Introducing CData Connect AI for ChatGPT

Interact with your enterprise data — warehouses, databases, lakehouses — through conversation, with accurate context and centralized controls.

Every day, analysts and business users run into the same three blockers. The first is access: data lives in warehouses, databases, and lakehouses, and other enterprise sources that ChatGPT can't reach, so users export CSVs, paste summaries, and get answers that are already behind the actual data. The second is structure: even when an AI model can reach enterprise data, it sees raw schema — tables named for how storage was designed, not for how questions get asked. Metric definitions vary by team. Relationships span joins no end user should have to know about. A model working from raw schema doesn't answer the business question; it answers the database question. The third blocker is governance: giving an AI model broad access to enterprise data without row-level filtering, access policy enforcement, and a full audit trail is an IT non-starter — and it should be.

Today, we're addressing all three. We're announcing CData Connect AI for ChatGPT, available now in the ChatGPT app directory. Connect AI gives ChatGPT governed, real-time access to your enterprise data — cloud data warehouses like Snowflake and BigQuery, relational databases, data lakehouses, and hundreds of enterprise systems. The connectivity spans your sources without moving data. The context is defined by you: virtual datasets and MCP toolkits built around how your business asks questions, not how your storage was designed. The control stays with IT: access policy enforced at run time, with a full audit trail on every result returned.

Add Connect AI from the ChatGPT Directory
Start your free Connect AI trial

What is the ChatGPT app directory?

The ChatGPT app directory is a curated ecosystem of third-party integrations built on MCP, accessible directly inside ChatGPT, that extends conversations with live data and real-world actions. OpenAI's intent is to shift ChatGPT from a question-answering tool into a platform where people move from ideas to action — with apps surfaced at the right moment, in the flow of the conversation.

Connect AI brings your enterprise data into ChatGPT — on your terms

Business teams need ChatGPT to understand their data well enough to answer the question they actually asked — and IT needs to know exactly what got returned and to whom.

Connect AI makes that possible because it sits between ChatGPT and your data sources — federating queries in real time, enforcing access policy, and returning results in the shape your business actually uses, not the shape your database happened to store them in.

Most enterprise data doesn't map cleanly to how questions get asked. Metric definitions vary by team. Relationships span tables that no end user should have to mentally join. Connect AI lets you define virtual datasets across sources, encode those relationships, and attach semantic context — so when a user asks, "how did feature adoption track against our Q1 release schedule," ChatGPT is working with a curated, governed view of that data, not raw schema.

What this looks like in practice:

• "Which customer segments dropped in engagement after recent release changes?" ChatGPT analyzes 3,200 active accounts across your customer data, correlates the drop with the v4.2 reporting API change on March 3rd, and surfaces which tiers were affected — and which weren't.

• "The March close is $2.3M below forecast. Where is the variance coming from?" Connect AI federates across your deal and contract records; ChatGPT identifies the specific drivers — missing contract execution dates, professional services shortfall — and how many deals each affected.

• "Which suppliers drive lead time variability across our top 10 products?" ChatGPT pulls six months of procurement and inventory records, ranks suppliers by standard deviation, and flags the four driving 68% of total variability.

How it works

Connect AI surfaces your enterprise data to ChatGPT as purpose-built MCP tools based on the standardized relational layer CData applies to every connected source. In Connect AI, administrators define virtual datasets that span one or more underlying sources, configure the relationships between them, and apply business definitions that contextualize the data for a language model. Governance is enforced by authenticating users to the source system, layering additional downscoped permissions, and exposing tightly defined datasets and tools as unique MCP Servers.

When a user asks a question in ChatGPT that involves enterprise data, Connect AI handles the query federation — reaching the underlying sources, joining across them if necessary, applying row- and column-level filters based on the user's access profile, and returning results. The user gets an answer. The data never leaves its source. Every interaction is logged with a full audit trail: who asked, what was returned, when.

IT administrators control all of it. Access policy is defined in Connect AI and enforced at run time — not delegated to the user or the AI model.

Example prompts to get started:

• "Pull churn rate by contract tier from Snowflake and cross-reference against support ticket volume from the warehouse — are high-volume accounts churning at a higher rate?"

• "What does 90-day retention look like for users who completed onboarding versus those who didn't, across our three regional databases?"

• "Combine inventory levels from the ERP with sales velocity from the data lakehouse — which SKUs are at risk of stockout in the next 30 days?"

• "Compare average deal cycle by industry vertical using CRM data alongside the product usage data in BigQuery."

Getting started

Add CData Connect AI from the ChatGPT app directory, connect your data sources, and define the virtual datasets and MCP toolkits your teams will need. Define access controls, audit logging, and semantic configuration in Connect AI — all before the first question gets asked in ChatGPT.