Based on an Infoquest Expert Voices interview with Paul Scully, Former UK Minister for Technology
Agentic AI in government services is no longer a theoretical horizon. It is arriving within the next 12 to 18 months, and most governments are not ready for it. That is the view of Paul Scully, who spent nine years as a UK Member of Parliament and served as Minister for Technology, overseeing the UK’s AI White Paper, semiconductor strategy, and the landmark AI Safety Summit at Bletchley Park.
Having navigated the intersection of policy, technology, and public trust at the highest levels, Scully offers a clear-eyed assessment of where governments are falling short and what they need to do now.
From Paper to Digital-First: How Government Technology Has Evolved
When Scully became a minister in 2020, civil servants were still printing out every document for him to sign by hand — sometimes dozens of pages in a single sitting. COVID changed that almost overnight, forcing a pivot to tablets and PDFs. It was not exactly a digital transformation, but it was a start.
The deeper problem, he argues, is that most governments have simply digitized existing paper processes rather than redesigning them from the ground up. When digital systems do not talk to each other, the benefits remain shallow. Scully draws a direct comparison to open banking, which fundamentally changed how financial services share data. That same logic, he says, needs to apply across every layer of government.
Government Digital Maturity: Legacy Constraints vs. Blank Canvas Advantage
How starting position shapes digital transformation speed and ambition
| Dimension | Legacy-Constrained Governments (e.g. UK, Japan) |
Blank Canvas Governments (e.g. GCC, emerging markets) |
|---|---|---|
| Starting Point | Decades of mismatched analog and digital systems that do not interconnect | Building on a near-clean slate with fewer legacy constraints |
| Public Trust in Digital Gov | Deep-rooted distrust of government ID and data-sharing programs | Higher cultural acceptance of unified government apps (e.g. Saudi Arabia, UAE) |
| Procurement Model | Reliance on 3–4 large IT corporations; limited SME innovation in government tech | Opportunity to design open, competitive procurement from the outset |
| Data Integration | Siloed systems — health, planning, welfare rarely share data across departments | Ability to design unified data architecture from day one |
| Digitization Approach | Often “digitizing paper” rather than rethinking processes as digital-first | Can design digital-first processes without replicating analog workflows |
| Entrepreneurial Culture | Risk-averse institutional culture; government failure seen as politically costly | Stronger appetite for risk-reward innovation, Vision-driven policy momentum |
| Agentic AI Readiness | Piloting underway (e.g. Scotland) but systemic integration is years away | Positioned to deploy agentic AI across connected systems with less friction |
The Three Primary Drivers of Public Sector Transformation
Scully identifies three core forces pushing governments toward digital transformation: efficiency, usability for citizens, and better planning. Each one is deeply connected to the others. The NHS in the UK holds what is arguably the largest single unified medical dataset in the world. Yet a patient referred from a GP to a hospital still has to explain their entire medical history from scratch, because the records do not follow them through the system.
Planning is equally broken. Developers in the UK spend tens of millions of pounds submitting tens of thousands of pages for major projects. AI-assisted systems could streamline that process dramatically. Scully points to digitally advanced governments across the Middle East and Far East, where major developments happen significantly faster. The link is not coincidental.
Why Trust Is the Real Barrier to Government Technology Adoption
Governments cannot move at startup speed. They cannot fail the way a business can. That tension sits at the heart of why trust is the defining challenge in public sector technology adoption. Citizens routinely share personal data with Facebook, Amazon, and X without hesitation, yet resist giving the same information to the government.
Scully points to Ukraine’s DIA app, a government-built identity platform used by over 80 percent of Ukrainians, as an example that simply would not work in the UK right now. The public trust has not been built. Alongside that, cybersecurity remains a serious vulnerability: the NHS suffered a major breach in part because it was running an outdated version of Windows. And a reliance on a small number of large IT corporations creates concentrated procurement costs and limits genuine innovation.
Bridging the Digital Divide With Social Tariffs and Post Office Networks
Upskilling the population is the long-term goal, but it cannot happen overnight. As an interim solution, the UK introduced social tariffs: subsidized broadband and mobile connections for lower-income households. Scully also sees untapped potential in the UK’s 11,500-strong post office network, a physical infrastructure already embedded in rural communities that could serve as an access point for citizens who lack connectivity or digital skills.
The broader principle is that government does not need to do everything itself. Scully frames it around three questions: what can government do directly, what should it get out of the way to let others do, and where should it simply convene and signpost? Organizations like Google Digital Garage already deliver digital skills training at scale. The role of government is to enable them, not duplicate them.
Why Agentic AI Will Be the Next Government Technology Wave
Of all the technologies on the horizon, Scully is clear: agentic AI in government services will be the defining shift of the next 12 to 24 months. The Scottish Government is already piloting AI systems to process public consultation responses, a task that previously took six months to a year. That is just the beginning.
Scully’s vision extends further: agentic AI working through entire legislative pipelines, moving a policy from public consultation through multiple stages of preparation with far less manual intervention. The speed gains are significant. But so is the change in what government work actually looks like day to day.
The Agentic AI Legislative Pipeline
How agentic AI can accelerate government decision-making from public consultation to policy
Government publishes consultation documents and invites responses from citizens, businesses, and civil society. In traditional settings, this generates thousands of written submissions over weeks or months.
Agentic AI ingests and synthesizes thousands of consultation responses — a process that traditionally takes 6 to 12 months of civil servant time. The Scottish Government is already piloting this approach. AI identifies themes, dissenting views, geographic patterns, and key stakeholder concerns at speed.
Agentic AI assists in drafting policy frameworks and initial regulatory language, drawing on the analyzed consultation data. Human ministers and policy advisors review and shape outputs. Ethics panels — including ethicists, philosophers, and sector specialists — provide oversight at this stage.
Refined policy outputs are prepared for formal legislative process — bill drafting, parliamentary review, and public scrutiny. AI tools reduce the time between consultation and bill-ready text. Human accountability and parliamentary oversight remain non-negotiable at this stage.
Policy moves into implementation. Agentic AI monitors outcomes, tracks service delivery metrics, and surfaces data for the next consultation cycle — creating a self-reinforcing loop of evidence-based governance. The human element remains paramount throughout.
Ethics, the Online Safety Bill, and Building AI Guardrails
Scully’s nine years working on the Online Safety Bill offer a critical lesson for AI governance. Early drafts risked restricting adult free speech online; it took a significant rethink to refocus the legislation on child protection. The lesson he draws is direct: governments should have been thinking about social media safety 15 years earlier, in its infancy. With AI, there is still a window to get ahead.
His approach to AI regulation focuses on training existing sector-specific regulators to evaluate AI’s outcomes rather than regulating the technology itself. Equally important, he argues for bringing ethicists, philosophers, and theologians into advisory panels alongside data scientists. Tunnel-vision problem-solving, he warns, is exactly how technology goes wrong at scale.
The Human Element: Advice for GCC Leaders and Digital Transformation Consultants
Scully’s closing message is simple: never forget the humans. Technology is not a process for process’s sake. When governments get digital transformation wrong, real people bear the cost. When they get it right, it is communities, families, and neighbors who benefit. That human calculus should sit at the center of every decision.
For leaders in the GCC specifically, Scully sees a genuine structural advantage. Starting with fewer legacy systems and without the accumulated weight of decades of mismatched digital infrastructure, the Gulf states can build in a way that older economies simply cannot. The entrepreneurial culture of the region, its appetite for risk and reward, is precisely the mindset that ambitious digital transformation requires. The blank canvas, he says, is a gift. Use it.
