Editorial
Research Report
Building the Case for Modern Public Sector Messaging
The Gist
- What is Databricks CustomerLake? An agentic CDP built natively inside the Databricks lakehouse that eliminates the separate customer data platform by bringing identity resolution, audience building and activation into the same governed environment where enterprise AI models already run.
- Why does the pricing model matter? CustomerLake carries no platform fee — Databricks captures value downstream through compute billing. Independent CDP vendors cannot compete on price with a company that earns revenue whether or not you buy its product.
- What should enterprise teams do now? Inventory what your current CDP does against what the platform now handles for free, run a beta as negotiation leverage, and address identity governance and agent permission boundaries before autonomy defaults are set for you.
The marketing technology industry names a shift after the part of it that has the best demos and makes the best headlines. We called the last decade "composable" after an architecture, and we are now calling this one "agentic" after a mechanism. Both labels describe the mechanism and miss what actually changed.
With Databricks’ announcement of their CustomerLake CDP, the customer data platform stopped being a separate system and became a function of the platform where the data already sits.
CustomerLake, the agentic CDP the company announced at its Data + AI Summit and built natively inside the lakehouse, asks: Does a customer data platform need to exist anywhere other than where your data and your agents already are? Many enterprises with upcoming contract renewals would do well to figure out the answer.
Table of Contents
- What Is Databricks CustomerLake?
- How Composable CDPs and the Data Warehouse Converged to Make CustomerLake Inevitable
- What CustomerLake Actually Does: Orchestration, Not Just AI
- Why CustomerLake's No-Platform-Fee Pricing Is an Existential Threat to Independent CDPs
- What Snowflake, Google, Microsoft and Adobe Are Likely to Do Next
- Where Independent CDPs and Neutral Vendors Can Still Compete
- What Enterprise Teams Should Do Before Their Next CDP Renewal
- Frequently Asked Questions About Databricks CustomerLake and the Future of CDPs
- Why Long-Term Architecture Decisions Matter More Than the AI Label
What Is Databricks CustomerLake?
Databricks CustomerLake is a natively integrated customer data platform built inside the Databricks lakehouse. It combines Customer 360, identity resolution, audience building and AI-driven campaign activation in a single governed environment — eliminating the need to copy data into a separate CDP. For enterprises already running data and AI workloads on Databricks, it removes the reconciliation layer that has historically sat between the warehouse and the CDP.
How Composable CDPs and the Data Warehouse Converged to Make CustomerLake Inevitable
The architecture has moved in one direction for most of a decade. Composable CDPs from Hightouch, Census, GrowthLoop and Simon Data taught the market to stop copying customer data into a separate system and to activate it directly from the warehouse via reverse ETL. Snowflake and BigQuery concentrated that data in one place. The pitch worked well enough that the data platform was positioned to build the same capability itself.
CustomerLake realizes that capability by bringing Customer 360, identity resolution, audience building and activation into the same governed environment where many enterprises already run their AI models. Plus, this no longer requires copying data across vendors. Having personally watched enterprise teams spend several quarters reconciling identities between a CDP and the warehouse that feeds it, I see that this design eliminates the reconciliation by eliminating the second system.
Related Article: Meet the Newest Martech Member: Databricks, Via Agentic CDP
What CustomerLake Actually Does: Orchestration, Not Just AI
The terminology I prefer to use when talking about things under the broader umbrella of “AI” helps to focus the conversation. Orchestration (not “agentic”) is the capability that matters here, with AI coordinating and executing across systems, tools and channels, on a dial ranging from human-directed workflows at one end to unsupervised execution at the other. Agentic is one mechanism for delivering orchestration, sitting at the far end of that dial.
Here’s one example. The infinite loops that stand in for one-off sends are really orchestration moved into your data layer in CustomerLake’s “infinity campaigns.” Agents that scrub profiles turn raw records into golden profiles using deterministic, probabilistic and AI matching. Campaign Agents then build audiences, recommend next-best action and optimize against a goal you specify. “Agentic” refers to how this work gets done. Think of it as moving orchestration to where your data lives instead of to a separate platform.
This implementation of orchestration follows the trend I mentioned earlier by using the lightest mechanism that clears the task's accuracy bar, and reaching for the agentic end only when the task's openness demands it. One big reason to do this is that you pay for that reach twice, once for compute and once for verification. Databricks built that discipline into its economics, saying it will run smaller models tuned to specific marketing tasks rather than spend compute on frontier models for every customer interaction.
