Following a wide-ranging panel discussion at this week’s Think Digital Government, public sector leaders are being urged to focus less on the novelty of AI and more on whether it will truly improve the way citizens interact with government.
Speakers from government, regulators, and industry explored the areas where AI is already delivering value within government. The discussion also highlighted the growing recognition that successful AI implementation requires governments to invest in data quality, governance, and public trust.
Speakers warned that ultimately public trust in government AI systems will determine whether implementation is successful. Alex Jones, interim director of the AI Incubator (i.AI), argued that discussions about AI should start with understanding the role governments themselves play in society.
“Government is not just a service provider,” he says. “Government is not a business. Government is not a startup.”
Rather, he argued, government exists because some responsibilities, such as justice, education, and public safety, are “too important to be left to the private sector or the market alone.”
Jones suggested that public trust in AI depends on two factors. One is that the public believes that the government is competent enough to use technology responsibly, and the other is that the public believes that its incentives remain aligned with the public interest.
“The public needs to make sure that government incentives align with their own incentives when using AI,” he said.
Chris Gullick, chief data and AI officer at Ofgem, said his organization was approaching AI through the lens of a regulatory mandate rather than a technology experiment.
“We start with the regulator’s mandate: protection of energy consumers, energy security, and net zero,” he said.
For Garrick, the important question is not simply whether AI will make processes cheaper or faster, but whether it will improve societal outcomes.
“What do we use AI for to benefit society, not just make things as cheap and fast as possible?” he added.
Governments still struggle between experimentation and caution
The panel also explored the tension between fostering innovation and maintaining adequate safeguards around the use of AI. Matthew Hills (pictured), head of government and public sector for the UK and Ireland at ElevenLab, said trust was particularly important for technologies that interacted directly with the public.
“When trust is lost, so is the purchasing entity that purchased the solution,” he said.
If you like this content…
He argued that organizations often underestimate the difference between experimenting with AI tools internally and safely deploying them in production.
“When it comes to actual production, that’s where the problem comes in,” Hills says.
Meanwhile, Garrick described the challenge of balancing the enthusiasm of staff eager to use AI tools with the concerns of others who remain skeptical or apprehensive about the technology.
Ofgem is focused on supporting staff to become “confident experimenters” through controlled opportunities to safely test AI tools, he said.
One effort involved bringing together technical and non-technical staff to build a simple AI agent using Copilot Studio.
“This kind of activity brings something useful and gets people engaged and experimenting and excited about how it can help them in their daily work,” Garrick said.
AI should reduce friction, not replace human judgment
Speakers emphasized that AI should augment civil servants, rather than replace human responsibilities and decision-making. Jones pointed to existing government AI projects focused on reducing administrative burden, such as tools to reduce note-taking time for social workers.
The panel also discussed when AI may not be appropriate at all. Garrick said recruitment remains one sensitive area because of the risk of bias and poor decision-making.
“Recruitment is an area where I am completely against the use of AI,” he said.
However, he noted that AI could still play a useful supporting role, such as helping unsuccessful candidates receive better feedback on their resumes and applications.
Jones also suggested that there are many “human-first” processes in which governments should be wary of relying too heavily on AI, especially when systems remain inaccurate or decisions have a significant impact on individuals.
A consistent message throughout the discussion was that governments should not pursue AI because it is trendy or politically attractive, but to truly improve services and outcomes for their citizens.
