Databricks launches agentic AI for retail planning

Databricks has unveiled an agentic AI framework designed to overhaul the traditional Monday morning joint business planning process for consumer packaged goods and retail partnerships. By transitioning teams from static reports to intelligent decision systems, the platform aims to reclaim the forty analyst hours lost each week to manual data stitching and reduce the three-to-five-day lag between signal and action. The system continuously streams point-of-sale, shipment, and inventory data, allowing retail partners to analyze millions of item and store combinations across thousands of SKUs and stores overnight.

This automated workflow relies on several core Databricks technologies to ensure context, control, and model flexibility. Genie Ontology, a context layer introduced at the company's 2026 summit, builds a self-improving knowledge graph from existing tables, queries, and dashboards. It utilizes an authority score called OntoRank to resolve conflicting data definitions. Security and governance are managed via the Unity AI Gateway, a control plane built on Unity Catalog that enforces rate limits, spend caps, and user permissions. Data sharing between manufacturers and retailers is handled securely without copying through Delta Sharing.

For practitioners, the system shifts the workflow from manual analysis to automated recommendation. Databricks has evolved its natural language assistant into Genie One, a virtual coworker that compiles daily briefs from calendars, inboxes, and governed data. Additionally, Genie Agents allow teams to convert recurring prompts into shareable, policy-compliant agents. Instead of spending joint planning meetings arguing over conflicting spreadsheets, sales and supply chain practitioners receive pre-analyzed watch-outs and drafted purchase orders for human approval. Databricks claims organizations can implement this system, starting with a ninety-minute discovery workshop, to run their first live AI-driven Monday report on shared data by day ninety.