How Agentic Data Prep & Analysis Works

Phase 1 – Preparation: The user describes a business goal in natural language (example: "segment marketing leads by campaign"). The AI agent reads the knowledge graph, thinks through the logic, and writes a data pipeline. Intermediate steps visible to the user include joins, filters (e.g. "status = new"), and segment definitions against account detail data.

Phase 2 – Review: The agent surfaces its work as a visual pipeline that the user can inspect step by step — showing exactly how the data was prepared, what was joined, and what was filtered. Users can refine any step before proceeding.

Phase 3 – Analysis: Once the prepared data is validated, the agent generates insights and visualizations. Example output: "There are 3 segmentations of the marketing leads with interesting insights." Results are rendered as charts and summaries the user can act on immediately.