From batch campaigns to AI-native CustomerLake
Databricks CustomerLake is an agentic CDP platform that unifies customer data, identity, AI models, and activation inside a lakehouse architecture to power continuous, always-on customer engagement instead of isolated batch campaigns. In its private preview, Databricks positions CustomerLake as an AI-native marketing technology layer that sits directly on the enterprise lakehouse, so the system of record and the system of decisioning are the same environment. Agents analyze behavior and determine offer, channel, and timing in near real time, while marketers control how much autonomy those agents have through human-in-the-loop approvals. According to Forrester, CustomerLake represents “a true, ground-up build of AI native marketing technology” that blends CDP, decisioning, and orchestration into one stack. This reframes the CDP category away from data collection alone and toward an engine for ongoing customer data personalization across channels.
Agentic CDPs: a new architecture for personalization
Agentic CDPs change marketing workflows by turning manual, rule-based segmentation into continuous AI-driven decision loops. Instead of marketers building static segments, exporting them, and loading them into execution tools, CustomerLake’s agents can draft campaign briefs, assemble audiences on request, resolve identities, and send actions to downstream platforms as one end-to-end process. Databricks describes specific agents, such as campaign and profile agents, that work natively with data in the lakehouse and can also call external models or tools through APIs and Model Context Protocol. This approach shrinks the distance between data, intelligence, and activation. It also means identity resolution and enrichment are no longer preprocessing chores; they become live inputs into agent decisions about what to do next for each individual, making customer data personalization a continuous, adaptive process instead of a periodic project.
Lakehouse architecture CDP: data gravity meets activation
CustomerLake is built as a lakehouse architecture CDP, reflecting the idea that marketing should run where enterprise data already lives. Databricks argues that “data gravity” has become a constraint: moving large governed datasets into separate martech platforms adds duplication, reconciliation, and latency before actions reach customers. By connecting agents to data, models, and identity directly in the lakehouse, CustomerLake aims to cut that friction and collapse some traditional CDP middleware into the core data platform. Unity Catalog governance extends to marketing workflows, providing consistent permissions, lineage, and audit trails across analytics and engagement. Integrations with partners such as Adobe, Meta, The Trade Desk, Braze, and Iterable then turn the lakehouse into a centralized decision hub that can activate insights across the broader martech and adtech ecosystem without repeated data copying or custom pipelines.
Always-on engagement and agent-led operations
Databricks promotes CustomerLake as a foundation for always-on marketing, replacing campaign flights with what it calls “infinity campaigns” that constantly analyze and act. In this model, AI agents listen for new customer signals in the lakehouse, update profiles and identity graphs, then trigger messages or offers through connected channels with minimal delay. Early pilots with brands such as Circle K and Getnet by Santander are designed to show that teams can move from insight to action without first loading data into a separate CDP. Databricks also stresses unit economics: instead of relying on large general-purpose models for every interaction, CustomerLake is intended to use smaller, task-specific models tuned for ongoing automation. The result is an agentic CDP platform where always-on customer engagement is a default behavior, and marketing operations increasingly resemble continuous software-driven optimization rather than a calendar of campaign drops.
Private preview today, inflection point tomorrow
CustomerLake is still in private preview, so its long-term impact will depend on how well enterprises accept agent-led automation in marketing. Forrester describes the launch as a “stress test” of enterprise appetite for agentic AI and a signal that attention is shifting away from standalone CDPs toward more consolidated, AI-rich stacks tied closely to core data infrastructure. If CustomerLake proves that agents can safely manage identity, personalization, and activation on top of a lakehouse, other vendors will be pushed to build similarly AI-native marketing technology rather than bolt-on features. The private preview status therefore matters less as a limitation and more as evidence that the market is entering an experimentation phase, where agentic CDPs move from concept to operating model and set expectations for tighter alignment between data engineering and marketing execution.
