AI’s promise in the enterprise has always come with a frustrating caveat: most projects never move beyond pilots. Research from MIT finds that 95% of AI pilots fail, largely because of challenges in accessing and governing enterprise data. Billions are wasted every year on proofs of concept that can’t scale. Manish Patel, Chief Product Officer at CData, believes the problem is finally being solved. In our conversation, he explained how Connect AI, a new SaaS application from CData, enables governed, real-time connectivity to hundreds of enterprise systems. “In this agentic world, connectivity is more important than ever before,” Patel said. “Without it, you can’t deliver reliable or secure outcomes.”

Governed, Real-Time Access

Patel contrasted Connect AI with legacy integration approaches, which often leave departments cobbling together their own connectors with service accounts. Connect AI centralizes this into a single governed layer. Users authenticate with their own credentials, permissions flow through to the AI layer, and enterprises gain the ability to audit, manage, and restrict what agents can see or do.

“The multi-layer protection is key,” Patel explained. At the user level, permissions are inherited from the source application. At the agent level, enterprises can configure guardrails—such as restricting an agent to read-only access or to specific objects within Salesforce. This ensures AI tools can analyze without introducing new risks.

Just as important is the real-time access. Patel emphasized that stale data can lead agents to make wrong decisions or generate incorrect insights. Unlike warehouse replication, Connect AI enables direct queries to live systems, ensuring accuracy for AI-driven workflows.

Practical Enterprise Use Cases

The shift isn’t theoretical. Patel shared an example of a construction company that connected its CRM and ERP to Connect AI and layered in a conversational interface with Claude AI. “For the first time, their GM could answer ad hoc questions in real time during quarterly reviews without running reports,” he said. Instead of waiting on BI dashboards, executives could simply ask questions and get live answers across multiple systems.

ISV partners are also embedding Connect AI. Patel highlighted one that helps enterprises migrate from legacy ERP platforms to NetSuite. By embedding Connect AI and MCP, they enabled both conversational interfaces and agentic workflows across migration tasks—without building their own orchestration stack.

MCP: The New Standard

“Underpinning Connect AI is the Model Context Protocol (MCP), which Patel described as ‘the new ODBC (Open Database Connectivity) or JDBC (Java Database Connectivity) for the AI era.’” Originally launched by Anthropic, MCP has quickly gained adoption by major LLM providers. Patel sees it as a crucial standard: “It gives us a consistent interface for applications and models, and it fits naturally with our DNA as a connectivity company.”

With MCP as the backbone, Connect AI offers enterprises a way to move AI beyond demos and into production, without losing governance or security. Patel noted that the first wave of disruption will hit business intelligence and analytics, where agentic workflows will make dashboards feel obsolete.

The Road Ahead

CData isn’t stopping with Connect AI. Patel teased that the company is exploring how its deep domain expertise could power pre-packaged agents for CRMs, ERPs, and file systems—helping smaller enterprises jumpstart AI adoption.

For now, the priority is clear: break the data access barrier that has stalled enterprise AI for a decade. “Connectivity has always been our foundation,” Patel said. “Now it’s the foundation for enterprise AI.”