Data concerns have delayed or derailed AI projects at 86 per cent of organisations in the past 12 months. That is despite 95 per cent naming better use of data to improve business outcomes as a top priority.
The figures come from The Information Supercycle: Aligning Data, Infrastructure and Capital for AI Scale, a report released by Iron Mountain on 20 October.
The report says 38 per cent of the organisations that experienced disruption suffered significant delays or outright cancellations, rather than minor slippage.
Government and public sector respondents were 30 per cent more likely than average to report significant delays or cancellations. Insurance was 10 per cent more likely. Healthcare and life sciences, and banking and financial services, were less likely.
Iron Mountain also found that 81 per cent believe AI cannot scale beyond pilots until siloed and unstructured content becomes AI-ready data.
Manual preparation is absorbing much of the effort. Four in 10 organisations spend at least 41 per cent of AI project timelines preparing data by hand. About one in five spend more than 60 per cent.
Only 39 per cent fall into the report's "digitised and structured" category, where dark data is centrally managed, indexed and readily usable for analytics and AI. The remaining 61 per cent are left with trapped data.
Lineage is the weak point
The report also exposes a traceability gap. Seventy per cent of organisations say they cannot effectively track the origin or journey of the data used in their AI models.
Respondents were asked how confident they were that the data used in their AI models was legally compliant, traceable, accessible, unbiased, accurate and complete. On none of the six measures did more than 44 per cent say they were completely confident.
Fragmentation remains the main obstacle. Fifty-six per cent cite information spread across disconnected systems or departmental silos, while 53 per cent point to legacy formats or physical archives.
The report argues that unreadable formats such as scanned PDFs and paper push organisations towards overly restrictive security policies. Those policies can block AI agents from information they need, delaying or derailing initiatives.
Responsibility for information assets is also dispersed. Oversight of physical records, IT assets and data provenance is split. Chief AI and chief data officers, CFOs, chief legal officers and CIOs each hold a significant share.
Budgets are tilted towards upkeep. Fifty-six per cent of respondents spend most of their data and IT budget on maintenance, while 15 per cent spend most of it on innovation.
Only 17 per cent are funding AI through reclaimed IT spending. Just 35 per cent say they can track spending on maintenance versus modernisation very well.
"AI readiness is about much more than deploying the latest technology," said Varun Gupta, Senior Vice President, Global Product and Solutions Marketing at Iron Mountain. "Enterprises need to know that the information powering their AI is accessible, current and governed, with clear ownership and permissions."
"As AI agents move from generating answers to taking actions across business workflows, these foundations become essential to scaling AI with confidence," Gupta said.
Iron Mountain says the consequences of fragmented information extend beyond poor outputs once AI begins acting on decisions. It lists security exposure, compliance and audit risk, higher discovery costs and the loss of institutional knowledge.
The company is calling for continuous lifecycle governance that gives an auditable view of information from creation to destruction.