Databricks has launched CustomerLake, an agentic customer data platform built inside its lakehouse. It uses a workforce of agents to analyse behaviour, decide and act in real time, replacing one-off campaigns with continuous loops. It puts the CDP where the data already lives and signals where martech is heading.
The data warehouse becoming the CDP is the quiet structural shift here. The campaign stops being a thing you launch and becomes a thing that runs itself.
Databricks has walked into the marketing business. On 16 June 2026 it launched CustomerLake, an agentic customer data platform built natively inside its lakehouse, with Integral Ad Science signing on as a launch partner.
The pitch is that the CDP now lives where the data already sits. CustomerLake handles identity resolution, audience segmentation, real-time profile serving and natural language access to customer data, and it runs a workforce of agents that continuously analyse behaviour, decide and act. Databricks calls the output infinity campaigns, continuous loops that react to customer context in real time rather than the old fire-and-forget send. It is in private preview with HP, Circle K and AB InBev among the early users.
Why it matters
For a decade the customer data platform was a separate box you bought, fed and tried to sync with everything else. Databricks is arguing that box should not exist as a box. If the data lives in the lakehouse, the activation should too.
CustomerLake is built to deliver personalised customer experiences a billion times a day through continuous agentic loops, not one-off campaigns
This is enterprise tooling and most Australian businesses will not touch it directly. The direction still matters to every one of them. The platforms are betting that agents will run the always-on personalisation that humans used to brief campaign by campaign. The businesses that win that shift are the ones whose underlying data is clean enough to trust an agent with.
What to do about it
Fix the data foundation before you chase the agent. An agentic CDP acting on messy data does not fix the mess. It automates it and scales it. Clean identity and clean definitions come first.
Know what you actually hold. Before any of this is useful you need to know which first-party data you have, where it lives and whether you are allowed to use it the way you want to.
Do not buy the enterprise dream on an SME budget. The principle, always-on personalisation off clean data, scales down. The platform does not. Take the idea, not the invoice.
Watch the privacy line. Agents acting on customer data a billion times a day is exactly the sort of processing that regulators and customers care about. Get consent and governance sorted before you automate.
The lesson for the rest of the market is simple. The value is moving to whoever owns clean first-party data and can act on it fast. That is a job you can start on today without a lakehouse, by knowing your numbers and tidying your data.