After a wave of consolidation over the past six years, more than a few martech minds have wondered whether CDP is a better concept than a product.
Vendors in and around the digital experience space, in particular, were buying CDP, and in some cases this was because the two vendors and their products were often used in tandem by customers anyway. Why not join forces then?
If you were in the world of martech, you would remember those days. Twilio purchased Segment; SAP bought Emarsys; Contentstack purchased Lytics; Uniphore purchased ActionIQ, to name just a few.
These were not legacy platforms. They were CDPs created for a data problem. They were developed to collect and unify customer data and profiles, create audiences and activate campaigns. Their strengths lay in identity resolution and unifying the customer profile.
Last week, Hightouch published a blog post about his vision of an agent-based CDP. A day later, Databricks announced CustomerLake, its CDP agent.
The concepts discussed by Hightouch and Databricks have a lot in common. But at a high level, what they say is this: the future is not customer profiles, but customer decisions.
Your customers search everywhere. Make sure your branding introduces himself.
The SEO toolkit you know, plus the AI visibility data you need.
Start your free trial
Start with
The agentic vision of CDP 3.0
If the unifying customer profiles were CDP 1.0, the composability was CDP 2.0, and the agentic CDP was CDP 3.0, and consists of unified customer data + AI decision + autonomous execution.
While CDP 1.0 saw data as the problem to be addressed, it can be argued that CDP 3.0 sees humans as the obstacle. We are too slow to analyze data and make decisions. AI agents, on the other hand, move quickly and constantly.
“Agentic AI offers the path to not only implement new capabilities that extend the mandate of the CDP, but also to develop a new paradigm for generating insights, targeting audiences, making decisions and orchestrating customer journeys.”
Who will own the agentic layer and where will they live?
Where Hightouch and Databricks diverge is in their philosophies on how all of this plays out.
“Five years ago, we thought we were building a better CDP architecture. In reality, we were laying the foundation for intelligent agents. By moving audiences, journeys, and activations into the warehouse, composable CDP connected marketing directly to the richer business and customer context.”
Hightouch’s vision is for agents to do their work in the data warehouse, without copying data. This is true to the company’s roots in composability and also adds an agent layer to the conversation. For Hightouch, the agent layer in CDP 3.0 is located in a marketing platform above the data platform.
Databricks’ CustomerLake is an example of the company’s philosophy that data warehouses, in this case Databricks’ Data Lakehouse technology, can also serve as an application platform. Databricks already did this with enterprise security when it launched Lakewatch in March 2026. With CustomerLake, it’s bringing this philosophy to marketing. (It’s worth noting that CustomerLake can import data from sources outside the Data Lakehouse.)
Databricks sees an advantage in building your CDP on top of the data platform because the governance, AI, and business context are already there. Don’t copy it, don’t move it, just do the work there.
Is the CDP large enough for both models?
At first glance from here in the cheap seats, it appears that Hightouch and Databricks are on a collision course. Let’s put aside for a moment the other players that will enter the CDP 3.0 space as it matures. (BlueConic’s acquisition of Blueshift last week has a similar story about adding AI agents and actions to customer data.)
In reality, Hightouch and Databricks can coexist because they tend to target different organizations.
For Databricks, which sells a data platform, the primary buyer is often the data, AI and platform teams on the technology side. Hightouch sells most often to CRM, marketing and lifecycle teams.
This means that each supplier often has a different starting point in their potential organizations. For Hightouch there is an existing data warehouse on which to position its product and an existing martech stack in place. For Databricks, there is an existing lake house and an enterprise-wide AI strategy in place.
Since Databricks works at an enterprise level, it is looking for customers with data engineering and AI maturity. Hightouch is looking for maturity in marketing operations.
Databricks’ corporate focus means the time to value is potentially longer than it is for Hightouch customers.
But these significant differences mean that the two companies will often be fishing in different ponds. Hightouch targets large DTC operators, consumer financial services, retailers, subscription businesses, and travel and hospitality companies where the CMO will likely play a role in the decision.
Databricks, on the other hand, will find itself in conversations with global financial services companies, telecommunications, large healthcare systems, and large enterprises with a centralized data organization in which the CIO will likely be heavily involved.
In other words, either approach is unlikely to win. Each approach can be successful for the right client.
The best we can hope for is that these platforms deliver on their promises and that both marketers and their customers win.
