Skip to content

Escaping the data trap: how the public sector can power the AI supercycle

5 November 20263 min read
Guest Insights
Digital data network visualisation representing public sector AI and data connectivity

Across central government, local councils, and NHS trusts, digital leaders attempting to shift from proof-of-concept AI initiatives to systemic delivery encounter an entrenched structural barrier: the primary bottleneck to scaling AI in the public sector is not model capability, performance, or algorithmic sophistication, it is what is termed ‘legacy data debt’, along with trapped capital and fragmented infrastructure. What this amounts to is that while advanced foundational models are increasingly accessible, public bodies remain constrained by decades of accumulated unstructured information, unmapped legacy databases, and isolated departmental silos that lack the interoperability, governance, and quality required for effective AI deployment. Addressing this challenge requires moving beyond isolated technical fixes to systematically remediate underlying data estates, release locked financial resources, and establish modern, compliant data architecture.

While UK digital strategy prioritises joined-up public services and secure data sharing, the operational reality on the ground often tells a different story. Up to a third of stored data is ROT (Redundant, Obsolete, or Trivial), while 52% is classified as dark data[1] which is unmapped, ungoverned, and carrying unquantified compliance risk. When departments rush to train or prompt models on these repositories, the outcomes are predictable:

86% of organisations have experienced project disruptions directly caused by poor data quality[2].

“We are drowning in disconnected data. When critical information is trapped inside isolated departmental silos, AI projects stall out entirely, leaving our organisation blind.” — Finance Leader, Public Sector

Aligning with UK direction: the "fund, feed, power" model

To align with central government missions including secure digital public infrastructure, AI transparency, and Greening Government Commitments, public bodies must embrace a four-part framework:

  • Fund Innovation via Legacy Remediation: Today, only 17% of organisations recycle savings from legacy IT back into AI. Millions remain locked inside outdated operations wasted on powering inefficient servers, paying premium storage for dark data, and maintaining physical archives of expired records. Through secure IT asset disposition (ITAD), hardware lifecycle reclamation, and systematic estate pruning, organisations recover residual value and cut recurring run costs. This frees immediate OpEx to self-fund AI compute and data modernisation without waiting for new Treasury allocations.

  • Feed Models with Governed, Traceable Data: Meeting the civil service mandate for transparent and trustworthy algorithms requires verifiable data lineage. Digitising vital physical records and purging ROT allows agencies to curate high-integrity datasets fit for ingestion into the National Data Library and cross-government platforms.

  • Power AI sustainably: Running live AI models will soon consume most enterprise computing power, creating a serious clash with UK net-zero targets. To scale responsibly, organisations need energy-efficient data centres and a cleaner data estate. Storing less junk data means using less power, making good data management an essential green practice.

  • Unify Commercial and Data Governance: In the public sector, siloed procurement frequently stalls technology adoption. CDOs, CIOs, and Accounting Officers must bridge the gap between compliance, commercial lifecycle planning, and data readiness. Establishing cross-functional oversight ensures that data assets and ITAD savings are treated as shared balance-sheet resources rather than departmental afterthoughts.

Building a resilient, trustworthy public service

For the UK government to truly benefit from AI, it has to get the basics right. By cleaning up redundant data, responsibly retiring old hardware, and keeping decisions transparent, public bodies can build AI tools that people actually trust and that deliver real value to citizens.

Read the complete findings and framework in the Iron Mountain Information Supercycle Report.

[1] Simon Jelley, “How to keep your data from ROT-ing in the cloud,” Forbes Business Council, Forbes, January 12, 2023.

[2] Iron Mountain, “The information supercycle: Aligning data, infrastructure, and capital for AI scale,” October 2026.