We are thrilled to announce the introduction of Agentic Data Intelligence as part of the watsonx.data intelligence SaaS offering on IBM Cloud and AWS Mumbai.
This release includes two main capabilities that are delivered together:
- Managed MCP Server (Model Context Protocol Server) which exposes governed data intelligence capabilities to external artificial intelligence agents.
- Data Intelligence Chat Agent, embedded within the watsonx.data intelligence user interface.
Both components are available as part of the existing watsonx.data intelligence experience.
What was holding teams back
Enterprise data platforms already manage large volumes of metadata, including catalogs, lineage, data quality metrics, business glossary terms, and governance policies. Despite this, two recurring challenges remain.
First, users often struggle to work efficiently with data intelligence tools. Identifying relevant assets, understanding data quality, or assessing downstream impact typically requires navigating multiple interfaces or relying on data experts.
Second, artificial intelligence agents operating on enterprise data often lack governed context. Without direct access to authoritative metadata and semantic definitions, agents may make incorrect assumptions about tables, columns, and relationships.
At the same time, data and analytics workflows are increasingly incorporating agent-based automation, where artificial intelligence systems are expected to retrieve context, reason over it, and perform tasks programmatically. Agentic Data Intelligence was introduced to support both interactive use and agent-based workflows using a shared, governed metadata foundation.
What Agentic Data Intelligence provide
Agentic Data Intelligence exposes watsonx.data intelligence capabilities through two complementary access paths: programmatic access for artificial intelligence agents and conversational interaction for users.
Programmatic access for AI agents
The watsonx.data intelligence Managed MCP Server exposes governed metadata as structured, callable tools using the open Model Context Protocol.
Through this MCP server, AI agents can:
- Find and understand data across the enterprise catalog
- Automate data governance/management workflows
- Conduct lineage analysis for impact and root cause assessment
- Query data products to get answers to business questions
- Manage the lifecycle of data products
The Managed MCP Server is hosted by IBM and is included with the watsonx.data intelligence SaaS entitlement. An open-source implementation of the server is also available for organizations that choose to self-host.
Conversational access to governed data
The Data Intelligence Chat Agent is embedded directly within watsonx.data intelligence and provides a conversational interface for interacting with catalog metadata and governance information.
Common interactions include:
- Searching for data assets using natural language
- Retrieving business context, ownership information, and data quality metrics
- Exploring lineage and upstream or downstream dependencies
- Querying business glossary terms and documentation
- Guiding new SaaS trial users through initial platform usage
This capability is intended for data consumers and data stewards who want faster access to trusted data context without deep familiarity with the product interface.
Operational outcomes
By introducing conversational and agent-based access to the data intelligence layer, this capability supports several outcomes:
- Reduced dependence on manual catalog navigation
- Faster access to lineage, quality, and ownership information
- Consistent enforcement of existing access controls and governance policies
- A standardized integration surface for artificial intelligence agents
These outcomes are achieved without duplicating metadata pipelines or deploying separate governance systems for users and automated agents.
Illustrative usage scenarios
Data discovery for analysts without deep catalog knowledge
A business analyst needs to identify datasets related to customer transactions that meet data quality requirements. Instead of navigating multiple catalog views and filters, the analyst uses the Data Intelligence Chat Agent embedded in watsonx.data intelligence to search using natural language.
The Data Intelligence Chat Agent returns relevant data assets along with ownership details, business context, and data quality information, all derived from existing catalog and governance metadata. This allows the analyst to locate appropriate data assets without requiring detailed knowledge of technical asset names or catalog structure.
Lineage-based impact analysis before data changes
Before making changes to a data pipeline, a data engineer needs to understand downstream dependencies. An artificial intelligence agent connected to the Managed MCP Server retrieves upstream and downstream lineage for the selected asset.
The agent identifies dependent tables, reports, and data products using authoritative lineage data from watsonx.data intelligence. This information supports impact analysis and changes planning without relying on manually maintained documentation.
Text-to-SQL generation grounded in enterprise metadata
An artificial intelligence agent is tasked with generating a query for analysis. Instead of inferring table names, columns, and joins, the agent uses the MCP Server to retrieve catalog metadata and generate SQL queries grounded in real enterprise schemas.
This improves consistency by ensuring that generated queries reference registered assets, defined relationships, and governed data structures rather than assumptions.
Next steps
Visit our webpage to learn more about watsonx.data intelligence.
Start your 30-day free trial to experience Agentic Data Intelligence.
Explore our Demo Library to see watsonx.data intelligence in action.