Building Data as Product in the Age of Artificial Intelligence

IA FORUM MEMBER INSIGHTS: ARTICLE

By Junaith Haja, Senior Data Engineer, AMAZON

Gartner predicts that through 2026, organizations will abandon 60% of AI projects that are not supported by AI-ready data. Artificial Intelligence is no longer confined to research teams or innovation labs. It has become a strategic expectation for modern enterprises. Yet, despite growing investments in data platforms, AI tools, and specialized talent, many organizations are still struggling to turn AI ambition into sustained business value.

In most organizations, data is still treated as a byproduct of applications rather than as a product designed intentionally for consumers. As a result, data platforms scale, but clarity does not. AI, rather than solving this problem, tends to amplify it by accelerating the spread of inconsistency and confusion.

AI does not consume raw data. AI consumes context. Context includes business meaning, ownership, lineage, governance, trust, and reusability. Without these elements, even the most advanced models produce outputs that are difficult to explain, validate, or operationalize. In this sense, AI has raised the stakes for data strategy. It exposes weaknesses that were once tolerable and makes them impossible to ignore.

This is where the concept of Data as a Product becomes essential. Treating data as a product means applying the same discipline used for customer-facing products. Data products are built with clear consumers in mind, whether those consumers are analysts, executives, machine learning models, or operational systems. They have defined ownership, quality expectations, and success metrics. They are discoverable, documented, and designed to be reused across use cases. When data is built this way, it becomes an asset that compounds in value rather than a liability that grows in complexity.

To ground these ideas, the session used Databricks as a reference example, not as a recommendation, but as a concrete way to illustrate how product thinking can be operationalized. Executives resonated strongly with the discussion around Unity Catalog and its role in establishing shared context. By centralizing governance, defining ownership, and standardizing business definitions, platforms like this help shift data from being an abstract technical artifact to something people can understand and trust. The technology mattered less than what it enabled: a common language and a shared understanding of data across analytics and AI.

Advanced AI and SQL capabilities were also discussed, but the conclusion was clear. These features only deliver value when built on top of well-governed, well-understood data products. Without that foundation, AI increases speed but not clarity, and automation simply accelerates poor decision-making.

A recurring question from executives during the session was how to choose between platforms such as Snowflake and Databricks. The discussion reframed this question entirely. The real decision is not which platform to adopt, but which data products to build, who they are for, and how success will be measured. Both platforms are capable. Both can support analytics and AI at scale. Both can also fail if adopted without a product mindset. Technology decisions should follow clarity of purpose, not precede it.

Organizations cannot tool their way out of a mindset problem. Companies that succeed with AI treat data teams as product teams rather than service teams. They align data initiatives to business outcomes instead of delivery milestones. They invest in context, ownership, and governance as first-class concerns. Success is measured not by data volume or pipeline count, but by adoption, trust, and decision impact.

AI is forcing a reckoning. Fragmented data, unclear ownership, and inconsistent definitions are no longer manageable at scale. Leaders now face a clear choice. They can continue investing in platforms while treating data as exhaust, or they can intentionally build data products that power analytics, AI, and decision-making across the enterprise. The organizations represented at the IA Forum lecture already had the technology. What they needed, and clearly recognized, was a shared mental model.

Once that mental model shifts, AI stops being an experiment and starts becoming a durable advantage. Until then our data evangelization and discussion efforts never stops.

Author Disclaimer: The views and opinions expressed herein are those of the Author alone and are shared in a personal capacity, in accordance with the Chatham House Rule. They do not reflect the official views or positions of the Author’s employer, organization, or any affiliated entity.