The power of AI in the public sector

by Andy Thomas - Principal Architect

Jun 18, 2025 | minute read

AI is everywhere in public sector conversations. But while the pilot projects keep piling up, few are turning into scaled operational use - embedding AI into day-to-day decision-making, powering frontline services, and automating complex back-office tasks.

Initiatives are kicking off well. But making the hard yards, the structured work of gaining sponsorship, securing funding, testing at scale, and managing acceptance across stakeholders is what really makes the difference.

The time for isolated pilots is over. We need scalable, reusable platforms that deliver across departments and citizen needs.

The power of AI in the public sector isn’t theoretical - it’s already here. But its impact depends on bold, collaborative action.

It’s a bit like rugby. You don’t win by scoring off every flashy play. You make ground through phases, one play at a time. Controlled execution.

AI rollout shares some of that rhythm. No more big bang transitions. Rather than long delivery cycles, AI demands small strides that demonstrate early value. And, just like in rugby, not all pitch gains are forwards. Sometimes you need to give up a bit of ground to maintain control and set up the next move. It’s all about keeping momentum towards the final try.

This is where government and industry can work side by side, demonstrating how co-design can deliver real-world benefits at scale. Sharing insight, testing ideas and building momentum together. It’s not about one side leading the other, but about collaboration that embraces the incredible pace of technological change and turns early promise into meaningful impact.

Why industry / government collaboration is crucial

In a fast-moving AI landscape, collaboration between government and industry isn’t just helpful, it’s essential.

Public sector teams know their policy context, citizen needs and service pressures better than anyone. But delivery at pace, particularly when it involves new AI tools and capabilities, requires tested frameworks, scalable platforms and real-world delivery expertise. That’s where industry can contribute. Industry brings its learning from the freedom to innovate, to try and fail, and to iterate rapidly. Lessons that are vital as government explores how to embed AI safely and at scale.

From secure, compliant platforms like Microsoft, AWS, ServiceNow and Salesforce to sector-wide experience in deploying ethical, explainable AI, industry brings both the tools and the track record. This includes helping government cut through challenges like hallucinations, bias and drift by building in strong governance from the outset.

And importantly, the most effective outcomes come from co-design, not handover. Sitting side by side, shaping delivery and sharing accountability. In complex public sector domains, this kind of partnership builds trust, accelerates outcomes, and ensures AI is used responsibly.

To make this sustainable, we need more than just good collaboration. We need the right structures. That means joint benefit models: frameworks that incentivise ongoing teamwork rather than one-off project delivery. Think shared risk and reward, outcome-based contracts, and delivery arrangements that evolve over time. These aren’t just commercial models, they’re cultural ones. They reinforce long-term partnership and support a product mindset, where services grow, adapt and improve continuously.

When done well, this blend of public purpose and private sector capability becomes a powerful enabler. It's how we navigate the complexity and move at pace, not with endless pilots, but with delivery that sticks.

Overcoming the barriers to AI adoption

If we accept that pilots are no longer the problem, and that most departments are already experimenting with AI in some form, then the real conversation is about scale.

The friction starts when trying to connect a promising prototype to the complexity of live government operations. Procurement pathways weren’t built for iterative, exploratory technologies. Legacy systems aren’t easily adapted to new models of intelligence. And teams often lack the time, capacity, or confidence to move forward at speed.

Then there’s the cloud question. The UK’s cloud-first policy signals clear intent, but the protection afforded to departments and arms-length bodies (ALBs) often doesn’t go far enough. Directors, Deputy Directors, CIOs and delivery teams remain exposed and under pressure to adopt modern platforms but without the confidence that their decisions will be backed if things go wrong. In areas like data sovereignty, cross-border storage or supplier compliance, too much risk is pushed down the chain.

A government-wide approach to platform accreditation can create a shared foundation of trusted, approved environments. When departments know the technology they’re using is centrally backed and futureproof, they’re far more likely to adopt it at scale. This is about creating safety, not just standards.

