By 
CData Software has launched Connect AI Gateway, a new control layer designed to govern how AI agents and employees interact with enterprise applications and data. The platform aims to address a growing enterprise challenge: as AI moves from generating answers to executing tasks, organizations need consistent controls over permissions, context, model selection, cost and the actions agents take.
Moving AI From Answers to Actions
The Gateway builds on CData’s existing data connectivity layer and its library of schema-aware connectors. Rather than treating enterprise applications as generic data sources, it uses information about schemas, relationships, objects and operations to expose business systems as governed tools for AI agents.
That approach is particularly relevant as enterprises deploy agents across CRM, ERP, databases, data warehouses and legacy systems. CData says Connect AI Gateway can expose these systems through an enterprise Model Context Protocol (MCP) layer, while also bringing customer-managed MCP servers under the same governance framework.
The company is also positioning enterprise context as something that should persist independently of individual AI models. Organizations can provide metric definitions, business terminology and semantic models, while approved corrections can become reusable context for authorized agents and users.
For developers, this potentially reduces the need to construct separate integration and context-management layers for every agent. It also provides a way to keep business definitions consistent when different models or applications are used.
“CData Connect AI gives our agents live, governed access to enterprise applications like SAP, and onboarding a new application went from weeks to hours. Our test automation agent uses it to work inside SAP within the permissions we set and drive regression testing autonomously. Test delivery went from 10 weeks of manual effort to about one, a 10x acceleration. Where CData is taking this with the Connect AI Gateway, with policy reaching every agent action down to the record, is exactly what we need as more agents come online,” said Padmakar Pasala, AI Software Lead, Adobe.
Governance Extends to the Record Level
Security and accountability become more complicated when an AI system can modify enterprise data rather than simply retrieve it. CData says the Gateway carries user or agent identity through each request and applies permissions across the model, tools and underlying data sources, including record-level entitlements.
The platform also validates changes against the business rules of connected systems before committing them and maintains an audit trail linking an interaction to its prompt, model, tool and applicable policy.
Cost is another part of the architecture. Schema-aware tools can filter, join and aggregate information at the source before sending relevant results into a model’s context window. Requests can also be routed to models based on the complexity of the task rather than automatically using the most expensive model.
CData’s internal research reports a 98.5% correct-response rate for prompts using its connectors across CRM, project management, data warehouse and ERP environments, compared with 65%–75% for the competing MCP approaches it tested. In another study involving 22 models, the company reported that governed tools produced the same correct answer across models, while the least expensive model cost 178 times less than the most expensive one. These figures come from CData’s own research and should therefore be considered vendor-reported results rather than independent benchmarks.
The broader significance is the separation of governance and context from any individual model or data platform. CData says its context layer can work with frontier, open-weight and locally hosted models, allowing enterprises to change models without rebuilding the surrounding business context.
That flexibility could become increasingly important as cloud-native and enterprise AI architectures become more heterogeneous. Instead of allowing each agent framework or model provider to establish its own path into corporate systems, organizations can place a common governance layer between AI and the applications where business processes actually execute.
Connect AI Gateway entered an early-access program on September 29, 2026. As enterprises move toward autonomous AI workflows, the focus is shifting from whether agents can access enterprise data to whether that access can be controlled, audited and economically managed at scale.
