Semi-finalists named in Civil Service AI & Data Challenge
Eight civil servants have been invited to develop standout ideas for using AI and data to improve public services. The shortlist was selected from more than 250 submissions to the Civil Service AI & Data Challenge, run by the Department for Science, Innovation and Technology (DSIT), the Cabinet Office, Global Government Forum, and NTT DATA.
Interdisciplinary, cross-departmental teams will now form around each idea with guidance from senior digital leaders. On 24-25 March in London, teams will attend a hackathon at the Innovation 2026 event, then pitch to a judging panel that includes chief digital and information officers from four major departments, a Ministry of Defence director general, the government's chief data officer and chief artificial intelligence officer, and NTT DATA's head of public sector. Four teams will progress to the May final, where one will receive £50,000 of digital project development.
The aim goes beyond a single winner: accelerate practical service improvements and save resources through responsible use of data and AI. As Minister for data and modern digital government Ian Murray put it: "This challenge shows how civil servants will put AI and data to work improving public services… test and learn, then scale what works." He added that the shortlisted ideas cut paperwork, reduce delays and free staff to focus on people.
The semi-finalists
- Casework compliance assistant
AI guidance and support with data analysis, verification and processing for Department for Work and Pensions benefits teams.
Marlon Woodley, Counter Fraud, Compliance and Debt verification assurance officer, Department for Work and Pensions - Codified digital solution assurance
A cross-government rules engine to reduce duplication, fragmentation and complexity in assurance processes.
Kelcey Swain, data foundations for AI lead, Cabinet Office - Fraudulent document detection
AI-powered verification service to identify falsified or tampered evidence and data.
Eric Edward, trainee tax specialist, His Majesty's Revenue and Customs - Offshore wind adaptation to protect wildlife
Use cameras and AI to shut down turbines when protected birds approach, reducing collisions.
Sabrina Schalz, Nationally Significant Infrastructure Projects ecologist, Ministry of Housing, Communities and Local Government - Real-time weather event impact intelligence
Combine government and open data to track flooding and other weather events in real time.
Robert Cowling, senior hydrometeorologist, Environment Agency - FOI request assistant
Multi-agent AI to support triage, costing, data gathering and drafting of responses to Freedom of Information requests.
Matthew Pickering, stakeholder engagement lead, Department for Energy Security and Net Zero - Virtual personas for policy testing
Create AI personalities to explore views and reactions across demographics.
George Griffiths, data scientist, Foreign, Commonwealth and Development Office; and Amish Sarpotdar, economic adviser, Department for Energy Security and Net Zero - AI for capability discovery
Use AI to scale advanced analysis of complex open data for early capability insights.
Mark Brushett, Ministry of Defence
How the shortlist was built
The programme received 252 ideas last autumn after a call from programme champions, including the permanent secretaries of the Cabinet Office and DSIT. Initial reviews by the project team and senior data leaders across government were followed by workshops run by NTT DATA and 13 civil service organisations, producing 57 submissions to the judging panel.
Aimee Smith, the government's chief data officer, said: "Congratulations to everyone behind these ideas - we had more nominations than ever this year with over 60 high-quality submissions… We will continue to explore how every strong idea can realise its potential."
David Filmer, head of public sector at NTT DATA UK and Ireland and judging panel chair, said it was "incredibly difficult" to pick the semi-finalists. He noted that the emergence of large language models "has opened up huge opportunities to improve public services, save taxpayers' money and strengthen public servants' tools across a wide variety of roles, activities and agendas."
Why this matters for government teams
This shortlist is a practical signal of where AI can deliver value now. If you're leading policy, operations or DDaT, here's how to use it.
- Spot near-term wins: casework triage and guidance (benefits), FOI handling, fraud and document checks, environmental monitoring (wildlife protection), real-time incident intelligence (flooding), rules engines for assurance, policy testing with simulated personas, and capability scanning from open data.
- Prep the groundwork: confirm lawful data access, DPIAs, retention and sharing; set model risk controls and human-in-the-loop decision points; plan evaluation metrics from day one. See UK guidance on AI in government and the cross-government digital and data roadmap.
- Pilot fast, measure outcomes, iterate: short discovery, tight alpha, clear service-level goals (time saved, quality, risk). Upskill policy and delivery leads so decisions keep pace with the tech: try the AI Learning Path for Policy Makers.
- Work across boundaries: bring policy, ops, legal, security, procurement and assurance into the room early. The "codified assurance" idea is a cue to standardise checks so scaling doesn't stall.
- Build capability, not one-offs: document patterns, reusable components and governance so successful pilots can be replicated. For curated resources, see AI for Government.
Key dates
- 24-25 March 2026, London: Hackathon and semi-final pitches at Innovation 2026.
- May 2026: Final and £50,000 award for digital project development.
The headline: real projects, real constraints, and a clear path to scale what works. If you have a related use case, now is the time to line up data, guardrails and a fast pilot so you can move when the door opens.
Your membership also unlocks:
