Data Governance · 3 min read

Iron Mountain InSight targets finance’s 96% manual-prep burden

Iron Mountain’s InSight DXP update targets fragmented finance data; a survey finds 96% of organisations need manual preparation for AI workflows at scale.

By Grace Kim699 words

Key takeaways

  1. Only 34% of financial-services content is connected across systems, while 96% need manual AI-data preparation.
  2. In the 134-person finance and insurance cohort, 89% cite legacy content-management costs and 84% report skilled staff diverted to routine data work.
  3. Vanson Bourne surveyed 400 IT and data leaders across six countries for Iron Mountain.
  4. Finance and insurance respondents reported 32% higher IT-efficiency gains and 21% higher AI-readiness gains than cross-industry peers.
  5. Iron Mountain reports above 96% extraction accuracy on complex instruments and above 99% with human review.
Iron Mountain updates platform to aid AI in finance - CFOtech Australia
Iron Mountain updates platform to aid AI in finance - CFOtech Australia — sergeitokmakov / pixabay

By Grace Kim

Only 34% of financial-services content is connected across systems, according to Iron Mountain’s commissioned research, a fragmentation rate that helps explain why 96% of surveyed organisations need manual intervention before unstructured data can enter an AI workflow.

On November 18, 2026, Iron Mountain updated InSight DXP for financial-services organisations. The company says the release connects enterprise information with business systems and adds automated content-intelligence and governance tools. It aims to manage unstructured data, preserve business context and improve oversight across AI workflows.

The practical problem sits downstream of model choice. Staff in banks and insurers still classify, tag and prepare documents spread across disconnected repositories, the unglamorous work that turns “AI-ready” into a production requirement.

What does the survey say?

Vanson Bourne conducted the survey for Iron Mountain, covering 400 IT and data leaders in the United States, United Kingdom, Germany, Brazil, India and Australia. Respondents worked at organisations with at least 1,000 employees across banking and financial services, insurance, healthcare and life sciences, technology, manufacturing and the public sector.

A cohort of 134 financial-services and insurance decision-makers provides the basis for the findings on content connectivity, manual preparation and legacy-system costs:

  • 96% need manual intervention to prepare unstructured data for AI workflows.
  • 89%, or nearly nine in ten, say legacy enterprise content-management costs limit their ability to invest in AI and modernisation.
  • 84%, or four in five, report that highly skilled IT and data-science staff get pulled from strategic work to complete routine data tasks.
  • 45% identify data classification and tagging as a primary manual burden.

That last number deserves attention. With 45% naming classification and tagging as a primary burden, information hygiene looks less like housekeeping and more like capacity planning.

Financial-services and insurance respondents also reported stronger average gains than cross-industry peers: 32% higher gains in IT efficiency and 21% higher gains in AI readiness. Iron Mountain argues that those results may conceal deeper problems caused by fragmented information management. Better dashboards can coexist with worse foundations. A familiar pattern.

What changed in InSight?

Iron Mountain designed the latest InSight DXP version to strengthen links between information repositories and operational systems, reducing the manual work required to turn fragmented content into governed data for AI. The platform supports document-heavy processes including trade back-office operations, mortgage processing and commercial lending.

Iron Mountain reports extraction accuracy above 96% for complex financial instruments and above 99% when human review is included. The second figure depends on people. That fits the approach many businesses use in regulated environments: automate document processing while retaining checks for higher-risk tasks. Oversight survives, with a heavier workload.

Varun Gupta, Senior Vice President, Global Product & Solutions Marketing at Iron Mountain, framed the commercial problem bluntly: “Financial services firms are investing heavily in AI, but paying a steep operational tax to feed those models trusted, reliable information.”

He also said:

“Connecting that information, preserving its business context and embedding governance into workflows are essential to scaling AI with confidence.

To unlock AI’s full potential, financial institutions need a unified content intelligence layer that automates governance and structures dark data at ingestion so models and agents can act on reliable data.

In today’s AI era, your underlying data foundation is just as critical as the AI technology itself.”

What should data teams do?

Start with an inventory of repositories feeding priority AI workflows, then measure how much content remains disconnected and how much staff time classification consumes. Both need baseline figures before deployment expands.

Then test the platform against the financial institution’s own instruments, not the vendor’s most cooperative examples. Demand evidence for the claimed 96% and 99% extraction rates, define which cases require human review, and price the governance layer alongside the software. The research was commissioned by Iron Mountain, and the accuracy figures come from the company. Treat them as claims to validate, not a procurement shortcut.

Finally, connect content links before expanding model scope. In banking and insurance, AI deployment depends on the quality, classification, storage controls and governance of source information. Buying another model will not repair a filing cabinet.

Google News: Data Governance (Source)

Price tags

Share this article:

More from Grace Kim

Grace Kim

Show full bio

Market editor covering companies, deals and strategy in Analytics Digest is a professional daily at Analytics Digest.

18 articles

Read also

Next article »