Databricks announced CustomerLake on June 16, 2026, pitching it as an agentic customer data platform built natively inside Databricks. The product brings Customer 360, identity resolution, audience building, campaign automation, activation, and personalization into the same lakehouse environment where many enterprise data teams already run models and governance.
CustomerLake is available in Private Preview. Databricks named HP, Circle K, AB InBev, and Getnet by Santander as current customers, and says Campaign Agents and Profile Agents will support “infinity campaigns”: continuous loops that analyze behavior, decide on a next action, and activate across channels.
For B2B marketing ops, the question is less “is this a new CDP” and more “who gets control of the customer graph when AI agents start acting on it.” The SERP is already full of launch recaps. The useful test is whether CustomerLake can reduce data movement without turning every campaign decision into a data-team ticket.
Key Takeaways
- Databricks CustomerLake entered Private Preview on June 16, 2026 as an agentic CDP built inside the Databricks platform.
- The launch combines Customer 360, identity resolution, audience building, campaign automation, activation, and personalization in one governed data layer.
- Databricks says Profile Agents and Campaign Agents support continuous “infinity campaigns” rather than one-time journey builds.
- The B2B buyer test is control: identity proof, consent rules, activation handoffs, and marketer autonomy all need evidence.
- The strongest ranking angle is not “Databricks launched a CDP.” It is whether lakehouse-native CDP changes who owns campaign execution.
What CustomerLake Actually Adds
The Databricks launch blog describes two core agent groups. Profile Agents turn raw customer data into business-ready Customer 360 profiles, including identity resolution and enrichment. Campaign Agents build audiences, recommend next-best actions, activate across channels, and optimize against business goals.
The architectural claim is the important part. CustomerLake is governed by Unity Catalog and uses Lakehouse Federation to access data where it already lives, including Databricks, Snowflake, Google BigQuery, cloud storage, and operational databases. That puts it in direct contrast with CDP projects that copy customer data into a separate application before marketing can use it.
That is why CMSWire framed the launch as a move into the CDP market, not only a new agent product. Databricks is trying to make the enterprise data layer the place where customer intelligence and campaign action meet.
The Control Test Hidden in the CDP Pitch
Most CDP debates are sold as data debates: identity, profiles, segments, and destinations. CustomerLake moves the debate toward control. If agents can decide, act, and optimize inside the data platform, marketing leaders need to know which decisions remain marketer-owned and which shift to data teams, models, or platform policy.
The tension is familiar. Adobe AJO B2B Prime removed the Real-Time CDP requirement for mid-market Marketo customers, letting teams adopt AI orchestration before a full CDP migration. Databricks makes the opposite argument: do not skip the data foundation. Put the CDP where governed enterprise data already sits.
Twilio made a related move at the conversation layer. Its Conversation Memory and Orchestrator launch argued that agents need persistent customer context across channels. CustomerLake is the CDP-side version of that same thesis: identity, memory, decisioning, and activation all become more useful when the agent can read the same context the business already trusts.
What Marketing Ops Should Prove in Private Preview
Private Preview is where B2B teams should resist the demo narrative and ask for measurable operating proof.
- Identity proof: Ask how Agentic Identity Resolution handles conflicting records, stale fields, household or account-level merges, and human override. If the matching logic is not explainable, the activation layer inherits risk.
- Consent and suppression proof: Test whether the agent respects regional consent, channel opt-outs, account exclusions, and sales-owned suppression lists before it recommends any audience.
- Activation proof: Map which destinations receive the decision and how quickly performance data returns. A lakehouse-native CDP still fails if activation feedback takes too long to improve the next action.
- Marketer autonomy proof: Measure how many campaign changes a marketer can make without a data engineer. If every prompt creates a data-ticket queue, the CDP moved but the bottleneck did not.
The closest internal comparison is Salesforce Agentforce Marketing, where the real issue is permission scope rather than email drafting. CustomerLake turns the same question toward the customer data layer: which agents can touch identity, audiences, and activation, and who audits the result?
Databricks has a credible wedge because many enterprises already trust it as the governed data platform. But a trusted data platform is not automatically a trusted marketing operating layer. CustomerLake will win B2B marketing attention only if Private Preview customers can show faster activation, fewer duplicated records, cleaner governance, and fewer data-team handoffs in the same quarter.
Frequently Asked Questions
Databricks CustomerLake is an agentic customer data platform built natively inside Databricks. It combines Customer 360, identity resolution, audience building, campaign automation, activation, personalization, governance, and agent workflows in the lakehouse environment.
Databricks announced CustomerLake on June 16, 2026 at Data + AI Summit. The product is available in Private Preview, with HP, Circle K, AB InBev, and Getnet by Santander named as current customers.
Traditional CDPs often copy data into a separate marketing system. CustomerLake keeps the CDP inside the governed Databricks data foundation, then adds Profile Agents and Campaign Agents that can prepare profiles, build audiences, recommend actions, and activate campaigns from that shared context.
Start with identity accuracy, consent and suppression logic, activation latency, marketer autonomy, and audit records. The first proof point should not be whether an agent can build a campaign. It should be whether the agent can act on governed customer data without creating risk or new handoffs.
