents act on live data across more than 350 enterprise systems, including NetSuite, Salesforce, SAP, Snowflake, and ServiceNow.
Connectivity, Context, and Control
Enterprise AI agents fail for predictable reasons. They cannot reliably access systems of record; they lack usable context about the data they retrieve, or they operate outside established security and governance boundaries. Each limitation is manageable in isolation. Together, they prevent agents from moving beyond pilots into production.
This is the gap CData Connect AI is designed to address inside Microsoft’s agent ecosystem. The platform presents a MCP layer that allows agents built in Copilot Studio and Agent 365 to interact with enterprise systems in a consistent, governed way.
Connect AI provides a single MCP platform that exposes every API endpoint and version across enterprise applications through one maintained toolset. This allows agents to operate across platforms. Multisource queries run through a unified interface that manages schema translation, protocol differences, pagination, and query optimization.
Context, meanwhile, determines whether an AI agent retrieves data or understands it. Connect AI provides that understanding by exposing schemas, metadata, entity relationships, and business logic directly from source systems. Agents can read, edit, and track changes in files alongside structured data, treating both as first-class inputs.
As AI agents move into production, governance becomes an operational requirement. An identity-first model, implemented through Connect AI, preserves source system permissions as agents operate, allowing existing RBAC, OAuth, and SSO policies to flow through unchanged. AI-specific controls limit agent actions, with full auditability.
Amit Sharma, CEO of CData, said, “We’re eliminating the three barriers that have prevented enterprises from deploying truly intelligent agents at scale.”
He explained that Connect AI “doesn’t just connect to data sources—it teaches AI agents the schemas, relationships, and business logic native to each system, enabling sophisticated multisource analysis that was previously impossible”.
What Separates Useful Agents
AI agents derive their value from the data they can access and understand.
As agents move closer to systems of record, that dependency becomes structural. Productivity gains give way to operational impact, and tolerance for stale data, partial context, or opaque access controls disappears.
Sabin Nair, group product manager at Microsoft, described the impact of Microsoft’s integration with CData’s Connect AI as “empowering,” noting that customers could build agents in Copilot Studio that connect seamlessly to hundreds of enterprise data sources, while allowing IT teams to retain full visibility and control through Agent 365’s governance.
The growing emphasis on standards governing how agents interact with enterprise systems reflects a deeper structural change.
As conversational interfaces become execution layers, integration becomes core infrastructure. The long-term implication is a reordering of enterprise software, where access to live data and usable context matters as much as control.
What This Means for ERP Insiders
ERP access is shifting from screens to agents. That change turns conversational tools into operational entry points, where latency, permissions, and data integrity directly affect business outcomes. Integration and governance become prerequisites for scale.
Integration quality now determines AI value. As agents move into production, usefulness depends less on model sophistication and more on whether enterprise systems are connected, contextualized, and governed at scale.
Standards now matter more than tools. Scalable platforms will be defined by open, consistent standards for data access and control. ERP leaders should view MCP-style connectivity as foundational infrastructure.
