What happened
HP described an enterprise marketing-data transformation built on Databricks lakehouses, CustomerLake and AI/BI Genie. The company says it consolidated fragmented data, created governed semantic definitions for marketing metrics, expanded intent-data ingestion and is bringing identity signals into its Databricks environment. HP reports that comparable governed data-sharing requests now take about 48 hours instead of roughly three months, while audience preparation can take about two hours instead of two or three months.
HP says its broader data program consolidated hundreds of data lakes and repositories into a smaller number of Databricks lakehouses, using central standards with federated execution. Marketing teams gained governed access to media performance, market-share, technographic, intent and propensity data while domain experts retained ownership.
The company is building semantic layers for Genie that include domains, metric views, instructions and reusable skills. One media metric view contains 26 dimensions and 10 measures intended to give dashboards, business users and AI systems shared definitions for media-spend and performance metrics.
HP says it expanded intent-data ingestion from 75 topics to approximately 20,000 topics and is using Genie Code to plan how those signals can be clustered into richer account profiles. It also plans to bring identity signals back into Databricks for measurement, audience activation and model improvement.
According to the source, a data-sharing request that took three months, about 15 email exchanges and three escalations in 2024 can now produce a governed Delta Sharing asset within 48 hours. HP also says CustomerLake made its team audience-ready in two hours rather than the two or three months expected from previous CDP implementations.
Source details: databricks.com ↗
Why it matters
The example shows how AI deployment depends on governed business context, not just access to raw data. If HP’s reported results are representative, combining data definitions, identity, analytics and activation in one environment could reduce duplication and make AI-assisted marketing more usable. However, the source is a Databricks case study and does not independently verify the claimed savings, performance or business impact.
The central implication is architectural: HP is treating a customer-data platform as a set of capabilities built on an existing governed data foundation rather than as a separate system that duplicates pipelines, definitions and controls. That can be important for AI agents, which may retrieve data but still produce unreliable results if they lack metric definitions, relationships and governance.
The source also describes a B2C-to-B2B use case. HP says CustomerLake can connect its large consumer customer base with a smaller population of known business contacts through a shared identity layer, while Lakehouse Federation can incorporate data from other platforms.
These are company-reported outcomes and expectations, not independent test results. The source does not provide a detailed cost baseline, deployment timeline beyond the examples cited, model-accuracy measurements, conversion results or evidence that the potential 40% reduction in CDP total cost of ownership has been achieved.
Interactive Mechanism: How It Actually Works
Explore the underlying technology behind this development interactively.
crm_get_transaction(id='4092').A route planner searches possible journeys using explicit rules. What does this illustrate about AI?
What to watch next
The key unresolved questions are whether HP achieves the reported cost reduction at scale, how accurately Genie and agent-based systems use the semantic definitions, and whether the consent-preserving identity approach reaches a meaningful share of customers. Access, pricing and general availability of the described CustomerLake and Genie configuration are not documented.
HP’s stated roadmap includes evaluating value from faster data access, broader customer context and more efficient activation. Evidence of realized revenue, marketing efficiency or model-quality improvements would clarify whether the architecture delivers more than operational speed.
Consent and identity resolution will be consequential. The source says HP intends to use a consent-preserving approach and suggests that reaching 10% of its consumer customer base could create a substantial commercial opportunity, but it does not quantify that opportunity or describe the consent controls in detail.
The source does not state CustomerLake or the described Genie configuration’s price, licensing terms, release status or availability to organizations outside HP. It also does not establish that other enterprises would see the same results.