Enterprise AI is rapidly evolving toward domain-specific, continuously learning agents that demand robust governance, scalable infrastructure, and accessible tools to empower both technical and business users. These advancements highlight a broader market shift, from one-off prompts to data-driven, adaptive agents that rely on high-quality, well-managed data to operate effectively and efficiently.
IBM has anticipated this shift. Over the past year, we’ve centered our data platform around watsonx.data and introduced two new offerings: watsonx.data integration and watsonx.data intelligence. Both tools are built to address the foundational data challenges that must be solved before AI agents can deliver real value. These tools help clients successfully deploy agents and AI across IBM, Databricks, or any other platform they’re investing in. IBM’s approach stands out in simplifying agentic AI lifecycle management with watsonx Orchestrate, connecting and coordinating agents across systems and acting as a central nervous system for enterprise AI, including custom agents and tools built with Databricks alongside others.
watsonx.data integration focuses on data delivery. It gives data teams a single control plane for multiple styles of integration, such as batch pipelines, change data capture, streaming and file ingestion, regardless of whether the destination is Db2, Delta Lake on Databricks, object storage or an event bus. With built-in flexibility and adaptability, it helps eliminate fragmented tooling and minimizes technical debt as storage paradism evolve, laying a scalable foundation for future-proof data architecture. Unstructured data integration capabilities also help organizations and data engineers capitalize on the unrealized value of files, documents, and other type of media which are fundamental for generative AI and agents, and these capabilities are available independently in watsonx.data integration as well as the broader watsonx.data platform.
watsonx.data intelligence tackles data context. It equips enterprises to infuse trusted for agents and AI use cases by automating governance, lineage, and quality shared across hybrid environments for both structured and unstructured data. Built with generative AI at its core, the platform accelerates metadata creation, enriches data catalogs, and surface relevant assets through intelligent semantic search, thereby reducing manual effort and enabling faster time to insight. Its conversational Data Intelligence Assistant acts as an agent that guides users through data discovery, helping both technical and non-technical teams make sense of complex data landscapes. By delivering explainable, high-quality data, watsonx.data intelligence lays the foundation for responsible and scalable adoption of AI use cases for enterprises.
Together, products in the watsonx.data platform bring agents to life. At the Databricks summit, IBM demonstrated and proved how the platform can unlock the value of unstructured data to create an agent within minutes.
-
Using watsonx.data integration, a user can ingest files such as invoices directly from storage solutions from Box from a low-code drag and drop interface
-
Native capabilities are built in to ensure security through propagation of Access Control Lists (ACLs), and removal of PII
-
Unstructured documents can be parsed to extract, annotate, classify, chunk, embed, and store our data into a vector store such as Milvus on watsonx.data
-
As data is processed, watsonx.data intelligence extracts and creates semantic understanding on all the derivative elements like key value pairs, and embeddings, and maps them to business terms and other data classes
-
The AI-ready data can directly be used within watsonx orchestrate, allowing business users to query the data using natural terms such as invoice IDs and purchase orders to extract meaningful insights
For customers who are invested in Databricks, try out the following features today:
Why do these innovations matter? For data engineers, tool sprawl is already a headache. Surveys show teams multiple services just to move data from systems of record into a lakehouse and onward to AI workloads. watsonx.data integration unifies that work, and therefore engineers spend less time connecting disparate siloed data sources and more time optimizing transformations that boost model accuracy. For business leaders, agents promise productivity leaps, yet industry research warns that ungoverned AI erodes trust and slows adoption. Gartner’s midsize-enterprise CIO survey found that 76 percent of respondents plan to use AI to launch new products or services, but only 40 percent of consumers trust the technology’s output without additional controls. watsonx.data intelligence makes those controls native, providing lineage and quality metrics that auditors can easily verify.
IDC’s Agentic Evolution brief reports that over a half of enterprise applications now include AI assistants, and 1 in 5 embed full agentic capabilites with expected growth. This rapid adoption underscores that leaders must combine open ecosystems with robust governance to ensure responsible AI adoption. This market truth aligns with IBM’s strategy: we meet customers where their data lives, keep formats open, and layer AI assistance on the integration and governance steps that have historically slowed innovation. Databricks’ announcements validate the direction, showing strong demand for turnkey agent frameworks and large-scale vector search. IBM ensures the underlying data and policies are ready the moment those frameworks arrive.
At the end of the day, flashy agent builders and advanced vector search tools tend to draw the most attention, but it’s the underlying data foundation that determines whether these systems actually work in practice. IBM’s approach focuses on making it easier to integrate, prepare, and manage data so that teams can build agentic AI systems that are reliable, transparent, and scalable. Whether your workloads live in watsonx, Databricks, or across both, having a consistent and well-governed data layer is what makes the difference between AI prototyping and releasing something that’s ready for production. As organizations move beyond experimentation, it’s that data foundation that will truly differentiate agents that deliver real value.