Data integration firm CData has introduced its Connect AI Gateway, a new platform extension designed to serve as the control point between AI agents and enterprise IT systems, as companies struggle to demonstrate return on investment from their AI initiatives. The Chapel Hill, North Carolina-based company, which develops data connectivity and integration technology, built the Gateway as an extension of its Connect AI MCP (Model Context Protocol) server that launched in September 2025. The new offering aims to address stalled AI projects facing high costs, disappointing productivity gains, and insufficient visibility into how AI tools are actually being used.
The Gateway creates a single control point for managing the models, tools, data, and actions that AI agents and their human collaborators access across an organization. It connects agents to enterprise systems through governed tools, applies company context at each step of execution, enforces existing permissions, and routes requests to whichever model handles each task most efficiently. A new "context engine" gathers, creates, and stores semantic definitions—including business terminology, metrics, and existing semantic models—for AI-guided processes to use, giving the Gateway better understanding of the schemas, objects, relationships, and operations within connected systems. The context engine includes a self-learning loop that continuously improves performance by making data corrections and adjusting prompt patterns within agentic memory, and the system logs every step of execution from AI down to source systems.
Marie Forshaw, CData's senior vice president of product marketing, told CRN that a common theme in the market shows companies rapidly adopting AI only to ask themselves, "Is this helping? Is this increasing my productivity? Where is the ROI of everything I've invested in AI?" Many organizations are coming up short on answers to those questions, she said. According to Forshaw, the lack of AI visibility, observability, and governance capabilities is at least partly responsible for the less-than-expected productivity, because fearful organizations are limiting AI's access to corporate systems and sensitive data. "We see this as a way to go the last mile in terms of AI governance and in terms of connectivity for AI initiatives," she said about the new offering.
The Gateway addresses what CData identifies as a central problem: companies need better control over AI agent behavior without blocking access to the enterprise systems and sensitive information that make AI productive in the first place. Organizations that restrict AI access to protect data end up undermining the productivity gains they purchased AI to deliver, creating a governance-versus-performance trade-off that leaves executives unable to justify continued investment. By inserting a governed layer between agents and systems, the Gateway lets companies track exactly how AI uses their data, enforce existing security policies automatically, and measure actual productivity impact—turning AI from an opaque cost center into a visible, controllable tool. CData built its technology into offerings from Google Cloud, BigQuery, and Palantir, and this year released a free edition for AI developers alongside the expanded platform capabilities. The Gateway is currently available through an early access program.
For channel partners and enterprise buyers, the strategic question isn't whether to govern AI—it's whether governance infrastructure can be retrofitted onto existing deployments or must be architected from the start. Organizations that wait to address visibility and control may find themselves forced to choose between abandoning sunk costs or accepting ungoverned risk indefinitely.