Databricks CustomerLake is Databricks’ new Agentic CDP, built inside the lakehouse rather than as a bolt-on, betting that banks would rather govern and understand their data once than maintain a separate CDP copy of a customer record they already own.
Every bank has a customer record. Most have too many versions sitting in different systems. That’s not just frustrating; it’s a dangerous environment for any regulated company.
That’s why Databricks’ Data Intelligence Platform enters the industry at an interesting moment for the CDP market, and CustomerLake, its newest and most CX-specific application, is the sharpest current test of whether the platform underneath it can actually back up the claim of better data management.
Instead of forcing banks to continue chasing a single customer view through dedicated CDPs stacked on top of a data warehouse, Databricks is arguing that’s backward: govern and understand the data properly first, inside the platform banks already run their core analytics on, and the customer view falls out of that rather than needing its own separate system.
In the customer data space for finance, this isn’t an architecture argument anymore. It’s a control argument. If zero copy customer data can cut duplication without weakening activation, and if customer data governance can travel with every profile, audience, and AI recommendation, Databricks has a serious banking CX story.
Related Articles
- Customer Data Management: The Complete CX Guide
- Single Customer View: Where CRM Falls Short
- Databricks CustomerLake: Inside the AI-Native CDP
TL;DR: Is CustomerLake Ready for Banking CX?
- The pitch is architectural, not just a feature list: CustomerLake avoids the usual CDP copy problem by keeping profiles and activation close to governed data, not a separate replica.
- Financial services now has a second, sharper proof point: Clearlake Capital just partnered with Databricks and West Monroe to run AI-enabled investing across deal origination, due diligence, and portfolio operations.
- Platform-wide evidence backs the real-time claim: Databricks’ Lakebase scaled a workforce agent to 130,000 consultants at DXC, cutting task time by 94%.
- The gap: CustomerLake is still in Private Preview, and public, named banking outcomes for the product specifically are still thin.
What Is the Databricks Data Intelligence Platform?
The Databricks Data Intelligence Platform is Databricks’ lakehouse for running data engineering, warehousing, analytics, governance, and AI against the same data estate. It uses Delta Lake for storage, Unity Catalog for permissions and lineage, and Mosaic AI for model and agent development.
For regulated companies like banks, the intelligence platform offers the architecture. Payment data, product holdings, fraud alerts, CRM records, and risk attributes can stay connected in one secure system, without forcing the team to build different versions of the customer.
Customer Lake sits on top of that foundation, applying the platform capabilities to identity resolution, customer profiles, audiences, and activation.
What Is DatabricksIQ?
DatabricksIQ is the layer that gives the platform context about the data inside it. It reads metadata, lineage, queries, usage history, table relationships, and company terminology to work out how different assets relate.
That context appears in Genie, where users can question data in plain English, and Genie Code, where technical teams can generate code, fix pipeline errors, and build dashboards. It also improves search inside Unity Catalog and helps Databricks tune workloads around the way the environment is actually being used.
In finance, DatabricksIQ can connect terms such as disputed payment, available balance, vulnerability flag, and product eligibility to the tables and rules behind them. It still relies on the bank defining those terms properly in the first place.
What Is Databricks CustomerLake?
Databricks CustomerLake is Databricks’ new Agentic CDP, launched in June 2026 and built inside the Databricks lakehouse. It launched at a time when the CDP market was already being pulled toward AI agents, composable data stacks, and cleaner governance, and Databricks has walked straight into that fight with a very pointed claim: customer profiles don’t need to live in another copied database to become useful.
The AI-native product brings Customer 360 profiles, identity resolution, audience building, campaign automation, activation, and personalization into Databricks. It’s still in Private Preview, so buyers shouldn’t treat it as battle-tested CDP replacement material yet. Launch customers include HP, Circle K, AB InBev, and Getnet by Santander.
CustomerLake isn’t another marketing dashboard with a fresh coat of paint. Profile Agents, Campaign Agents, Agentic Identity Resolution, the Real-Time Profile API, Genie, Lakeflow, and Unity Catalog all come in the same package. Databricks wants teams to build profiles, shape audiences, activate records, bring in data, question it, and govern it without bouncing between platforms.
Learn more about customer data management use cases in 2026 here.
Why Is Financial Services the Strongest Vertical Test for CustomerLake?
Databricks isn’t focusing exclusively on finance here, but financial services is the harshest place to test CustomerLake because banking data has consequences baked into it. A retail brand gets a segment wrong, and someone gets an odd offer. A bank can trigger a complaint, expose sensitive information, or recommend a product the customer should never have seen.
The customer record also has to reconcile product holdings, transactions, risk signals, complaints, consent, vulnerability, and eligibility. Each attribute can change what the bank is allowed to say or do, which makes stale context far more dangerous than a weak marketing segment.
