IBM watsonx.data intelligence brings trusted context to AI agents in self-managed environments

AI agents are rapidly becoming embedded in how organizations operate, automating tasks, analyzing information and accelerating decision-making. But their effectiveness depends on one critical factor: access to trusted, well-understood data.

Without a clear understanding of the underlying data and how it is defined, related and governed, agents are forced to rely on incomplete signals, increasing the risk of inaccurate results.

To address this, IBM is introducing Agentic Data Intelligence in watsonx.data intelligence, expanding its availability beyond the SaaS offering announced last April to self-managed deployments. This enables organizations to run it on their own infrastructure with full control over data, governance and security requirements, while bringing trusted business and operational context directly into AI agent interactions.

AI agents are most effective when connected to the enterprise knowledge layer, with access to business definitions, lineage, governance policies and data quality insights. Much of this information already exists but remains disconnected from where agents operate, limiting reliable interpretation and reducing trust in AI-driven workflows.

Agentic Data Intelligence is designed to meet this need by bringing watsonx.data intelligence capabilities into AI agent interactions. Using the Model Context Protocol (MCP), an emerging standard for connecting AI systems with enterprise data and tools, agents can dynamically access trusted metadata and governance context from watsonx.data intelligence at runtime.

This allows agents to move beyond raw data access and operate with a shared understanding of business meaning, relationships, ownership and policies. These integrations work across existing tools and applications, including IBM Bob, watsonx Orchestrate, Claude, GitHub Copilot, enterprise copilots and custom AI applications.

With this model, agents can access:

• Cataloged data assets
• Business definitions 
• Lineage relationships and dependencies
• Data products and ownership information
• Governance policies and controls

As a result, agents can discover relevant data faster and produce outputs that are more accurate, explainable and aligned with enterprise standards.

While trusted context is essential, many data intelligence and governance tasks require structured processes and best practices to be executed effectively. Agentic Data Intelligence addresses this through Agent Skills, which enable agents to act on data consistently.

Agent Skills provide AI agents with reusable workflows for completing common data tasks. Rather than determining each step independently, agents follow predefined sequences that reflect how this work is carried out in practice and can be applied across similar scenarios.

These workflows cover activities such as:

• Data product creation
• Data cataloging and metadata enrichment
• Lineage analysis and impact tracing

Many of these tasks involve multiple tools, dependencies and ordered steps. Without clear guidance, execution can become inconsistent, with steps missed or performed out of sequence.

Agent Skills define how this work should be executed, including step-by-step actions and the underlying best practices. Because these workflows are reusable, they can be applied repeatedly, improving consistency while reducing the need to design or coordinate processes each time.

Tasks that would otherwise require coordinating multiple tools are handled through a single, guided workflow, reducing manual effort and enabling more predictable outcomes.

When a user interacts with an AI agent used within the organization, Agentic Data Intelligence enables that agent to use operational context from watsonx.data intelligence to interpret and respond to the request.

The agent identifies relevant data assets and business terms, retrieves lineage and data quality signals to understand relationships and reliability, and verifies ownership and governance policies before generating a response. For common and recurring tasks, Agent Skills guide how the agent performs the workflow, applying predefined steps and established practices.

Because this happens at runtime, agents operate on current, governed data, not static or duplicated copies. This makes data governance, lineage, quality, and data products part of how AI agents interpret requests and produce results.

With access to semantic understanding, AI agents can support key data and governance workflows across the organization.

• Regulatory audit readiness: Data stewards can ask agents for business context, quality metrics, and lineage information for a given asset. This allows them to respond to questions and prepare for audits without navigating multiple systems.
• Governance workflow automation: Data stewards can onboard new data sources through a single interaction, enriching metadata, classifying sensitive data, and applying governance policies.
• Lineage-driven impact analysis: Data engineers can ask an agent to analyze lineage and assess upstream and downstream dependencies before making changes, reducing risk and improving decision-making.

To see how AI agents interact with governed metadata through the MCP Server, watch the demo below.

Agentic Data Intelligence is available in watsonx.data intelligence for self-managed deployment, giving organizations full control over their data, governance and security. This is particularly relevant in regulated or security-sensitive industries.

At the same time, by making operational context available to the AI agents already in use, teams can extend existing workflows without redesigning their foundations. This accelerates adoption, reduces integration overhead and helps organizations realize value faster.

The next phase of enterprise AI will depend not just on better models, but on how effectively those models use trusted organizational knowledge.

Agentic Data Intelligence makes governance, lineage, data quality, and workflow expertise part of how AI agents operate, providing a scalable foundation for AI-driven workflows grounded in enterprise reality. By embedding trusted context directly into agent interactions and equipping agents with reusable skills, organizations can move faster while maintaining control, transparency and trust.

Explore more about IBM watsonx.data intelligence

Read the Agentic Data Intelligence documentation for deployment guidance

Learn how Docling for IBM watsonx transforms complex documents into AI-ready data