The recent merger of the Government Digital Service (GDS) and Central Digital and Data Office (CDDO) under the Department for Science, Innovation and Technology (DSIT) could mark a pivotal moment for progress. It’s a prime opportunity to rethink how government works with technology platforms. Both have established enterprise solutions and emerging tools via innovation agreements and sandbox frameworks.

If we want departments and ALBs to move confidently, we must give them the freedom to adopt trusted platforms without fear of recourse. That means backing their decisions with centrally endorsed environments and opening the door to new tech that meets the right standards, even if it’s not yet widely adopted. This isn’t about a lack of ambition. It’s about systemic blockers - some technical, some cultural, many policy-driven.

We need more than just platforms. We need shared ways of working, consistent guidance from the centre, and space to build delivery confidence. That’s how we move from early-stage experimentation to operational impact.

"Our assets are screaming data at us, but we’re not listening."

AI in the public sector is often framed around contact centres, chatbots and service desk automation - and rightly so. These are visible, citizen-facing areas where AI can make a real difference. But behind the scenes, in the world of Operational Technology (OT), lies an even bigger opportunity.

Critical National Infrastructure is everywhere we look. Rail, road and air networks. National grids for electricity, water and gas. Operated by national departments, local government and private sector alike. All of these networks rely on a vast array of systems and sensors.

With the rapid evolution of tools such as Agentic AI (systems capable of acting autonomously toward goals) and Data Fabric, coupled with more established technologies like Robotic Process Automation (RPA), we can now harness this data to tackle some of the most visible and urgent challenges in public infrastructure.

Imagine a world where water companies can pinpoint underground leaks before they become bursts, using existing flow and pressure sensor data to track abnormal loss and dispatch crews faster. In 2023–24, over a trillion litres of water were lost to leaks across England and Wales. Even a modest 10% reduction could save 100 billion litres annually.

Or imagine less traffic congestion. Not through costly new roads, but by using AI to optimise existing traffic signal timing. In Manchester, AI-powered traffic light systems have cut travel times by up to 30%. In Hull, AI helps manage congestion and air quality in real time.

What about rail companies adjusting capacity based on demand, freeing up freight paths to reduce lorry movements. Shifting more freight to rail could reduce CO₂ emissions by up to 76% per tonne moved.

These opportunities don’t require futuristic tech. They just need better use of the systems and data we already have.

But OT isn’t like IT. It’s risk-averse, highly specialised, and often disconnected from the rest of government’s digital estate. That’s why it’s critical that AI adoption here is done with sector expertise, strong governance, and tight integration with operational decision-making.

Create confidence through pre-approved platforms

This isn’t about creating separate routes. It’s about giving teams a solid starting point with trusted platforms, while still keeping the door open for fresh ideas and new tech to prove their worth.

What we’re really talking about is a government-wide approach to trusted platforms. A set of approved environments that are secure, compliant and scalable, paired with a more agile mechanism for assuring emerging tools. Platforms that delivery teams can trust, and that central government stands behind.

These foundations reduce friction. They avoid repeated evaluation and procurement for the same capability. And they send a powerful signal: this is a place where AI can happen safely.

At the same time, frameworks like innovation agreements, sandboxes, or test-and-learn procurement routes can give emerging tech a path into government without asking departments to take on disproportionate risk.

This model isn’t about mandating solutions from the centre. Local leaders know their challenges best. But with a clear platform strategy and a culture of safe exploration, we create the conditions where good ideas scale quickly and AI becomes a normal part of digital delivery.

From promise to performance

The time for isolated pilots is over. We need scalable, reusable platforms that deliver across departments and citizen needs.

The power of AI in the public sector isn’t theoretical. It’s already here. But its impact depends on bold, collaborative action.

Now is the time to act. Whether you're shaping strategy, leading delivery or building platforms, there's a role to play in turning AI's potential into public value.