Databricks Enters Martech with CustomerLake, an Agentic CDP Built Inside the Lakehouse

After weeks of speculation, Databricks made its move into the marketing technology market today. At its Data + AI Summit in San Francisco, the company unveiled CustomerLake, an agentic Customer Data Platform (CDP) embedded natively in the Databricks Lakehouse. It marks another expansion for a company that stepped into the security market just months ago with Lakewatch.

The pitch

The core argument is architectural. For years, enterprises have stitched together a sprawl of customer data infrastructure, yet marketers still cannot quickly answer basic questions like which customers are about to churn or which campaign actually drove revenue. Databricks frames this as a problem of architecture, not strategy. Traditional CDPs sit outside a company's core data and AI platform, creating yet another system to integrate, govern, and reconcile, plus another place where sensitive data has to be copied.

CustomerLake's answer is to put the CDP directly where the data, the AI models, and the governance already live: inside the lakehouse, governed by Unity Catalog.

CEO Ali Ghodsi tied the launch to a broader shift. His framing is that marketing stops being a sequence of discrete campaigns and becomes a continuous loop of agents that analyze, decide, and act on every customer in real time. He also notes that the customers being marketed to now increasingly include AI agents acting on consumers' behalf.

What's in it

CustomerLake bundles the usual CDP capabilities, including Customer 360, identity resolution, audience building, activation, and personalization. The headline is two categories of agents:

• Profile Agents turn raw customer data into business ready Customer 360 profiles inside Databricks. At the center is Agentic Identity Resolution, which blends deterministic, probabilistic, and agentic approaches to unify scattered records into trusted profiles.
• Campaign Agents push marketers away from static campaigns toward what Databricks calls "infinity campaigns": continuous engagement loops. A marketer starts by defining a goal, and the agents identify the audience, factor in constraints, recommend the next offer or channel, activate across destinations, and optimize on performance.

Humans still own the strategy and guardrails while agents handle execution at scale.

An open ecosystem play

Rather than building a walled garden, Databricks is leaning on interoperability, offering bidirectional pipelines into existing martech and adtech tools. Launch partners include Adobe, Meta, Braze, Acxiom, Epsilon, The Trade Desk, LiveRamp, and others. For enterprises trying to consolidate a bloated stack, Databricks is also pitching a consumption based pricing model as a cheaper alternative to traditional licensing.

Early adopters cited include HP, Circle K, and Getnet by Santander, each pointing to the value of building customer intelligence on a data foundation they already trust rather than copying data into yet another system.

The takeaway

CustomerLake is Databricks betting that the next era of marketing is agentic, and that the winning architecture keeps customer data, AI models, and agents in one governed place. Whether it displaces entrenched CDP vendors comes down to execution, but the logic matches how Databricks has expanded all year: meet customers where their data already lives.

CustomerLake is currently available in Private Preview.

This post is based on Databricks' official announcement at Data + AI Summit on June 16, 2026. Capabilities described as available in Private Preview may evolve before general availability.