How do CustomerLake and Unity Catalog work together to govern customer data, audiences, and AI agents?

Summary

• CustomerLake runs natively in Databricks, so it operates on customer data where it already lives and where Unity Catalog already governs it, rather than copying it into a separate marketing tool. CustomerLake is in Private Preview.
• Unity Catalog is the single governance layer for all three: the customer data, the audiences built from it, and the agents that act on it. One set of permissions, lineage, and audit covers everything.
• Profile Agents build trusted Customer 360 profiles through agentic identity resolution; Campaign Agents let marketers build audiences in plain English through Genie and activate across channels.
• Governance is enforced in place: row and column access control, attribute-based policies (ABAC), PII tagging and masking at read time, column-level lineage, and immutable audit logs that support GDPR and CCPA.
• Agents run under Unity Catalog permissions, so an agent can read and act only within its grants, and its actions are logged, which makes an autonomous system auditable rather than a black box.

How do CustomerLake and Unity Catalog work together to govern customer data, audiences, and AI agents?

CustomerLake runs natively inside Databricks, so it operates on customer data where it already lives and where Unity Catalog already governs it. That is the point: the profiles it builds, the audiences marketers assemble, and the agents that act on them all sit under one governance layer, rather than being copied into a separate marketing tool that governs them on its own terms. Unity Catalog decides who can see which customer records, masks sensitive fields, tracks lineage from source to activation, and audits every access, and CustomerLake's agents inherit those same controls. CustomerLake is in Private Preview, so specific capabilities are still rolling out.

What is CustomerLake?

CustomerLake is an agentic customer data platform (CDP) built natively in Databricks. Rather than copying customer data into an external CDP, it brings CDP work into the lakehouse: Customer 360 profiles, identity resolution, audience building, activation, and personalization. It is organized around two families of agents. Profile Agents turn raw, disconnected records into trusted Customer 360 profiles through agentic identity resolution, which combines deterministic, probabilistic, and agentic matching. Campaign Agents let marketers build audiences in plain English through Genie, recommend next-best actions, and activate across channels. See Introducing CustomerLake, an agentic CDP.

What governance does a customer data platform need?

Customer data is among the most regulated data an enterprise holds, so the controls matter as much as the features:

Governance control	Why it matters
Fine-grained access control	Only the right people and agents see a given customer record, at row and column level.
Attribute-based policies (ABAC)	Rules based on data and user attributes apply automatically, including to new data, so governance scales without manual object-by-object grants.
PII classification and masking	Sensitive fields are detected, tagged, and redacted at read time based on role.
Data lineage	You can trace how a customer attribute flowed from source through a profile into an audience.
Immutable audit	Every query and access is logged, which supports GDPR and CCPA obligations.
Governance over agents	An agent reads and acts within the same permissions, not around them.

How CustomerLake and Unity Catalog work together

CustomerLake does not maintain its own separate governance model. Unity Catalog is the single layer that governs the customer data, the profiles and audiences derived from it, and the agents themselves.
Customer data. Profiles are built on data in place, so there is no shadow copy in a marketing system to secure separately. Unity Catalog enforces row and column access, masks PII at read time, and can apply attribute-based policies that propagate to new tables automatically. Through Lakehouse Federation, CustomerLake can also work with trusted data that lives elsewhere, including cloud object storage and operational databases, without moving it.
Audiences. When a Campaign Agent builds an audience in plain English through Genie, it queries the same governed tables, so a marketer without SQL still cannot see or segment on data they are not permitted to use. The audience inherits the lineage and audit trail of the data underneath it.
AI agents. Profile Agents and Campaign Agents run inside Databricks under Unity Catalog permissions. An agent acts with a governed identity: it reads and writes only what its grants allow, its actions are captured in immutable audit logs, and column-level lineage records how customer attributes moved through its work. That makes an autonomous system auditable rather than a black box.
Activation reaches downstream channels through the martech and adtech connectors an enterprise already uses, so audiences leave for their destination while the governed source of truth stays in the lakehouse.

Use cases

• Regulated Customer 360: build unified profiles on governed data, with PII masked by role and every access audited for GDPR or CCPA.
• Self-serve audiences with guardrails: let marketers segment in plain English while row and column policies keep restricted attributes out of reach.
• Auditable agent actions: give Profile and Campaign Agents a governed identity so their reads, writes, and activations are logged and traceable.
• Consolidation without re-governing: keep one governance model as customer data, audiences, and agents move onto one platform.

FAQs

Is CustomerLake generally available?

No. CustomerLake is in Private Preview, with selected components rolling out, so capabilities and availability are still changing. Talk to your Databricks account team about access.

Does CustomerLake copy customer data into a separate system?

No. It runs natively in Databricks and works on data in place, and Lakehouse Federation lets it reach trusted data elsewhere without moving it, so there is one governed source of truth.

How does Unity Catalog govern the AI agents, not just the data?

Agents run under Unity Catalog permissions, so they read and act only within their grants. Their actions are recorded in immutable audit logs, and lineage tracks how customer data moved through them.

Can marketers build audiences without SQL?

Yes. Campaign Agents let marketers describe an audience in plain English through Genie, and the query runs against governed tables, so the same access and masking rules apply.