AI ambitions collide with government’s data reality

The race to adopt AI across government and national security organisations is accelerating rapidly, but many departments are still struggling with the fundamentals needed to make those ambitions operational.

That’s the message from a new paper by Aker Systems, which warns that many public sector AI programmes risk stalling unless organisations first modernise the data infrastructure, governance and operational practices sitting beneath them.

Written by Stephen Dowdeswell, technical pre-sales solution architect at Aker Systems, the paper focuses on government, defence and national security environments, where data quality and operational resilience carry significantly higher stakes than in many commercial deployments.

According to the report, only 14 percent of organisations currently have a fully AI-ready data platform, while around half cite legacy systems and scalability challenges as major barriers to progress.

The paper argues that many organisations are still focused too heavily on AI applications themselves, while underestimating the importance of the underlying infrastructure needed to support them reliably and securely.

Dowdeswell warns that weak data foundations create significant operational risks: “In high-stakes environments, the impact of poor data goes far beyond cost. It affects operational effectiveness and strategic outcomes,” he writes.

Legacy systems remain a major barrier

The paper highlights the challenges many government organisations face in trying to modernise AI capabilities while continuing to operate fragmented legacy estates.

Many public sector systems were originally designed for transactional processing, record management or static reporting rather than real-time AI-driven decision-making. As a result, organisations often struggle with siloed data, inconsistent standards and infrastructure that cannot easily support modern AI workloads.

The paper argues that cloud migration alone is not enough to solve those issues.

Instead, Dowdeswell says organisations need integrated and interoperable data environments capable of supporting secure, scalable and continuously updated data pipelines. That includes improving governance frameworks, standardisation, metadata management and cross-organisational data sharing practices.

The report also stresses that trust in data is becoming increasingly important as AI systems move closer to operational decision-making.

For government and national security organisations, poorly governed or unreliable data could undermine the effectiveness of AI systems rather than improve it.

AI adoption moving beyond experimentation

A key theme running through the paper is that many organisations remain stuck in an “experimentation phase” with AI. Pilot projects and proofs of concept continue to emerge across public services, defence and security operations, but scaling those initiatives into operational systems remains difficult.

Dowdeswell argues that organisations now need to move beyond isolated experimentation and focus instead on building sustainable operational capability. That requires long-term investment in foundational disciplines such as secure infrastructure, governance, interoperability and operational resilience.

The paper also argues that AI readiness should be viewed as an organisational capability rather than simply a technology deployment exercise.

“Being AI-ready is not a destination,” writes Dowdeswell. “It’s an operational capability.”

According to the report, organisations that fail to address the underlying quality and governance of their data environments may struggle to operationalise AI systems effectively, regardless of how advanced the models themselves become.

National security implications

The paper highlights defence and national security environments, where the consequences of poor data governance or infrastructure weaknesses may be especially severe. In those environments, organisations often need to manage highly sensitive information across multiple classifications, systems and operational teams.

Dowdeswell argues that AI-ready infrastructure therefore requires more than centralised storage or analytics capability alone.

Instead, organisations need secure-by-design architectures, stronger data lineage and governance models capable of balancing operational agility with security and compliance requirements.

The report suggests that many organisations are still underestimating the complexity involved in building those environments.

“Without modern, scalable and secure data infrastructure. AI initiatives risk becoming isolated experiments rather than operational assets,” writes Dowdeswell.

That warning comes as governments globally continue to position AI as central to economic growth, productivity and public service reform.

But the paper argues that AI transformation cannot be separated from wider infrastructure modernisation efforts.

Many public sector organisations are simultaneously managing legacy technology debt, rising cyber security pressures and large-scale digital transformation programmes while also attempting to introduce generative AI capabilities.

The report suggests that without stronger data foundations, those ambitions may become increasingly difficult to sustain.

Infrastructure before innovation

Ultimately, the paper argues that successful AI adoption depends less on the AI models themselves and more on whether organisations can build trusted, resilient and scalable data ecosystems around them.

That requires government organisations to treat data as strategic infrastructure rather than simply a by-product of operations.

Dowdeswell says organisations must focus on creating environments where data can move securely, consistently and reliably across systems while maintaining trust, compliance and operational resilience.

“AI is only as good as the data and infrastructure that support it,” he writes.