Executive Summary

Across central government, local government and arm’s-length bodies, data leaders are converging on a shared conclusion: the era of data strategy as an end in itself is over. Chief Data Officers, Heads of Data Platforms and transformation leaders are no longer being asked simply to define ambitious data visions, they are being asked to demonstrate that years of platform investment can now support AI at scale, in ways that improve services, reduce cost and build public trust.

This shift is the clearest signal to emerge from recent conversations with senior public sector data and digital leaders, and it has significant implications for how government organisations should prioritise investment over the next 12 to 24 months. This paper sets out four themes shaping government data strategy today, examines the legacy modernisation bottleneck that sits behind much of the AI readiness gap, and looks at how AI-powered migration tooling including AWS Transform is changing the timeline for closing it.

The Changing Shape of Government Data Strategy

For much of the last decade, government data strategy has focused on foundational questions, how to consolidate legacy systems, how to build central platforms, how to establish common standards. Those questions have not gone away, but they are increasingly being asked in service of a different goal. Organisations are no longer building data platforms for their own sake, they are trying to establish the conditions under which AI can be adopted safely, at scale, and with measurable benefit to citizens and public services.

This is a meaningful shift in emphasis. It means data quality, architecture and governance are being judged less against internal engineering benchmarks and more against a single practical test. Can this organisation trust its data enough to put AI in front of real decisions and real services? For many public sector bodies, the honest answer today is 'not yet', and closing that gap has become the primary task facing data leadership.

AI Readiness Has Become the New Data Strategy

Across departments and arm’s-length bodies, a consistent pattern is emerging: AI ambition is running ahead of data readiness. Leaders describe a familiar set of constraints holding back progress data quality that has not been addressed at source, architectures that were never designed to support real-time or federated access, governance frameworks built for a slower, more manual world, and a persistent gap in the skills needed to operate modern data and AI capability day to day.

Importantly, the conversation has moved beyond pilots and proofs of concept. Government organisations are now asking how to operationalise AI safely and repeatably within environments that carry real regulatory, security and public accountability obligations. That is a materially harder problem than running a single experiment, and it requires data foundations, governance models and delivery capability to mature together rather than in sequence.

Closing the Legacy Gap: Where AWS Transform Fits

Much of the AI readiness gap described above sits in a specific, familiar place: the legacy estate. Government organisations continue to run significant workloads on infrastructure that predates cloud-native design entirely. VMware-virtualised data centres, mainframe systems handling core transactional processing, and Windows-based applications tied to on-premises SQL Server estates. These are not simply old systems; they are systems that were never built to support the real-time data access, integration and governance that AI initiatives now require.

Historically, modernising this estate has been slow, resource-intensive and high-risk multi-year undertaking that competes for the same skilled engineering capacity needed to keep day-to-day services running. This is one of the main practical reasons AI ambition and data readiness have diverged; organisations often know what needs to change, but lack the delivery capacity to change it fast enough.

AWS Transform, AWS's agentic AI service for large-scale migration and modernisation, is materially changing that equation. Rather than a single tool, it operates as a set of purpose-built AI agents that handle discovery, assessment, wave planning, code transformation and validation across four areas of the legacy estate:

·        VMware migration to Amazon EC2, removing dependency on third-party virtualisation licensing.

·        Mainframe modernisation, supporting refactor, reimagine or replat form patterns for core transactional systems.

·        Full-stack Windows modernisation, covering .NET, the UI layer, SQL Server and deployment together rather than as separate projects.

·        Custom transformation of code, APIs and frameworks that fall outside a pre-built migration pattern.

The scale of what this changes is significant. AWS reports that the same agentic approach has been used internally to migrate more than 10,000 applications, saving an estimated 4,500 developer-years and$260 million annually. Across its early customer base, AWS Transform has analysed over a billion lines of mainframe code and saved more than 810,000 hours of manual migration effort. For full-stack Windows modernisation specifically, AWS reports modernisation completing up to five times faster, with operating costs reduced by up to 70%.

For government organisations, the practical implication is that legacy modernisation no longer needs to be treated as a multi-year prerequisite that delays AI adoption. It can run as a parallel, AI-accelerated workstream compressing the gap between “our architecture isn't AI-ready” and “our architecture is AI-ready” from years to months, and freeing scarce internal engineering capacity to focus on the governance, integration and skills work that still requires human judgement. Modern, cloud-native architectures produced through this kind of modernisation are also inherently easier to integrate and federate than siloed on-premises estates directly supporting the interoperability challenge explored below.