Why CustomerLake's No-Platform-Fee Pricing Is an Existential Threat to Independent CDPs
The pricing applies the same rule and stands to reorder the market. CustomerLake carries no platform fee because Databricks captures the value downstream in compute; it bills every time it resolves an identity or builds a profile. An independent CDP charges for the CDP because that is its only revenue line. You cannot out-price a company that earns money whether or not you buy its product.
David Raab of the CDP Institute called the move smart and "a challenge for many conventional CDP vendors." The man running engineering for CustomerLake, Tasso Argyros, built ActionIQ before this and now says the CDP "as middleware, is going to go away." A founder of the category is calling for the category to dissolve.
Related Article: David Raab's Take on CMSWire: The Fight for Martech's Most Valuable Layer Has Begun
What Snowflake, Google, Microsoft and Adobe Are Likely to Do Next
Gartner views this as an architectural change, not a phase that will pass. They’re advising CMOs to “consider CustomerLake an infrastructure decision” and expect that by 2030, 80% of newly deployed CDPs will be native integrations built into the data platform, rather than separate entities. Snowflake is clearly next in line. Both Google and Microsoft are integrating audience tools into their core clouds directly, and Adobe isn’t fighting this integration because its platform is built on Databricks. If Adidas and AT&T most are already committed then the music will play before the product even ships. Now it’s game-on for who will own the layer where the magic happens.
Where Independent CDPs and Neutral Vendors Can Still Compete
A data platform gets governance and scale for free. It doesn’t get channel trust for free, nor can it convincingly play both sides. CustomerLake federates to Snowflake and BigQuery via Lakehouse Federation, but this significantly weakens the naive “cross-cloud” argument, and the query still lands within Databricks’ control plane/governance/billing. Nor will the lakehouse ever have any incentive to build a truly neutral intelligence layer that can operate across clouds without playing favorites. Hightouch saw that argument coming and took to Twitter to publish its own agentic-CDP thesis the day before the keynote: cede the data layer, claim ownership of everything else. “Hightouch will remain the place where marketing gets done.”
Regulation and data sovereignty are second-party constraints that provide a second avenue of attack. A hyperscaler cannot serve a regulated (or sovereignty-constrained) vertical as cleanly as a specialist can. I expect half of this trend to play out in both directions over the next eighteen months: vertical, agentic CDPs that know one industry’s compliance landscape backward, and a neutral layer that abstracts agentic workloads across competitive, warehouse-native platforms. Whoever cleanly solves for multi-platform governance will have a sustainable business.
Trust and permissions are the other grounds on which a data platform does not get for free, and they are where autonomy carries the most risk. Under orchestration, trust means a verifiable audit trail of the actions an agent took, checkable after the fact rather than taken on the agent's own report. Permissions mean a clear answer to what the agent may do and what it may spend, a question that becomes increasingly significant the moment execution moves toward the unsupervised end of the dial. Vendors that solve those two problems will retain customers; a data platform cannot easily take them.
What the Databricks CustomerLake Launch Means for Enterprise Data and Marketing Teams
The following table highlights the most important lessons, actions and strategic considerations emerging from Databricks' launch of CustomerLake and the broader shift toward lakehouse-native customer data platforms.
