Particle.news
Download on the App Store

Databricks Recasts Lakehouse as Real-Time Foundation for Agentic Applications

The company unveiled LTAP, expanded Lakebase, and launched CustomerLake to let AI agents act on a single, governed copy of business data with built-in controls and Microsoft integrations.

Overview

  • Databricks announced at its Data + AI Summit on Tuesday that it is shipping LTAP (Lake Transactional/Analytical Processing) to unify transactional and analytical workloads on the lakehouse so agents can read and write live data without persistent ETL or replicas.
  • CustomerLake, an agentic customer data platform built inside the lakehouse, is available on Azure Databricks and uses Profile Agents and Campaign Agents to assemble Customer 360 profiles and automate personalized campaigns.
  • The company expanded Lakebase with enterprise features such as Git-style branching for quick full-fidelity database forks, real-time mirroring from transactional Postgres into the lake, and Lakehouse//RT for sub-second analytics on lake data.
  • Unity Catalog AI Gateway and new contextual service policies extend governance to models, agents, MCP services and external tools with unified spend controls, auditing, and fine-grained runtime rules for agent actions.
  • Databricks broadened Agent Bricks and Genie with managed agent memory, document intelligence, multi-model support and Microsoft integrations like OneLake catalog federation (GA) and an Excel add-in preview to bring governed lake data into business workflows.