Databricks has introduced CustomerLake, an "agentic" customer data platform that runs natively inside the Databricks lakehouse, the company announced.

The launch marks Databricks' formal entry into the CDP category, a market long dominated by standalone SaaS vendors. By building CustomerLake into its existing data platform, the company positions the new tool against incumbents such as Segment, mParticle, Treasure Data, Adobe Real-Time CDP and Salesforce Data Cloud.

What does "agentic CDP" mean in practice?

The "agentic" label, increasingly common in 2024–2025 martech launches, describes software that uses autonomous AI agents to perform tasks a data or marketing engineer would normally script. In a CDP context, that typically means agents that build segments, reconcile identity records, or trigger activation paths in response to plain-language prompts from a marketer.

Databricks' use of the term — rather than "AI-powered" or "automated" — places CustomerLake in the same product category as agentic offerings from vendors such as Salesforce (Agentforce) and HubSpot (Breeze Agents).

Why does embedding inside Databricks matter?

Most CDPs operate as a separate vendor stack. They ingest customer data from a warehouse, build profiles in their own database, then sync audiences back out to campaign tools. That architecture carries a known cost: nightly syncs, duplicate identity graphs, and a per-record or per-event fee that scales with activation volume.

A CDP built inside the same lakehouse where the customer already stores transactions, events and model features removes those steps. For practitioners, the model resembles the warehouse-native CDPs Hightouch and Census, which activate from a warehouse rather than a separate profile store.

Direct implications for a Databricks customer:

• No external data egress charges for moving records into the CDP
• Identity, consent and PII controls inherit from existing Unity Catalog policies
• Segments, model scores and journey logic share storage with downstream analytics
• Activation latency drops because no overnight batch syncs audiences

What does the launch change for marketers?

The most direct impact lands on marketing operations leaders who already pay Databricks for analytics and ML. A native CDP gives those teams a path to retire a separate Segment, mParticle or Adobe contract and consolidate customer data tooling on a single platform.

The agentic framing also raises the operational bar. A marketer who can prompt the system in natural language to "build a high-intent audience for the Q3 launch" reduces dependency on the data engineering team that previously owned audience logic.

What is still unknown

Databricks has not disclosed pricing model, edition availability, or general-availability date in the announcement. Practitioners evaluating the product will need answers on three fronts before treating CustomerLake as a replacement for an existing CDP:

1. Identity coverage — which paid-media IDs, device IDs and B2B identifiers CustomerLake resolves natively
2. Activation reach — which ad platforms, MAPs and CRM systems the tool pushes audiences to without a separate connector
3. Governance — how consent, regional residency and PII masking are enforced under Unity Catalog

Martech Beacon will update this story as Databricks publishes pricing, GA timing and a full feature list.