Why data governance is the missing link in enterprise AI ROI

Bridging the AI value gap

Despite enterprise AI adoption accelerating, many organizations are struggling to demonstrate tangible returns. According to research conducted by Iron Mountain and FT Longitude, fewer than 10% of businesses report that AI has materially boosted revenue or profitability—a challenge rooted not in the capabilities of AI models, but in the underlying quality and structure of enterprise data.

In a feature by The Washington Post and an executive interview with Investment Reports, Mithu Bhargava, Executive Vice President & General Manager of Digital Solutions at Iron Mountain, highlights why strong data management is essential to bridging this gap.

"The issue isn't the AI models themselves—it's the state of enterprise data," notes Bhargava. "AI learns from data, so if that information is inaccurate, outdated, fragmented, or poorly governed, the outcomes will be unreliable."

Establishing enterprise-wide ownership

Bhargava points out that a primary obstacle facing modern organizations is that fewer than half are effectively making unstructured, “dark” data accessible for AI operations. Furthermore, because data often spans decades of legacy physical records, digital files, and siloed systems, no single department truly owns it.

To bridge this disconnect, leaders must move beyond treating AI as an isolated IT project. Sustainable AI adoption requires company-wide governance, shared accountability across business units, and ongoing human oversight. By building trust into every stage of the data lifecycle, enterprises can minimize compliance risks and ensure AI-driven decisions are transparent, trusted, and directly tied to business outcomes.

Read the full feature in The Washington Post and explore Mithu’s complete interview with Investment Reports.