Data-related problems delayed or derailed AI projects at 86% of organizations over the past year, according to a new report from Iron Mountain released October 19. 

The research, The Information Supercycle: Aligning Data, Infrastructure and Capital for AI Scale, was unveiled at Gartner IT Symposium/Xpo in Orlando and examines the information-management challenges companies face as they expand AI adoption.

The findings point to a persistent problem for chief data officers (CDOs): organizations want to use data to improve business outcomes, but fragmented systems, legacy formats and manual preparation leave much of their information insufficiently prepared, accessible or organized for AI.

Although 95% of respondents said using data to improve business outcomes was a priority, only 39% reported having data that was digitized, structured and readily usable for analytics and AI. 

The report also finds that 70% of organizations cannot effectively track the lifecycle of data used in their AI models, highlighting an AI governance challenge as businesses seek to deploy AI more widely.

Data Preparation Continues to Slow AI Delivery

Manual work remains a substantial burden. Four in 10 organizations spend at least 41% of their AI project timelines on manual data preparation, according to the report.

Fifty-six percent of respondents cited data spread across disconnected systems or departmental silos, while 53% identified legacy formats or physical archives as obstacles to making information usable for AI.

For CDOs, these findings point to the operational consequences of fragmented data environments. When information is difficult to locate, reconcile or prepare, teams spend more time making it usable and less time developing and deploying AI applications.

Limited Data Lineage Creates Governance Risks

The finding that 70% of organizations struggle to track the lifecycle of data used in AI models raises a separate concern: visibility into how information is sourced, managed and used.

Without that visibility, organizations may find it difficult to establish whether data is appropriate for a particular use, whether access is authorized or how information contributed to an AI-generated output. 

These limitations can complicate compliance reviews, audits and investigations into AI-related incidents.

This makes the relationship between AI governance and data governance increasingly important. AI governance establishes controls around the development and use of AI systems, while data governance helps ensure the underlying information has appropriate ownership, access permissions, quality standards, and traceability.

Data Readiness Needs To Be a Priority

As organizations expand their use of AI, CDOs face pressure to address information-management issues that can undermine projects before models reach production. 

That includes improving accessibility, reducing fragmentation and establishing clear visibility into how information moves through the enterprise.

The findings also have implications for organizations deploying AI agents that can take actions across business workflows. Such systems need access to information that is reliable and authorized for the intended purpose, alongside controls that make its use traceable.

For CDOs, the challenge is to make these capabilities part of the enterprise data environment rather than treating them as project-specific fixes. Without that foundation, organizations risk encountering the same barriers as they move from AI experimentation to broader deployment.