Expert Mode: Why Your Modern Data Warehouse is Sabotaging Your AI Strategy
This article was based on the interview with Ravi Shankar, Senior Vice President and Chief Marketing Officer at Denodo by Greg Kihlström, AI and MarTech keynote speaker for The Agile Brand with Greg Kihlström podcast. Listen to the original episode here:
As marketing leaders, we find ourselves in an interesting position. We are champions of innovation, the primary sponsors of multi-million dollar AI initiatives, and the stewards of the customer experience. We’ve fought the good fight for budget, invested heavily in modern data platforms like Snowflake and Databricks, and have been assured that our data house is finally in order. We’ve been sold the dream of a centralized data repository, a single source of truth that will unlock unprecedented personalization and predictive power. And yet, for many of us, the reality falls short. Our AI projects stall, our campaign insights arrive a week too late, and true real-time CX remains frustratingly out of reach.
The temptation is to blame the AI models, the algorithms, or the talent gap. But the real culprit is often hiding in plain sight, masquerading as the solution itself: our data infrastructure. The very platforms we’ve relied on to solve data fragmentation may be inadvertently creating newer, larger, and more expensive silos. The paradox is that in our relentless pursuit of centralization through physical data consolidation, we’ve created a system that is often too slow, too rigid, and too disconnected from the real-time pulse of our business. It’s a strategic disconnect that silently sabotages our most ambitious goals, and it’s time we, as business leaders, understood the architectural shift required to fix it.
The Paradox of Modern Centralization
For decades, the answer to data chaos has been to centralize it. From databases in the 80s to data warehouses in the 90s and cloud lakehouses today, the pattern is familiar. The problem is that with each new wave of technology, the promise of a single, unified platform has failed to materialize in the complex reality of enterprise IT. Marketing has its stack, finance has its own, and operations runs on something else entirely. As leaders, we look at this and wonder why, after all this investment, the landscape is still so fragmented. Ravi Shankar points out that this isn’t a failure, but a feature of how enterprise technology is deployed.
“If there is only one platform that you need as advertised by these vendors to store all the data, why is it that companies have Snowflake, Databricks, Oracle, Teradata? They have multiple of these technologies. That’s the paradox… IT provisions a repository just for them, for that particular purpose, and makes it available to them. So when they do that for marketing, sales, finance, and so on, they create multiple silos.”
This is a crucial point for marketing leaders to grasp. The issue isn’t that our IT counterparts are incompetent; it’s that they are serving specific business needs with purpose-built tools. The analytics-heavy lakehouse that powers our BI dashboards is fundamentally different from the operational databases running our e-commerce platform. The result is a collection of powerful, yet disconnected, “centralized” platforms. We haven’t eliminated silos; we’ve just built bigger, more expensive ones. And the traditional method of bridging them—painstakingly extracting, transforming, and loading (ETL) data from one to another—is where our agility goes to die.
The High Cost of Data Latency
This architectural problem is not some abstract IT concern; it manifests in very real, very painful ways for marketing. The lag time between an event happening in the real world—a customer downloading an ebook, a lead engaging with a campaign—and that data being available for analysis is the gap where opportunity is lost. We talk about real-time personalization and in-the-moment optimization, but our infrastructure often forces us to make decisions based on last week’s data. Shankar illustrates this with an all-too-familiar marketing scenario.
“By the time all this data is loaded and the analysis is done and I gain the insights, the initial system…has become out of sync with the source already. So the information that I get is not relevant anymore… Consider…a retailer. I’m running a day-long promotion…I want to understand [if] this campaign is working in the first half of the day so that way I can fine tune it… But the problem is…it’s not going to be ready for us…to kind of do that real-time campaign check-in and ultimately both the retailer and the consumers also lose.”
This is the silent killer of marketing ROI. We launch a campaign, and by the time the data is collected from our CRM, marketing automation platform, and web analytics, moved to the lakehouse, and processed by our analytics team, the critical window for action has closed. We’re left managing campaigns by looking in the rearview mirror. For any AI initiative, this latency is fatal. AI models are only as good as the data they’re fed, and if that data is stale, the predictions and personalizations will be suboptimal at best, and value-destroying at worst. The business demands speed, but our physical data consolidation approach is an anchor, not an engine.
Thinking Logically, Not Physically
If moving all the data to one place is the problem, the solution is elegantly simple: don’t move the data. This is the core principle of a logical data management approach, powered by technologies like data virtualization. Instead of collecting all the data into yet another repository, you create a virtual layer that connects to the data where it lives. This layer provides a unified, real-time view across all your disparate systems—your CRM, your data warehouse, your cloud applications—without the costly and time-consuming process of physical data replication. Shankar uses a brilliant analogy to clarify the difference.
“The logical approach allows the data to reside wherever it is…and it provides a unified view of that information…without having to pull all the data into yet another repository… The logical data integration is like taking your cup to the source, because all you need…is just that one cup of water. You’re not going to drink 20 gallons of water. It is much more faster for me to go, get that one cup and come back and give it to you, rather than…lifting this 20-gallon pitcher and then bringing it all the way back to you.”
This is a fundamental shift in thinking. We’ve been conditioned to believe that data must be physically co-located to be useful. The logical approach proves otherwise. It’s about being smart and efficient—requesting just the answer you need (“How many leads from the new campaign converted in the last hour?”) and letting the system query the underlying sources and bring back the result, rather than moving terabytes of raw data just to answer a simple question. This is what enables the 10x acceleration in AI rollouts and 75% reduction in integration time seen by companies that adopt this model. It’s not magic; it’s just smarter plumbing.
Future-Proofing Your Strategy with Data Abstraction
Perhaps the most compelling benefit for us as strategic leaders is not just solving today’s speed-to-insight problem, but building an architecture that is resilient to future change. The technology landscape is in constant flux. The data platform that is best-in-class today may be legacy tomorrow. A logical approach creates a “data abstraction layer” that insulates the business from this underlying complexity. It acts as a stable, consistent middleman between the business users who need data and the complex web of IT systems that store it.
“A data abstraction is a middle layer…that disintermediates the business users from the IT… business users come to this data abstraction layer and consume the data from there. And at the same time, the IT can move at the speed that they need to under the cover of the data abstraction to slowly change out the systems, modernize and so on… Business users don’t care where the data comes from and the IT can take the time to modernize it.”
This decoupling is profoundly strategic. It means your marketing analytics team, your AI models, and your CX platforms can all consume data from a single, stable, logical access point, regardless of whether IT decides to migrate from one cloud provider to another or swap out a legacy CRM. It allows the business to move at its own rapid pace, while giving IT the freedom to evolve the technical infrastructure without causing massive disruption. This is how you build an agile, future-proofed organization. You stop tying your business logic directly to your physical data storage and create a flexible layer that can adapt to whatever comes next.
The path to becoming a truly data-driven, AI-enabled marketing organization is not paved with more ETL scripts and bigger data warehouses. It requires a fundamental re-evaluation of how we access and integrate information. The shift from a physical-first to a logical-first data management strategy is the key to unlocking the agility, speed, and governance that our modern initiatives demand. It’s about working smarter, not harder, and recognizing that connecting to data is often far more powerful than collecting it.
As marketing leaders, we can no longer afford to view data architecture as a problem for the IT department to solve. Our ability to deliver personalized experiences, optimize campaigns in real-time, and generate value from AI is directly constrained by it. The next time a major technology initiative is on the table, our first question shouldn’t be about the features of the new AI model. It should be about the plumbing. We need to ask: is our data strategy an accelerator, or is it an anchor? The success of everything we hope to achieve depends on the answer.
