The morning panel session on Day 2 in Citizen Experience at the DigiGov Expo 2025 brought together an impressive lineup of digital leaders to discuss AI's transformative potential in public services. Nadira Hussein (Chief Executive, Socitm) chaired the discussion with panellists David Johnson (Head of Innovation, HMRC), Ed Pikett (Chief Digital Officer, RedRock Consulting), Eilidh McLaughlin (Deputy Director for Digital Ethics, Inclusion and Assurance, Scottish Government), Emily Ball (Deputy Director for AI Adoption and Service Transformation, Government Digital Service), and Marta Pocztowska (Chief Digital Officer, Thurrock Council).
The session opened with a provocative challenge to conventional thinking. Rather than simply making services available "anytime, anywhere," one panellist suggested the real goal should be passive, proactive government engagement; where citizens never have to apply for passports, driving licences, or complete tax returns again. These processes should happen in the background, freeing citizens to focus on more meaningful activities. This sparked an immediate ethical reflection: have we asked the right question? Whilst transforming digital services is praiseworthy, are we addressing the needs of those at the hard cutting edge of society? For someone living in a 10-storey block of flats with a disabled mother and a broken lift for three weeks, the priority isn't digital tax returns; it's basic infrastructure. This challenge to consider whether AI is the answer to every problem set an important tone for the entire discussion.
The conversation then turned to practical applications already delivering results. HMRC has been using AI for several years through natural language processing and machine learning for digital assistants, correspondence routing, and contact data analysis. The organisation is now focused on generative AI for automated code translation, improving call centre operations, and enhancing internal processes, all of which translate directly into better public service. Scottish Government demonstrated transparency through their AI Register, where all public sector organisations must document their AI use publicly. Their FOI redaction tool addresses the 6,000 annual freedom of information requests, each taking approximately half a day to answer – a substantial time saving that prompted questions about whether proactive data publication could avoid FOI requests altogether.
Thurrock Council provided perhaps the most compelling evidence of AI's practical impact in adult social care. Tools including magic notes for meeting transcriptions automatically populate reports, documentation, and assessments, achieving a 47% reduction in administrative time and enabling a 22% reduction in agency staff; significant achievements given severe financial constraints and intervention status. The council also focuses on addressing barriers to digital engagement, including language barriers and accessible formats for those with conditions like ADHD or dyslexia.
The discussion on personalisation highlighted how AI can enhance human connection rather than replace it. In the context of Citizens Advice, where advisors typically serve three clients per day, AI provides greater context before meetings, past interactions, records, and specific situations. This is particularly valuable when people seek advice from positions of anxiety and stress. The goal isn't replacing human advisors but providing them with better information for more informed, empathetic support. Translation and transcription capabilities also make content more accessible, particularly important in diverse areas like Lewisham, which has more languages spoken than any other local authority.
GDS's AI Exemplar Programme illustrates the balance between efficiency and user experience. Some tools improve internal processes and shield users from organisational complexity, whilst others directly interact with citizens. There's particular interest in using AI to help less digitally confident users navigate complicated transactions and enable proactive, personalised interactions through better data joining. Discharge summaries in healthcare demonstrate benefits beyond simple time savings; they help clinicians discharge patients more quickly whilst improving the quality of output and community care provision.
The shift from reactive to preventative services offers perhaps the greatest opportunity for savings. By combining data from public health, hospitals, GPs, mental health referrals, hospital discharges, housing, and evictions, local authorities can identify at-risk residents. Preventative intervention costs significantly less than emergency housing, hospital discharges, or social care packages. However, this requires improving data quality and developing the capability to view data collectively. As one panellist emphasised, data isn't AI itself, but it's the beginning and local government has huge amounts of data that needs better leveraging.
Creating environments where people feel able to test, learn, and experiment emerged as crucial for AI adoption. At HMRC, rolling out Microsoft Copilot demonstrated how quickly staff become reliant on tools for summarising Teams calls and content creation – removing licences at trial end provoked significant outcry. This bottom-up adoption could prove powerful, as frontline staff understand where processes need improvement. The discussion acknowledged an interesting contradiction: surveys show concerns about using AI too quickly and worries about job security, yet feedback about actual tools is excellent and demand is high for faster rollout.
Several participants emphasised values-based approaches centred on people. As one speaker noted, quoting a former boss: "It's about people and stuff – sort the people out, the stuff will happen." This means genuinely hearing diverse voices, embedding civil service values at leadership level, and demonstrating those values in practice. The discussion highlighted the need for public conversation about AI in government services, bringing citizens into discussions about appropriate uses, concerns, and reassurance measures.
Importantly, compliance with existing legislation; particularly data protection and records management; cannot be overlooked. Simply completing a data protection impact assessment isn't sufficient; organisations must consider what records need retaining when using AI to summarise documents, ensuring public transparency and confidence. The importance of unglamorous but essential infrastructure also received attention. Common components like digital identity, secure communication channels, and structured data access aren't AI themselves, but they're fundamental for AI to work effectively and for building trust.
The session concluded with recognition of improving connections between local and central government after years of separation. The overarching message was clear: AI is a powerful tool to achieve longstanding goals of inclusive, joined-up, user-centred government services that reduce burden on citizens and businesses. Success requires maintaining focus on people and values, ensuring proper guardrails and transparency, investing in foundational infrastructure, and fostering innovation whilst managing legitimate concerns about pace and consequences. As the audience was reminded, effective public service transformation through AI isn't just about technology; it's about asking the right questions, addressing real citizen needs, and building trust through demonstrable commitment to transparency and ethical deployment.