As we close out 2025, we're taking a moment to look back at what's been an incredible year for CData's AI journey. In this final episode of Vibe Querying with MCP for 2025, hosts Stan and Cam reflect on the milestones, partnerships, and real-world impact that defined CData's evolution in the AI space.
Watch Now: Vibe Querying with MCP - Episode #12
The core message: AI success requires AI-ready data
2025 proved one fundamental truth: AI ROI isn't constrained by models – it's constrained by data access and context. Throughout the year, CData focused on solving this challenge through Model Context Protocol (MCP) technology and the Connect AI platform, providing enterprises with governed, real-time access to their business data.
This episode explores the timeline of CData’s releases, major ecosystem partnerships, impressive usage metrics, and key findings from our State of AI Data Connectivity Report – all demonstrating that access to AI-ready data is the true differentiator for enterprise AI success.
Introducing MCP, CData Connect AI, and Vibe Querying
Model Context Protocol (MCP) is a standard protocol designed to connect AI models to external data sources, tools, and workflows securely and efficiently. MCP enables AI models to access real-time data and interact naturally with that information through conversational AI or agent-driven actions.
CData Connect AI enables AI clients like Claude, ChatGPT, and Gemini to work with live data from 350+ data sources through a managed MCP platform. This combination gives AI models access to the work data they need to be genuinely useful in business contexts—moving beyond simple chatbots to provide genuine analytical intelligence and workflow automation.
Vibe Querying is a conversational approach to data exploration and analytics where you don't need intimate knowledge of data schemas or prebuilt pipelines. You simply ask questions in natural language, and the MCP server provides live access to data while the AI uses its knowledge to provide answers.
CData's 2025 AI timeline: From downloadable MCP to enterprise partnerships
Q2: Early foundations with downloadable MCP servers
May 1st: CData launches first batch of downloadable MCP servers
When Anthropic introduced Model Context Protocol in late 2024, it signaled that AI was moving beyond static prompts toward real tool-driven interaction with live business systems. At CData, this resonated immediately – MCP was essentially formalizing something we'd been thinking about for years: conversational analytics where AI doesn't just talk about data but can safely interact with it.
On May 1st, we released our first batch of downloadable MCP servers for popular data sources like Excel, SAP, QuickBooks, Shopify, and Google Drive. These weren't just analytical tools – they enabled full read-write capabilities, allowing AI to both query and take action with enterprise data.
Key milestones:
350+ downloadable MCP servers released across four batches
June 24th: Final batch of downloadable MCP servers released
3,000+ downloads within the first quarter
Critical learnings from real usage patterns directly informed Connect AI development
What mattered more than download numbers was how people were using these servers. We watched real usage patterns – what tools were getting called, where people hit friction, and what they expected to just work. That feedback directly shaped what became Connect AI.
Q3: Connect AI launch and Foundations 2025
September 24th: Official Connect AI release
During our annual Foundations event, CData launched the first managed MCP platform. This wasn't just a hosted version of our downloadable servers – it was a comprehensive platform designed for enterprise-scale AI data connectivity.
Connect AI addressed the key challenge we identified from our downloadable MCP users: enterprises needed remote MCP capabilities that could scale, maintain governance, and provide reliable uptime. The platform launched with:
Hosted MCP infrastructure eliminating local installation complexity
Enterprise-grade security and permission models
Unified interface for 350+ data sources
Production-ready reliability and performance
The rapid adoption validated our approach – enterprises were ready for managed MCP that could bridge their AI systems with real-time business data at scale.
Q4: Partner ecosystem growth and incremental releases
The fourth quarter brought continuous improvements that may not have made headlines but significantly enhanced the platform's capabilities:
Connection error URLs: When connection issues occur, error messages now include direct URLs to the Connect AI dashboard for immediate troubleshooting. IT teams particularly appreciated this streamlined error resolution.
Data source-specific instructions: This game-changing feature reduced tool usage by roughly 50%. Through a new getInstructions tool, AI models receive upfront context about data sources – common use cases, metadata information, table relationships, and frequently accessed tables. For example, when connecting to Salesforce, the AI already knows that Account, Opportunity, and Contact tables are most commonly used, along with their relationships.
Major ecosystem collaborations: Q4 culminated in collaborations with three of the biggest names in enterprise AI (Databricks, Microsoft, and Anthropic), positioning Connect AI as the connectivity layer for the AI ecosystem.
2025 partnerships: Securing CData's place in the AI ecosystem
Databricks partnership - November 6th
CData was named a Databricks Agent Bricks launch partner, with Connect AI featured in the Databricks Marketplace. This partnership embeds Connect AI into Databricks workflows, enabling enterprises to build and deploy AI agents with governed, auditable access to business data directly from within Databricks systems and beyond.
Microsoft collaboration - November 18th
The Microsoft collaboration announced Connect AI's integration with Microsoft AI agents and Copilot offerings. This partnership positioned Connect AI as a native connectivity layer for Microsoft's AI ecosystem, allowing enterprise-grade real-time data access across 350+ systems into Microsoft Copilot environments – all while maintaining critical governance and security requirements.