Databricks’ April 14th, 2026, Financial Services Outlook reports that around 94% of financial services firms are piloting or using generative AI across customer functions, fraud, and risk workflows. That means the standard for the data underneath those systems is rising.
Every copied profile gives consent, eligibility, and risk context another chance to drift. Regulators won’t care which system missed the update. They’ll still expect the bank to explain the decision.
CX Awards 2026 Are Open
See which CRM, data, and customer intelligence platforms are shortlisted across every category this year.
Can Databricks Replace a Dedicated CDP for Banks?
For some banks, yes. For most, CustomerLake changes where the CDP job happens before it removes the dedicated product altogether.
CustomerLake can take on a lot of CDP work, from Customer 360 and identity matching to audiences, activation, campaign automation, personalization, and real-time profile access. Its best case is a bank where Databricks already holds the important customer, payments, risk, product, and service context. The less data copied out, the fewer arguments later.
In a Databricks-heavy estate, CustomerLake may replace the existing CDP. Elsewhere, it is more likely to sit underneath one, supplying governed profiles while the incumbent handles the campaign tools marketers already know.
Many companies are already moving away from pulling every record into one database, and toward connected data linked through shared identifiers. CustomerLake fits that architectural shift, even though the operating model will differ from bank to bank.
Salesforce Data Cloud, Adobe Real-Time CDP, and Twilio Segment still have an advantage where day-to-day marketing work gets messy. They offer more packaged consent workflows, activation templates, channel connectors, and tools built for people who don’t spend their days inside a data platform. When quarterly targets are looming, speed and usability can beat a cleaner technical design.
The decision buyers can’t sidestep is whether Databricks becomes the customer-data command center or remains the governed data source beneath another engagement system.
How Does Databricks CustomerLake Avoid Copying Data Into Another CDP?
CustomerLake avoids the usual CDP copy problem by keeping profiles, identity logic, audiences, models, and activation close to the data’s governed home instead of exporting a duplicate into a separate customer database.
For banks, copied profiles age badly. Consent changes, risk scores move, disputes are resolved, and customers become ineligible for offers. A stale copy can remain technically available long after it has stopped being safe to use.
Databricks says CustomerLake can use Lakehouse Federation to work with trusted data across Databricks workspaces and connected systems before anyone starts duplicating records. Banks still need to push on the real-time claim. Some CX stacks lose 5% to 15% of expected events and still call the feed live.
The useful test is deliberately unpleasant. Kill a source, break an event, change a consent status, then force a write-back into the governed profile at speed. If zero copy customer data survives that mess, the payoff is concrete: fewer copies, fresher customer context, and fewer systems inventing their own version of governance.
What Role Does Unity Catalog Play in Customer Data Governance?
Unity Catalog is where CustomerLake’s banking argument becomes credible. Without it, the Lakehouse vs CDP debate would amount to a data-team preference dressed up as CX strategy.
Banks can’t stop at identity. They need a trail for every attribute: source, user, policy, and consent state when a decision was made. Databricks used Data + AI Summit 2026 to expand Unity Catalog with cross-cloud and cross-region governance, a four-level namespace, and unified discovery. DatabricksIQ then reads that governed metadata, giving the platform context about what a field represents instead of leaving governance as a side process.
In late May 2026, Informatica, now part of Salesforce, announced deeper interoperability with Unity Catalog managed tables, too. That’s good ecosystem validation: another major data vendor is connecting to the governance layer rather than asking customers to route around it.
A bank should test whether those controls survive real work. Compliance teams need to trace audience inclusion, suppression, retention, regional access, and AI recommendations without relying on a manual explanation assembled after the event.
CustomerLake’s value depends on Unity Catalog holding that line while marketers, service teams, fraud systems, and agents act on the same profile.
What Evidence Supports Databricks’ Financial Services Case?
Databricks has a credible financial-services platform story, but the public record for CustomerLake itself is still much thinner.
Getnet by Santander is the CustomerLake reference worth watching. Its payments and merchant context fits the product neatly, but Databricks hasn’t yet published the kind of production metrics that would show faster activation, stronger profile matching, or lower service effort.
The wider platform has more substance. NAB uses Genie Spaces to examine customer disputes through transcripts, intent, sentiment, ambiguity, and resolution patterns. In Databricks’ published customer story, HSBC replaced 14 databases with one Delta Lake and cut complex analytics from six hours to six seconds. Coastal Community Bank reports 66% faster vendor due diligence reviews and 94% classifier accuracy, per Barb MacLean, Coastal’s SVP and Head of Technology Operations and Implementation, in Databricks’ customer story.
