Bridging the AI Pilot-to-Scale Gap: How the Public Sector is Turning Ambition into Action

The UK public sector’s biggest AI challenge isn’t enthusiasm – it’s execution. As Liz Cook, director for central government at ServiceNow UK, told public sector attendees at ServiceNow World Forum in London on Wednesday: “The pilot-to-scale gap seems to be one of the biggest difficulties facing the UK public sector.”

The government’s AI Opportunity Action Plan lays out four core goals: build AI-ready data and infrastructure, scale trusted pilots across services, grow skills and embed governance, and deliver citizen-centric innovation.

Yet, Cook noted that “while every department recognises the opportunity for AI, the pace of delivery is not matching the government’s ambition.”

At the event, leaders from Oxford University Hospitals NHS Trust (OUH) and the Department for Environment, Food and Rural Affairs (Defra) explained how they are moving from concept to capability, which includes confronting legacy systems, data silos, and skills gaps to make AI real.

Oxford University Hospitals: Building Secure Data Foundations

For OUH, the journey to scalable AI begins with data quality and accessibility.

Lee Massie, the Trust’s head of IT, explained: “We have so much opportunity to leverage the data that we hold for our patients and our populations. The challenge we had is that data was distributed across siloed systems… data quality is poorer because those systems weren’t implemented for data leveraging.”

OUH is now one of ten NHS organisations developing a Secure Data Environment (SDE) – a protected ecosystem that consolidates data for safe analysis and AI development.

“We’re bringing together about 30 different clinical systems across the region,” said Massie. “That equates to about 11 million patients’ worth of data – 40 million rows in total. That’s 2.6 million floppy disks stacked to the height of Mount Everest.”

The goal, he said, is to make data “available ethically to researchers and organisations” while engaging with “priority and underrepresented groups” to ensure fairness and trust.

With the NHS now aspiring to become “the best AI-enabled healthcare service in the world,” OUH’s SDE lays the groundwork for scaling innovation safely. But Massie was clear that technology alone won’t deliver transformation.

“Our ability to create information exceeds our ability to process it,” he said. “We need these technologies to support individuals – to take workload away from people, giving them the opportunity to do the higher, smarter things.”

He also stressed that building literacy across the workforce is crucial. “We still have colleagues who say, ‘isn’t using AI cheating?’” said Massie. “So that conversation about what it’s for – and how it complements your job – is vital.”

Data, Ethics and AI for Environmental Outcomes at Defra

At Defra, the focus is equally on data and people. Tim Wheelhouse, deputy director for HR digital & automation, described AI as “fundamental to Defra’s outcomes and challenges – between water, access to nature, and the circular economy.”

Defra is investing in common data standards and “reusable technology” to make sure AI is applied consistently and transparently across the department and its agencies. Wheelhouse noted that around 72 critical national systems were recently reviewed, and only 21 are AI-ready. To change that, he said, “we’re identifying our critical data assets, our shared data assets, and making sure we can share that across government.”

One major initiative is Defra’s Crop Map of England (CROME) geospatial project. The dataset contains approximately 32 million hexagonal cells that is now becoming open source. The aim, Wheelhouse said, is to enable the public and other agencies to “see our natural environment as an asset that belongs to the UK.”

Like OUH, Defra is tackling the human side of AI adoption through the Civil Service’s “AI for All” campaign, which is designed to get “470,000 people playing with AI,” said Wheelhouse.

“If we don’t have people at the heart of this, any form of change will fail. We’re building communities that are using AI around different use cases, and building that AI literacy and innovative culture.”

Defra is also leading on ethical and environmental responsibility in AI. “Trust isn’t just about ethics – it’s about transparency and openness,” said Wheelhouse. “From an energy and carbon perspective, we are being really responsible and transparent about that use.”

From Pilots to Scalable Value

Both spokespeople agreed that moving AI from pilot to scale requires clear evidence of value and strong governance. “Sometimes it’s hard to measure through randomised trials,” said Wheelhouse. “We have to hold on to tangible outcomes around productivity, patient outcomes and citizen outcomes.”

Massie added that success goes beyond financial returns. “It’s not just about time saved – it’s about better well-being, higher retention, and return on employee. We need AI to help us understand all those data points.”

For Cook, these examples reflect the promise of AI done right: where trusted data, skilled people, and ethical governance align to deliver meaningful outcomes. As she put it, “AI is no longer a conversation about potential – it’s about delivery. And that delivery must be at scale.”