Anthropic certification- December 9th
Anthropic certified CData Connect AI within the Claude connector directory. This certification enables all the latest Claude models, including Claude Opus 4.5 and Claude Code, to access live enterprise data securely through Connect AI. The partnership expands Connect AI's ecosystem reach by making it a trusted, governed data layer for Claude-based workflows, helping enterprises connect Claude to their entire portfolio of 350+ systems without having to build custom connectors.
These three collaborations with Databricks, Microsoft, and Anthropic cement CData's position as the trusted connectivity partner that major AI platforms rely on to bridge their systems with enterprise data.
2025 key adoption and usage metrics
Connect AI 2025 query volume
The scale of Connect AI usage in 2025 demonstrates real production workloads, not just demo environments:
191,262,071 total tool calls executed through Connect AI in 2025
78,633,525 total AI queries run through the platform
850 billion rows of data moved through Connect AI this year
These aren't just lightweight requests – they represent high-volume production workloads running through partnership tools like Microsoft Copilot Studio, Claude, Claude Code, and Databricks Agent Bricks.
Connect AI 2025 top data sources
When we look at where this massive query volume originated, the diversity of data sources tells an important story:
NetSuite: 13,635,061 queries
Bullhorn CRM: 11,143,030 queries
SQL Server: 9,664,916 queries
MongoDB: 6,287,465 queries
QuickBooks Online: 4,579,474 queries
Snowflake: 3,975,772 queries
What's striking is the mix – ERPs, CRMs, databases, and cloud warehouses. Connect AI isn't concentrated in one category or vertical. It's being used across entire data stacks, demonstrating the universal need for AI data connectivity across diverse enterprise systems.
Connect AI 2025 outages
For IT teams and stakeholders evaluating Connect AI, one metric stands above all others:
Millions of queries, hundreds of billions of rows of data, zero outages. At this scale, reliability isn't just a footnote – it’s foundational for AI success.
State of AI data connectivity report: Key insights for 2026
On December 3rd, CData launched the State of AI Data Connectivity Report, surveying over 200 AI and data leaders to understand the current state and future outlook of AI data infrastructure.
The AI readiness gap
Only 6% of AI and data leaders say their data infrastructure is fully ready for AI.
The AI readiness gap isn't about model access – it's about having the data context and control needed to actually run AI in production. Most organizations have invested heavily in AI models while their data infrastructure remains unprepared for the demands of real-time, multi-source AI applications.
The integration time tax
71% of AI and data leaders report their teams spend over a quarter of AI implementation time on data integration.
When engineering time goes toward building connectors and pipelines rather than innovating with AI products, it represents significant overhead that directly impacts time-to-value. Every hour spent on data plumbing is an hour not spent on AI outcomes.
Multi-source connectivity is the default
46% of software providers need real-time access to over six sources for a single AI use case.
Integration complexity is exploding. Multi-source connectivity has become a default requirement, meaning each AI use case becomes an entire architecture problem. More systems create more failure points and more context fragmentation, making robust data connectivity infrastructure essential.
Real-time data is universally critical
100% of enterprises agree that real-time data is essential for AI agents.
AI agents can't act reliably on stale content. The effectiveness of real-time integrations will decide who scales agents successfully versus who stalls out. This unanimous agreement underscores that batch processing and periodic updates are insufficient for modern AI applications.
The maturity divide
60% of highest AI-maturity companies have advanced data infrastructure
53% of lower AI-maturity organizations are hampered by immature data systems
AI maturity follows integration maturity. Companies with sophisticated data infrastructure consistently demonstrate higher AI success rates. This correlation suggests that organizations serious about AI ROI should invest where the correlation is strongest: in their data connectivity and infrastructure.
Download the full report: State of AI Data Connectivity Report
Looking forward to 2026
The progression throughout 2025 is clear: we started with downloadable MCP servers to learn from real usage in the market, evolved that into Connect AI – the first managed MCP platform operating at production scale – and secured partnerships with the biggest names in enterprise AI.
The focus throughout the year was performance, reliability, governance, and usability. It was really about moving from possibility to maturity.
Looking ahead to 2026, the focus only deepens:
More intelligence built on real data: As AI capabilities expand, the quality and timeliness of data access becomes even more critical
More trust and control as AI becomes operational: Production AI deployments demand robust governance, security, and auditability
More ways to build real systems on top of AI: The ecosystem will continue expanding with new tools, frameworks, and integration patterns
This year showed what's possible. Next year is about pushing that further.
From proving ground to production scale
CData's 2025 journey in AI demonstrates a clear path from experimentation to enterprise adoption. The downloadable MCP servers proved the concept, Connect AI scaled it to production, and partnerships with Databricks, Microsoft, and Anthropic positioned it as the connectivity layer for the AI ecosystem.
The metrics speak for themselves: nearly 200 million tool calls, over 78 million AI queries, 850 billion rows of data moved, and zero outages. But more importantly, the usage patterns and partnerships demonstrate that enterprises are successfully deploying AI with real-time access to their business data – securely, reliably, and at scale.
As we move into 2026, the message remains clear: AI success requires AI-ready data – not just AI models.
Join the conversation
Want to stay connected with the Vibe Querying community and explore how others are using MCP for business intelligence?
Stay tuned for more Vibe Querying episodes in 2026. Until then, stay curious and keep vibing with your data!
Ready to transform your AI data connectivity? Explore Connect AI at cdata.com/ai/ and discover how governed, real-time data access can unlock your AI potential.
