IBM says its data agent makes a pipeline in three minutes

IBM released Agentic Data Integration, a preview in watsonx.data integration, where a user describes the data in plain words. An agent then plans, makes, checks and runs the pipeline, and the user approves the plan before it runs. IBM says the full cycle of the agent takes less than three minutes, but it gave no test results.

How it works

The agent adds a plain-word layer to the integration engine of IBM. It finds the data sources and writes a plan. The user reads and approves the plan before the pipeline runs. Then the agent applies the governance rules and delivers a dataset.

The numbers

IBM says a data task takes one to four weeks at this time. Omdia research says that only 4% of organizations can add a new trusted data source for AI. They do it in less than one week. IBM estimates that almost half of the time of engineers goes to the maintenance of pipelines.

What is not known

The tool is a preview, and IBM gave no cost and no date for release. It gave no test of the three minute claim. The claims of lower cost are an estimate of IBM.

Insight: Data vendors open their systems to external agents

IBM lets agents from other makers tell the system to make pipelines. The connection is an MCP server, a standard connection between agents and tools. Workday did the same for HR and finance data in October 2026. This reads as a move to be the gate that agents use to get the data of a company. The first gate that a company chooses can be hard to change.