From Technology Investment to Demonstrable Outcomes

There is growing pressure on senior data leaders to show what earlier investment has actually delivered. Rather than launching new technology initiatives, many are being asked to evidence service improvement, productivity gains, better decision-making and improved citizen outcomes from the platforms and programmes already in place.

This changes what “success” looks like for a data programme. A well-architected platform is no longer sufficient on its own, it needs to be tied explicitly to an outcome a minister, permanent secretary or director general can recognise and defend. Organisations that can draw as straight line from data investment to a measurable result are in a much stronger position both to secure further funding and to build the internal confidence needed to expand AI use responsibly.

Data Sharing and Interoperability Remain Unresolved

Despite sustained investment over many years,sharing data effectively across departmental and organisational boundaries remains one of the hardest unsolved problems in government. Leaders continue to seek practical routes to integration, federated data models, cross-government collaboration, and common standards that actually get adopted rather than simply published.

This is not a new problem, but it is taking on new urgency. AI use cases frequently depend on data that sits outside any single organisation’s boundary: citizen records held by one department, service data held by another, and reference data held by a third. Without workable interoperability, AI initiatives stall at the same integration barriers that have constrained data programmes for years. Solving this pragmatically, rather than waiting for a perfect common standard, is increasingly seen as a precondition for progress.

Governance That Enables Rather Than Restricts

Perhaps the most consistent theme was the search for a better balance between governance and innovation. Public sector leaders are clear that governance cannot relax the requirements around trust, compliance and public accountability; they are not going away, and nor should they. But there is equally clear frustration where governance models create friction that is disproportionate to the risk being managed, slowing legitimate innovation without a corresponding improvement in safety or trust.

The organisations making the fastest progress are those redesigning governance as an enabler: frameworks that support AI initiatives by design, that maintain compliance without manual bottlenecks, and that give teams faster, well-controlled access to the data they need. This is less about loosening control and more about making control intelligently built into data platforms and workflows rather than layered on top of them as a separate, manual process.

What This Means for Government Organisations

Taken together, these themes point to a consistent set of priorities for organisations preparing for AI at scale:

·        Treat data quality and architecture as AI infrastructure, not back-office housekeeping. The two are now inseparable.

·        Use AI-powered migration and modernisation tooling, such as AWS Transform, to compress legacy modernisation timelines rather than treating them as a multi-year blocker to AI adoption.

·        Anchor every data initiative to a specific, measurable outcome that matters to service users and senior sponsors, not just to the technical estate.

·        Invest in pragmatic interoperability now, rather than waiting for a definitive cross-government standard that may never arrive.

·        Redesign governance so it is built into platforms and workflows, rather than imposed afterwards as a manual control layer.

·        Build internal capability deliberately, recognising that skills and operating-model change take longer to embed than the technology itself.

How Aker Helps

This is precisely the ground on which Aker works with public sector organisations. We help government bodies bridge the gap between data strategy and delivery, building the data foundations, real-time architectures and governance models that make AI adoption safe and practical, not just theoretically possible. Our experience spans complex, multi-stakeholder public sector ecosystems, giving us a practical understanding of the interoperability and governance challenges that are specific to government rather than generic to industry.

As part of this, we work alongside AWS Transform and the wider AWS migration and modernisation ecosystem to accelerate the legacy modernisation work that AI readiness depends on, combining agentic AI-powered migration with the governance, integration and change-management expertise that public sector programmes require. Automation compresses the mechanical work of code analysis, wave planning and transformation; our role is to pair that speed with the human oversight, assurance and stakeholder management needed to deliver it safely in a government context.

Rather than positioning ourselves as a technology vendor, we work as a delivery partner, helping organisations move from a well-articulated data strategy to a running, governed, outcome-generating capability and to do so in a way that stands up to the scrutiny that public sector programmes rightly attract.

Conclusion

Government data strategy is entering a new phase. The organisations that succeed over the next few years will not be those with the most ambitious strategy documents but those that can demonstrate real outcomes, share data safely across boundaries, and build governance that moves at the speed AI now demands.

The emergence of agentic AI-powered migration tooling has transformed legacy modernisation from a multi-year aspiration into an immediate priority for data and digital leaders across government. Rather than focusing solely on replacing legacy systems, organisations now have an opportunity to accelerate the integration and connectivity of services across government through AI-enabled modernisation. The challenge is no longer defining a compelling strategic vision but demonstrating the ability to deliver at the pace and scale that vision demands. Where strong digital foundations already exist, the priority should shift towards accelerating delivery, ensuring those foundation scan support AI adoption at enterprise scale. Ultimately, organisations that approach this as a delivery challenge, not just a strategic one, will be best positioned to translate AI ambition into measurable public value.