What capabilities should an agentic CDP include?

Summary

• An agentic CDP should include the core capabilities of any CDP. The vendor-neutral CDP Institute lists ingesting any source, capturing full detail, building persistent unified profiles, providing open access, responding in real time and governing the data.
• On top of that baseline, Databricks names three defining characteristics: agentic personalization through always-on Infinity Campaigns, embedding in the data foundation where governed customer context lives, and a design in which every function can be run by an agent or a human.
• Agents need context as well as records. Databricks calls this Golden Context: the unified profile plus what the business is trying to accomplish right now and what has already been tried with each customer.
• Governance has to cover agents. Databricks says an Agentic CDP integrated with the lakehouse governance layer runs every agent action under the same data entitlements and security boundaries as every other operation in the enterprise.
• Humans stay in control. Databricks says they still define the strategy, goals and guardrails while agents help scale execution.

What capabilities should an agentic CDP include?

An agentic CDP should keep the capabilities of a conventional CDP and add the ones that let agents operate on customer data. The CDP Institute baseline is a system that ingests any source, captures full detail, builds persistent unified profiles, provides open access, responds in real time and governs the data. Databricks argues that an agentic CDP adds agentic personalization, a position inside the data foundation, and an agent-native design in which humans set goals and stay in control.

What is an agentic CDP?

Databricks' founders define an Agentic CDP as one that powers always-on 1:1 personalization with Infinity Campaigns, is embedded in the data foundation where governed customer context lives, and is architected from the ground up for agents as first-class operators alongside humans. They contrast it with CDPs built for human-operated, batch-based campaigns. An Infinity Campaign, in their words, is an autonomous, continuously compounding engagement loop that adapts message, timing and channel as new context signals arrive.

Which capabilities should you evaluate?

Capability	What to look for	Basis
Ingestion	Batch and real-time customer, campaign and engagement data from marketing tools, advertising platforms and databases	CDP Institute baseline; Databricks
Identity resolution	Deterministic and probabilistic matching combined, a review path for hard cases, and support for the rules you already have	CDP Institute; Databricks
Unified profile and context	A persistent profile that also carries current business goals and the history of what was tried and how the customer responded	CDP Institute; Databricks founders
Audience building	Segments that marketers build and refine in natural language on governed data	Databricks
Decisioning	Next-best-action recommendations, real-time serving of customer attributes, and custom models	Databricks
Agent-run campaigns	Agents that build audiences, recommend actions, activate across channels and optimize toward goals within limits humans define	Databricks
Activation	Managed connectors that sync audiences and signals to martech, adtech and customer experience tools without custom pipelines	Databricks
Governance	Access controls, lineage and auditability across customer data, audiences and AI assets, with agent actions run under the same entitlements as other operations	Databricks
Openness	Works with the tools you already use and reaches data where it resides	Databricks

How does Databricks approach these capabilities?

CustomerLake is the Databricks Agentic CDP, embedded natively in Databricks and in Private Preview. According to Databricks, Profile Agents turn raw customer data into business-ready Customer 360 profiles, and Agentic Identity Resolution combines deterministic, probabilistic and agentic matching with LLM and human review for edge cases. Campaign Agents use governed customer context to build audiences, recommend next-best actions, activate across channels and optimize toward business goals, while humans define the strategy, goals and guardrails. Marketers use Genie to build audiences in natural language. Custom models deploy through Model Serving, native managed reverse ETL connectors handle activation, Unity Catalog governs the data and Lakehouse Federation reaches data outside Databricks.

FAQs

How is an agentic CDP different from a conventional CDP?

Databricks says conventional CDPs were built for human-operated, batch-based campaigns, while an agentic CDP is designed so that every function can be run by an agent or a human.

What does embedded in the data foundation mean?

Databricks says the lakehouse and the CDP should occupy the same layer, so that context is collected, managed and analyzed in one place rather than synchronized across systems. For CustomerLake, Databricks says CDP capabilities run where customer data already lives, so the data isn't moved or copied.

What is Golden Context?

Databricks defines it as the unified customer profile plus two things a Golden Record doesn't carry: what the business is trying to accomplish right now, and what has already been tried with this customer and how they responded.

Do humans stay in control?

Databricks says humans still define the strategy, goals and guardrails, and agents help scale execution. CustomerLake's identity resolution also includes human review workflows for edge cases.

How should I test identity resolution during an evaluation?

An article published by the CDP Institute advises using representative data and measuring both false merges and missed matches, rather than relying on a single vendor-reported match rate.

Is CustomerLake generally available?

No. Databricks announced CustomerLake on June 16, 2026, and it is in Private Preview.