The role of AI in HM Government’s innovation agenda

Pictured (left to right): Stephanie Calhoun, Lead Data Scientist, Opportunity Mission at 10 Downing Street; Shehroze Junejo, Executive Director & Chief Data Officer at UK Government Investments; Simon King, Director of AI & Innovation at the Department for Work and Pensions; and panel chair Jennifer Anderson, Chief Information Officer at the National Wealth Fund.

As Artificial Intelligence transforms workplaces across the public sector, government leaders and digital experts are navigating the opportunities, challenges, and cultural shifts required to embed AI successfully.

At a recent panel discussion, chaired by Jennifer Anderson, Chief Information Officer at the National Wealth Fund, she and other government leaders emphasised that AI must empower people, streamline decision-making, and foster trust - all while ensuring transparency and human oversight.

Making Government smarter

Shehroze Junejo, Executive Director & Chief Data Officer at UK Government Investments (UKGI), an arm’s length body of HM Treasury, highlighted how AI is revolutionising the use of Digital, Data, and Technology (DDaT) across the public sector.

“The AI boom has brought DDaT to entirely new populations of users,” he said. AI’s real potential, he argued, lies in answering the key question: What can it and can’t it do? “It opens us up to more difficult conversations - rubbish in, rubbish out. AI is only as good as the data.”

At UKGI, AI-driven tools are already generating quick wins. For example, Notebook LLM can transform documents into podcasts, fundamentally changing how teams consume information.

“We’ve been streamlining assessments of risks such as contingent liabilities. Using AI gives ministers data points that didn’t exist before to inform decisions,” Junejo explained. However, he warned of the need for a cultural shift: “We must be inquisitive, inclusive, and understand what was the black box of AI - or risk falling behind global leaders like Hong Kong.”

From concept to service

Stephanie Calhoun, Lead Data Scientist, Opportunity Mission at 10 Downing Street, described how the Incubator for AI is focused on identifying practical use cases to deliver meaningful improvements. 

“We’re a delivery team with 11 products in development,” she said, sharing examples such as Consult, an AI tool designed to overhaul the government consultation process. “Government runs hundreds of consultations annually, costing over £80m and significant staff time. Consult makes the process of analysing public responses much quicker and cheaper.”

The team’s Redbox project, born from a hackathon, enables civil servants to search official documents using a ChatGPT-like interface. “Anyone in Cabinet Office can sign up to access Redbox - it’s already helping private offices work faster and smarter,” Calhoun added.

All of this is open source and available on the Incubator for AI’s own Github page.

Private offices are small, dedicated teams that support ministers and senior civil servants in managing their official responsibilities. They coordinate diary management, briefings, correspondence, and communication between the minister, their department, and external stakeholders, ensuring the smooth flow of information and decisions.

Scaling AI responsibly

Simon King, Director of AI & Innovation at the Department for Work and Pensions (DWP), shared his department’s approach: safe acceleration.

“We’ve been developing AI tools for years, but a responsible AI framework is critical,” he said. “We don’t have technology projects; we have people projects. The key is ensuring AI has the right guardrails - humans must remain in the loop.”

King gave examples of how DWP is already using AI:

• DWP Ask: Work coaches use AI with a natural language interface to navigate 70,000 pages of information, helping people access benefits quickly and accurately.
• Letter triaging: Large language models scan 23,000 physical letters monthly to identify urgent cases requiring escalation. “This system has been operational for six months and saves considerable time,” King explained.

However, scaling AI projects remains a challenge. 

“It’s easy to spin up a proof of concept, but scaling it is much harder. Someone said to me that they’ve got more pilots than British Airways!” he quipped. For successful AI adoption, King emphasised transparency and trust. “Citizens are concerned about how AI will be used. We must lean into openness, publish algorithms, and ensure AI projects are treated differently to mega-projects. Funding models must match the size and scope.”

Building trust in AI

Jennifer Anderson, CIO at the National Wealth Fund, the UK's sovereign fund for strategic investment headquartered in Leeds, highlighted a key barrier to adoption: anxiety around AI.

“Data is the ultimate junk in, junk out. There’s also the question of access - ensuring people don’t see data they shouldn’t.” She cited a survey by the National Audit Office showing that 70% of government organisations are either piloting AI or about to start. “The two main questions I hear are: Is my data safe? and Will AI replace my job?”

Calhoun agreed that anxieties stem from AI’s “black box” perception. “Having humans in the loop and focusing on transparency allows users to interrogate AI outputs and build trust.”

Junejo added that accountability is crucial: “AI empowers you, but decision-making remains yours. If someone thinks they could have done the task themselves, then you’ve implemented the AI well.”

King stressed the importance of energy efficiency and using AI appropriately. “Tools like ChatGPT use 30 times more electricity than a Google search. We need to focus on smaller, fit-for-purpose models and platform efficiency. How much risk are we willing to take?”

Junejo concluded by calling for curiosity and collaboration: “The challenge with AI is the same as any technology - we often make it sound too complicated. Have lots of conversations, listen to your users, and focus on how you reach decisions.”

When asked whether humans always add value, King reflected on the broader purpose of AI: “It’s not about replacing people - it’s about enabling them. The technology is only as powerful as the people using it.”

Acknowledgements: We would like to thank the HM Treasury team for inviting Government Transformation Magazine along to be a part of their Innovation Day 24.