| Key Area | What Happened | Why It Matters | Recommended Action |
|---|---|---|---|
| CDP Architecture | aDatabricks launched CustomerLake, a native CDP built directly into its lakehouse platform with identity resolution, audience management and activation capabilities. | Organizations can increasingly manage customer data, segmentation and activation within the same environment where data already resides, reducing integration complexity. | Assess whether your existing CDP provides capabilities that remain differentiated from emerging lakehouse-native alternatives. |
| Pricing Model Shift | CustomerLake does not include a traditional platform fee, instead charging based on compute consumption for activities such as identity resolution and profile creation. | The move challenges the conventional CDP pricing model and could pressure standalone vendors that rely on platform licensing revenue. | Compare projected compute-based costs against current licensing expenses using your actual customer data volumes and workloads. |
| Industry Direction | Analysts and vendors increasingly expect customer data capabilities to become embedded within cloud data platforms rather than operating as separate middleware layers. | Competitive differentiation is shifting from data collection toward orchestration, activation, governance and decisioning. | Review where your technology investments create unique business value beyond basic customer data management. |
| Agent Governance | CustomerLake introduces Campaign Agents capable of building audiences, recommending actions and optimizing campaigns with greater autonomy. | As marketing agents gain decision-making authority, governance, permissions and auditability become critical operational concerns. | Establish clear approval workflows, spending thresholds and monitoring controls before deploying agentic marketing capabilities. |
| Competitive Opportunities | Independent CDPs and orchestration vendors continue to emphasize strengths such as multi-cloud flexibility, compliance support and channel-specific expertise. | Large data platforms do not automatically solve every governance, trust or regulatory requirement facing enterprise organizations. | Determine whether compliance obligations, data sovereignty requirements or multi-cloud strategies justify maintaining an independent customer data layer. |
What Enterprise Teams Should Do Before Their Next CDP Renewal
If your enterprise is already invested in these platforms, the moves for the months ahead are:
- Inventory what you pay for against what the platform now does for free. Scope CustomerLake's coverage line by line versus the incumbent's invoice before any 2027-or-later renewal.
- Pilot first before you thinking about replacing. A beta is used as leverage in a negotiation and a classroom exercise. This is not a migration requirement this year.
- Address identity and governance first. Users will do what you enable them to do with your profiles. Velocity against bad data breeds insanity, and consistency at scale is the real job. The platform rarely delivers this by itself.
- Consider the level and location of autonomy in your system. The path of least resistance allows an agent to sell and approve its own work. Choose where a human needs to sign off, and insist on the audit trail, before the path of least resistance defaults for you.
- Reduce the bottleneck between data and marketing through upsklling. Focus on improving processes between the two teams will make or break your effort. The platform is the easy part.
Frequently Asked Questions About Databricks CustomerLake and the Future of CDPs
As lakehouse-native customer data platforms gain traction, enterprise teams are weighing what CustomerLake means for their current CDP investments, vendor contracts and data governance strategies. The following questions address the most common points of uncertainty.
Why Long-Term Architecture Decisions Matter More Than the AI Label
Labeling trends come and go. Electric fell out of “electric light” because electricity became table stakes infrastructure. Agentic and AI will fall out of fashion when they too become table stakes. Look for the fundamentals (orchestration and decisioning) not the buzzword vendor trying to sell it to you.
Webinar
Jun
25
The Hidden Cost of Fragmented Customer Communication
Discover why growing businesses are rethinking the systems, workflows and communication habits shaping customer experience.
Register
Webinar
Jun
30
How Modern Marketing Is Exposing the Limits of Legacy CMS
Why marketers are rethinking CMS workflows that slow publishing, personalization and campaign execution.
Register
Webinar
Jul
9
Why Some Dealers Are Pulling Ahead With AI
AI results vary. Learn where dealers win and where they get stuck.
Register
Webinar
Jul
14
Content Leaders Collective: Proving Content’s Business Impact
Join us as top content leaders look beyond the buzzwords to share how they actually prove ROI and scale what works.
Register
Webinar
On demand
The 5-Question CX Audit: Benchmark Your CX Operations for 2026
Built around insights from the 2026 CX Outsourcing Report, this live session puts the audit into practice.
Watch Now
Webinar
On demand
From Legacy to Launch-Ready: How Gainbridge Made Its Website a Marketing-Led Growth Engine
Join in to learn how a D2C annuity brand gave marketing full website ownership — without slowing down or risking compliance.
Watch Now
The industry built CDPs to solve data silos, only to create more silos. Data platforms are making the same promises today. Own your data, obsess on data governance and keep your teams coordinated. Stop buying into the marketing adjective. Buy into the long term architectural positioning.
The critical thing to include on your short list of critical decisions is how you want to position your customer data, and how far removed you want it to be from orchestration and decisioning. The industry offered CDPs to cure customer data fragmentation, only for the CDP itself to become another fragment. This latest approach to customer data makes similar promises yet opens new doors. The enterprise can own the data and the governance, keep the people who run both close to the work and let the surface get rebuilt around you as many times as it needs to.
Stop buying the adjective that describes the function. Buy the position that a platform holds in your technology stack, and focus on the speed, quality and cost of delivery.
Learn how you can join our contributor community.