TL;DR: In a recent episode of the Edgematics Data Enablers podcast, I made the case that most organizations already have the data they need. The problem is it was never designed to be used. Governed, reusable data products close that gap, and AI is accelerating the work, but only when governance foundations are already in place.
In 20 years working in data management, I have seen the same problem repeat itself in different forms. Organizations invest heavily in their data estates. They build pipelines, deploy platforms, hire analysts, and commission dashboards. And yet, when a business leader needs a timely, reliable answer, the journey to get there still takes weeks or months. The data exists. The problem is that it was never designed with practical use in mind.
That is the central challenge data products were created to solve. A data product is not simply a dataset or a report. It is a curated, governed, reusable asset built around a specific business outcome. It has an owner, a quality standard, and a lifecycle. And when it is done well, it democratizes access to insight across an organization rather than concentrating that access in a small technical team that becomes everyone’s bottleneck.
The question organizations are wrestling with now is not whether data products matter. Most data leaders have already crossed that threshold. The question is how to create them faster, make them easier to consume, and extend their value across the enterprise rather than sitting unused in a catalog.
“The shift from data access for the few to data empowerment for the many requires more than a new toolset. It requires a new way of thinking about what data is for and who it belongs to.”
Why do data teams still face the 80/20 problem?
Most data teams spend far more time preparing data than acting on it, and the cause is structural, not technical. The traditional approach to data products was painful by design, even if no one intended it that way. Multiple teams worked in silos. The technical teams owned the data and managed it according to their own priorities. The business teams knew what decisions they needed to make but had little visibility into whether the data to support those decisions existed, where it lived, or whether it could be trusted.
Glenda O’Keefe, Field CTO at Quest Software, on why the biggest obstacle to data-driven decisions is not a lack of data. It is data that was never designed to be used.
I speak with enterprise data leaders regularly, and one pattern comes up more than any other: teams spending the vast majority of their time finding and preparing data, with only a fraction left for the analysis that actually drives decisions. One customer I’ve worked with recently told me they were spending 80% of their time finding, preparing, and trying to understand the data, and only 20% actually getting results from it.
That ratio is not a talent problem or a process problem. It is a structural problem. When data is not productized, not curated, not governed, not made discoverable and reusable, every consumer starts from scratch. Every new use case reinvents the wheel.
The shift from data access for the few to data empowerment for the many requires more than a new toolset. It requires a new way of thinking about what data is for and who it belongs to.
“A mature data product requires four things: clear ownership, lifecycle management, quality assurance, and reusability by design.”
What does a mature data product actually require?
One of the first things I work through with customers is the question of definition. Every organization I speak with has a slightly different understanding of what a data product is. The technical teams tend to think in terms of pipelines and schemas. The business teams think in terms of reports and metrics. Neither is wrong, but the gap between them is where data products go to die.
A mature data product requires four things: clear ownership, lifecycle management, quality assurance, and reusability by design.
- Ownership means someone is accountable for the quality and relevance of this data product, and that person is not just a data engineer.
- Lifecycle management means the product is curated, updated, and retired according to a defined process.
- Quality assurance means the data has passed through validation and meets a defined standard before it is published.
- Reusability means the product is designed from the outset to be consumed by multiple teams, not built once for a single use case and abandoned.
That fourth point is the one I push hardest on. Reusability is not just a design principle; it is the mechanism through which data products create compounding value. The more a data product is used across the organization, the more valuable it becomes, because its quality signal strengthens, and its relevance to real business decisions becomes demonstrable. An unused data product sitting in a catalog is an asset on paper. A reused one is an asset in practice.
“AI does not eliminate the need for governance; it accelerates your ability to get to a state where governance is meaningful.”
How is AI lowering the barrier to data product creation?
Two shifts are making data product creation genuinely accessible to teams that were locked out of it before: natural language interaction with data, and a move from reactive to proactive quality management. What has changed most visibly in the last two years is the role AI plays in removing the technical barrier to data product creation. Historically, building a governed data product required deep engineering resources. The process was slow, documentation was inconsistent, and the business stakeholders who most needed the output had little ability to participate in the creation.
AI changes that equation in two important ways. The first is in creation speed. With the right platform capabilities, a data analyst can describe a business outcome in natural language and have the system scan the existing data estate, identify relevant assets, and generate a working data model in minutes rather than weeks. The second is in quality maintenance. AI can detect data drift, identify anomalies, and recommend quality rules proactively rather than waiting for a downstream consumer to discover that something is wrong. That shift from reactive to proactive quality management is significant, particularly at scale.
I want to be careful about overclaiming here, however. AI does not eliminate the need for governance; it accelerates your ability to get to a state where governance is meaningful. A mature, curated data estate gives AI the foundation it needs to produce reliable outputs. If the underlying data is ungoverned and inconsistent, AI amplifies that inconsistency at speed. The organizations seeing the most value from AI-assisted data product creation are the ones that invested in the governance fundamentals first.
In a follow-up post, I will look at what happens when organizations delay that investment, and what it actually costs them when the market moves faster than their data can.
This post expands upon my conversation with Sarit Bose of Edgematics on the Data Enablers podcast.
