Strengthening AI Foundations in the Public Sector: an Ada Lovelace Institute perspective

The Ada Lovelace Institute’s recent blog post, “Building blocks: four recommendations to strengthen the foundations for AI in the public sector” (May 2025) (here), offers a valuable framework for thinking about how AI can serve, rather than simply streamline, public services. As the government accelerates its digital transformation agenda, the Institute outlines both the motivations behind this shift and the structural supports needed to ensure AI delivers genuine public value.

According to the blog, the government’s vision is ambitious: “no person’s substantive time should be spent on a task where digital or AI can do it better, quicker and to the same high quality and standard.” This statement, drawn from the Civil Service’s reform plan, encapsulates a broader push to modernise the state through automation and data-driven technologies.

Drivers

The blog identifies three key drivers behind this momentum:

1. Improving user experience - the aim is to streamline interactions with government services, many of which could be made “more functional, connected and better designed”.
2. Boosting productivity and efficiency - with significant budgetary pressures, AI is seen as a tool to reduce costs. The Department for Science, Innovation and Technology (DSIT) estimates that up to £36 billion could be saved through automation and simplification.
3. Attracting investment - through a conscious decision to step back from regulation of AI, UK is positioning itself as a global leader in AI by opening up its public sector data. The government has described the current time as a “narrow window [of opportunity] to secure a stake in the future of AI.”

Caution

However, the blog cautions that the pace of deployment must be matched by investment in foundational infrastructure. Drawing on six years of research into the social and ethical implications of public sector data and AI, the blog identifies four key lessons for the deployment of AI to be successful:

1. Contextualise AI – the blog emphasises that a lack of clear terminology, inconsistent underlying data and the context in which AI systems are deployed all impact its effectiveness.
2. Learn what works - currently, the public sector lacks a comprehensive view of where AI is being successfully (or unsuccessfully) deployed across government in order to learn the necessary lessons for future deployment.
3. Deliver on public expectations and public sector values - successful use of AI depends on public trust.  Public procurement of AI to be unfit for purpose and good governance is crucial for ensuring AI tools are safe, effective and fair.
4. Think beyond the technology - AI will have transformative societal consequences and could be an instrumental tool in re-imagining the operation of the public sector.

Recommendations

Building on these findings, the blog makes four key recommendations: 

1. Create a What Works Centre - Build a central hub to collect and share evidence across the government and the wider public sector on the AI tools which work best in certain environments.
2. Strengthen Transparency Standards - Improve and mandate the use of the Algorithmic Transparency Recording Standard (ATRS) across government.
3. Support Ethical Procurement - Set up a taskforce to help local authorities buy AI tools that align with public values and central government standards.
4. Invest in Public Engagement - To foster public trust in evolving AI tools, government should fund research and dialogue to understand and reflect public attitudes toward AI in policy and design.

In conclusion

• whilst AI adoption in public services is accelerating, there’s still no clear vision for how it should serve the public good. Current efforts often focus on efficiency, but the real challenge is using AI to deliver public value. 
• promising pilots do not begin with technology. Instead, they begin with service outcomes and involve users in shaping solutions, including in understanding where AI helps and where its use should be limited.
• designing from the end-user up helps clarify where technology should prioritise personalisation, precision, or productivity. It also surfaces the trade-offs between different stakeholders (government, professionals, and the public) that must be carefully balanced.

If you would like to discuss how current or future regulations impact what you do with AI, please contact Tom Whittaker, Brian Wong, Lucy Pegler, Martin Cook, Liz Smith or any other member in our Technology team.  For the latest on AI law and regulation, see our blog and newsletter.

This article was written by David Harrison.