Clearlake Capital’s partnership with Databricks and West Monroe extends the finance case into deal origination, due diligence, and portfolio operations. DXC provides adjacent proof of scale: a Lakebase-backed workforce agent reached 130,000 consultants and reduced task time by 94%. Both strengthen the platform case, although neither is a CustomerLake banking outcome.
Databricks has enough evidence to justify a serious evaluation. It still needs a wider bench of named CustomerLake deployments in regulated finance before buyers can treat the product as a proven banking CDP replacement.
Is CustomerLake Ready for Regulated Financial Services?
CustomerLake is ready for serious evaluation. It’s not ready for a victory lap. CustomerLake remains in Private Preview. Lakehouse//RT is in Beta, and some Salesforce Zero Copy capabilities arrive across H2 2026 rather than being available today.
That makes readiness conditional. The governed lakehouse, Unity Catalog, and federation model are credible foundations for regulated data. The CustomerLake layer still needs more evidence around marketer workflows, operational reliability, and regulatory review in live banking environments.
Banks should separate platform strength from product maturity. Ask the right questions:
| Test Area | What the Bank Should Ask | Evidence to Request |
|---|---|---|
| Ownership | Will CustomerLake replace the CDP, sit beneath it, or feed it? Which system owns identity, consent, and the golden record? | Signed architecture and data-ownership map |
| Governance | Can compliance trace inclusion, exclusion, retention, regional access, and an AI recommendation? | Lineage views, policy tests, access logs, and audit replay |
| Activation | Which destinations are native, which need custom work, and how does the platform interact with Salesforce? | Connector inventory, failure handling, and measured activation time |
| Real time | How fresh are profiles, and what happens when events are late, missing, or duplicated? | Latency SLA, reconciliation results, and fallback procedure |
| Operating model | Can marketing and service teams complete routine work without continuous engineering support? | Hands-on sandbox test with real users and timed workflows |
| Proof | Which results come from CustomerLake in finance rather than the wider Databricks platform? | Named references, production metrics, and regulatory-review outcomes |
Banks don’t need another promise that governance works. They need proof it survives campaign launches, customer complaints, service actions, automated model decisions, and an auditor asking who touched what.
Databricks Belongs in the CX Data Strategy Conversation
For banks, most customer-data projects fail somewhere between the profile and the moment of action. The dashboard looks right. The segment looks useful. Then consent is unclear, the fraud signal is stale, the service team can’t see the same context the marketing team just used, and the whole pitch collapses under its own audit trail.
For customer data in finance, Databricks’ best argument is control. CustomerLake, Unity Catalog, Genie, Lakehouse//RT, and the Salesforce Zero Copy work all point at the same idea: govern the data once, and let every downstream use- marketing, service, risk, fraud- inherit that governance rather than rebuilding it.
But banks should stay demanding. Databricks CustomerLake still needs more public, named financial-services outcomes. Getnet by Santander is a good sign; Clearlake Capital’s new partnership extends the credibility into investment operations. Neither is a substitute for a wider bench of regulated banking references yet. That bench needs to grow before buyers can treat CustomerLake as a proven banking CDP replacement.
Explore the Best Data Strategy for CX Award
See how Databricks holds up against the competition in the CX Awards 2026.
Frequently Asked Questions
Does CustomerLake Include Consent Management Out of the Box?
CustomerLake carries governance and consent context through Databricks, but buyers should not assume it matches every packaged consent workflow in a mature CDP. Banks need to test preference capture, suppression, purpose limits, retention, and regional rules across real audience builds and activation destinations.
Can a Bank Run CustomerLake Alongside an Existing CDP?
Yes. CustomerLake can hold the governed profile and identity logic while an existing CDP continues to manage familiar campaign workflows and connectors. That coexistence model may be the safest first step for banks that want fewer data copies without forcing marketing teams through an immediate platform replacement.
Does Zero Copy Mean Customer Data Never Moves?
No. Zero copy reduces the need to create another persistent customer database, but data may still move through queries, caches, activation payloads, exports, or downstream applications. Banks should map each movement, confirm which policies travel with it, and test how quickly consent changes reach every destination.
Does CustomerLake Change the Contact Center Story?
It can, provided the bank connects it properly. Agents could receive current dispute status, digital behaviour, product holdings, and vulnerability context without hunting across separate systems. The value disappears if the profile is stale, the service desktop cannot consume it, or sensitive attributes reach staff without the right controls.
Where Does Salesforce Fit If Databricks Owns the Data Layer?
Salesforce doesn't vanish from the stack. It can remain the place where CRM work, service cases, routing, and agents meet the customer. Databricks can hold the governed data foundation underneath. The buying team needs to test the plumbing: Zero Copy access, write-backs, consent changes, and failure handling across Data 360, Marketing Cloud, Service Cloud, and Agentforce.
