The Announcement
CData Software has made three executive appointments aimed squarely at the enterprise AI data access market. Raviv Levi joins as Chief Product and Technology Officer, bringing experience from Sift and Cisco’s Security Business Group. Amit Naik takes the role of Vice President of AI Architecture, arriving from Calix with prior stops at PayPal and Oracle. Craig Sanchez steps in as Senior Vice President of Embedded Sales, previously with Vectara, Elastic, and Cloudera. Together, the hires signal that CData is positioning its platform as the connective tissue between enterprise data estates and the AI systems, both conversational and agentic, that increasingly need to query and act on that data in real time.
The Bigger Picture
The Bottleneck Nobody Talks About Enough
The executive commentary in this announcement gets something right that a lot of AI platform vendors get wrong. The model layer is not the hard problem anymore. Inference is fast, models are capable, and orchestration frameworks are maturing quickly. The hard problem is the one Levi named directly: responsible, live, governed access to enterprise data, across hundreds of fragmented sources, with the semantic context that makes AI responses accurate rather than plausible-sounding.
This is not a niche concern. ECI Research found that 50.7% of organizations rely on public AI tools such as ChatGPT and Copilot, while only 20.2% report enterprise-wide AI deployments built on a governed framework. That gap represents tens of thousands of organizations using consumer-grade AI tools against enterprise data without the governance rails that auditors, regulators, and security teams require. CData’s pitch is that it fills precisely that gap: live connectivity to 350-plus sources, semantic intelligence for context-aware responses, and built-in controls at every AI-to-data interaction.
What This Means for ITDMs
For IT decision-makers, the relevant question is not whether to use AI against enterprise data. That decision is largely made, or is being made right now without you. The question is whether it happens with or without a proper data layer underneath it.
CData’s model is differentiated in one specific way: it avoids complex pipeline infrastructure by providing live access rather than replication-first architectures. That matters because traditional ETL and data warehousing pipelines introduce latency, synchronization debt, and governance gaps that are tolerable for batch analytics but become genuine liabilities when an autonomous agent is taking action on stale data. For sectors with strict audit requirements, including finance, healthcare, and pharmaceuticals (CData’s customer list includes GSK), the ability to enforce governance at the connectivity layer rather than retrofitting it onto a pipeline is a meaningful architectural advantage.
The embedded sales hire deserves attention from a procurement standpoint as well. Sanchez’s mandate is to get CData’s connectivity layer into third-party AI-powered applications, meaning the data governance controls may arrive through a vendor’s product rather than through a direct CData deployment. ITDMs evaluating AI-enhanced SaaS applications should ask vendors whether their data access layer carries enterprise-grade governance or relies on the customer to build it separately.
What This Means for Developers
For developers and platform engineers building agentic AI systems, the Model Context Protocol (MCP) positioning is the most technically interesting element of this announcement. CData claims Connect AI is the industry’s first fully managed MCP platform, with integrations spanning Anthropic Claude, OpenAI ChatGPT, Microsoft Copilot Studio, Azure AI Foundry, and Agent 365. MCP is emerging as a candidate standard for how AI agents discover and interact with external data and tool surfaces, and a fully managed MCP layer abstracting 350-plus enterprise data sources is a meaningful accelerant for teams trying to build production-grade agentic workflows without hand-rolling every connector.
The operational complexity argument is real. According to ECI Research’s 2025 AI Builder Summit survey, two-thirds of enterprise AI leaders have already implemented multi-agent collaboration in live or pilot workflows, yet 44% of enterprise AI leaders have only moderate confidence that AI agents can act autonomously without human intervention. That confidence deficit is not primarily about model quality. It is about whether agents can reliably access the right data, with the right context, under consistent governance conditions. A managed data layer that handles connectivity, semantic grounding, and access controls addresses the operational substrate of that confidence problem directly.
Teams evaluating CData’s platform should probe three things in particular: how semantic context is generated and kept current as source schemas evolve, how fine-grained the governance controls are at the individual query or agent-action level, and what the latency profile looks like under concurrent agentic workloads across heterogeneous sources.
What’s Next
The Governance Gap Becomes a Compliance Event
The next twelve to eighteen months will likely see regulatory and procurement pressure force a sharper separation between AI deployments that have a proper data governance layer and those that do not. ECI Research found that 78.3% of surveyed organizations are subject to industry regulations such as HIPAA or GDPR, and as AI agents move from pilot to production, the question of how those regulations apply to autonomous data access will move from theoretical to immediately practical. Vendors that can demonstrate auditability at the data layer, specifically logging which agent accessed what data, under what permission context, and at what time, will have a compliance story that resonates with legal and risk teams, not just engineering.
MCP as Infrastructure
The broader significance of CData’s MCP positioning is that it treats MCP not as a feature but as infrastructure. If MCP achieves the kind of standardization that REST achieved for web APIs, the managed MCP platform category could become as foundational to agentic AI stacks as API gateways became to microservices architectures. That is not a guaranteed outcome; standards competitions in enterprise software are rarely clean. But the trajectory of adoption across the major AI platforms that CData has already integrated with suggests the standardization bet is a reasonable one. Organizations building agentic AI systems today should be designing data access architecture with that potential standardization in mind, rather than committing to bespoke connector infrastructure that may need to be replaced.
Snowflake AI Platform: Security, Microsoft & Multimodal
AI Is Stressing Open Source Infrastructure | ECI Research
AMD Ryzen AI Halo: The Case for On-Device Agentic AI
GitLab 19.0: Agentic DevSecOps and the AI Paradox
CData Bets on the Enterprise AI Data Layer | ECI Research
