Databricks for Retail and Consumer Goods

Agentic commerce is the shift from customers searching for products to AI agents handling discovery, recommendation, transaction and service autonomously on behalf of shoppers, retailers and suppliers. Industry analysts predict that by 2028, 50% of purchases will be initiated by AI agents or intent-based search. The retailers who control that experience will see dramatically higher customer lifetime value than those still relying on keyword search and static category pages.

Databricks enables this through Agent Bricks, which allows retailers to build production-grade AI agents grounded in their own enterprise data without months of custom ML engineering. Teams describe the agent's task, connect their data, and Agent Bricks handles the rest: synthetic data generation, evaluation, optimization and deployment. Lakebase provides the real-time transactional foundation these agents depend on, with millisecond-speed access to product catalogs, inventory and customer profiles, while Unity Catalog ensures every agent action is governed and auditable, giving retailers the confidence to deploy AI in customer-facing production environments.

Retail margin is under historic pressure. Research shows 30%–40% of promotional spending is wasted on consumers who would have purchased at full price regardless, and manual price reviews can't keep pace with market conditions that shift daily. Databricks powers Revenue Growth Management by enabling AI-driven optimization across pricing, promotions and assortment on a single platform: continuously adjusting prices based on real-time demand and competitive signals, measuring true promotion incrementality rather than vendor-claimed lift and optimizing SKU and shelf allocation by store to maximize revenue per square foot.

Genie puts this power directly into the hands of category managers and revenue growth managers, so a merchant can ask what the true incremental lift was from a promotion and get a data-backed, cited answer instantly. Lakebase serves the dynamic pricing APIs and promotion engines that need to read margin rules and competitive signals in real time and write decisions back to POS and e-commerce systems at transactional speed. Leading retailers achieve 2%–5% margin improvement through AI-driven optimization, and Prada reported a 20% increase in conversion rates through personalized AI applications built on Databricks.

The structural problem most retailers face isn't a shortage of data; it's that the data lives everywhere and belongs to no one. Customer data sits in CDPs, supply chain data in ERPs, promotion data in point solutions and store data in legacy POS systems. Each has its own governance model and its own copy of truth, producing slow decisions, stale insights and AI projects that never move from pilot to production. Databricks breaks down this fragmentation with a single, open lakehouse where all retail data, from clickstream to cold chain, is unified under one governance layer via Unity Catalog.

The impact is felt across every function: supply chain and merchandising operate from the same demand signals, marketing and finance from the same customer truth and AI models train on complete governed data rather than isolated subsets. Delta Sharing extends collaboration beyond the retailer's four walls, enabling real-time governed data exchange with suppliers and CPG partners without copies, proprietary lock-in or costly integrations. Walmart achieved 50% deeper supply chain insights with 300+ suppliers on Databricks using exactly this approach.

Supply chain collaboration in retail has historically meant emailed spreadsheets, proprietary EDI integrations and expensive data-sharing intermediaries, all of which create lag and governance risk. In an environment defined by tariff volatility and shifting consumer demand, that friction is no longer acceptable. Databricks enables a fundamentally different model through Delta Sharing: an open, zero-copy protocol that lets retailers share governed demand forecasts, sell-through data and inventory positions with suppliers. This happens in real time, without data copies, without requiring partners to adopt the same technology stack and without surrendering control of the underlying data.

On top of that foundation, AI-driven forecasting models combine POS data with real-time external signals like weather, events, port congestion and raw material prices to generate store-SKU-level forecasts that update continuously. AI/BI Genie gives supply chain planners conversational access to disruption signals and forecast accuracy without pulling data from disparate systems, while Lakebase powers the operational inventory applications that keep warehouse levels and order data in sync with the lakehouse in real time. Retailers on Databricks report 20%–30% reductions in forecast error and 60% reductions in out-of-stocks, while building supplier relationships that create genuine competitive advantage.