The University of Cambridge’s flagship mission on AI, ai@cam, has today announced the six projects selected for its inaugural Local Government AI Accelerator - a programme that establishes a new model for how universities and government can work together to advance AI innovation in public services.
Funded by the Ministry of Housing, Communities and Local Government (MHCLG), the 12-month Accelerator pairs Cambridge researchers directly with local councils to develop practical, proof-of-concept AI solutions to real operational challenges, from automating housing data collection to detecting fly-tipping using cameras on refuse vehicles.
The Accelerator pilots a new model for university-council collaboration by embedding councils as active partners from the outset and building shared capability across the sector. Each project will receive up to £25,000, dedicated technical support from machine learning engineers, and access to a structured community of practice.Crucially, the programme places public concerns directly into the research process. Building on our 2025 public dialogues on AI in local government, in May, residents will have a chance to shape how AI tools are developed through facilitated dialogue sessions.
“This partnership with MHCLG represents a new model for how universities and government can work together to advance AI innovation in public services,” said Jessica Montgomery, Director of ai@cam. “We’re building a community of practice where local authorities can share challenges and collectively advance what’s possible in public service delivery.”
The programme was shaped by insights from a workshop held in November 2025, where 21 local authorities shared their most pressing operational challenges, highlighting the significant potential for AI to provide practical support. These discussions brought into sharp focus the pressures facing councils: residents navigating fragmented services, vulnerable people at risk of falling through gaps in provision, and staff burdened by manual administrative work. The Accelerator has been designed to respond directly to these challenges.
The six funded projects
The six selected projects reflect these priorities in practice, addressing a range of operational challenges across local government.
AI-Enabled Surveys for Housing Trajectories
Principal Investigator: Dr Matteo Zallio, University of Cambridge
Council Partner: Greater Cambridge Shared Planning Service (Cambridge City Council and South Cambridgeshire District Council)
Preparing the annual Housing Trajectory is a statutory requirement for all English local authorities, yet the process of distributing, chasing, and analysing site-specific questionnaires sent to developers and housebuilders remains highly manual and time-consuming. This project, led by Dr Matteo Zallio at the University of Cambridge in partnership with the Greater Cambridge Shared Planning Service, will develop a human-in-the-loop AI-enabled workflow to automate questionnaire drafting and distribution, convert free-text responses into structured data, and provide live monitoring dashboards for planning officers. By freeing officers from repetitive administrative tasks, the project aims to deliver faster and more consistent housing delivery evidence, improve the experience for external respondents, and produce an open, reusable framework that other local authorities can adopt.
PRISM: Predictive Risk Intelligence for Social Housing Maintenance
Principal Investigators: Professor Ronita Bardhan and Dr Ramit Debnath, University of Cambridge
Council Partner: Cambridge City Council and South Cambridgeshire District Council (Housing Departments)
PRISM will develop a multimodal AI platform that fuses environmental, structural, and socioeconomic data to generate dynamic Risk Hotspot Maps for social housing in Cambridgeshire. By identifying properties most at risk of deterioration, the system will enable proactive intervention. The project embeds human oversight throughout, ensuring AI supports—rather than replaces—the professional judgement of housing officers.
MAPLE: Map Automation for Planning and Local Efficiency
Principal Investigator: Alexis Litvine, University of Cambridge
Council Partner: Greater Cambridge Shared Planning Service
Cambridge City Council and South Cambridgeshire District Council collectively process tens of thousands of planning-related map submissions each year, each requiring manual georeferencing and vectorisation by specialist GIS officers — a process that can take between five minutes and over an hour per map. MAPLE will develop an AI-assisted pipeline that automates map detection, georeferencing, and object extraction from submitted planning documents, outputting structured vector data compatible with existing GIS workflows. The project aims to reduce GIS processing time per map by at least 75%, freeing specialist capacity for higher-value work, accelerating application turnaround, and producing open-source tools and methods transferable to local authorities across the UK.
Deep Learning for Fly-tipped Waste Detection
Principal Investigators: Florian Urmetzer and Tyler Holderness, University of Cambridge
Council Partners: South Cambridgeshire District Council and Greater Cambridge Shared Waste Services
Fly-tipping places a significant financial and environmental burden on local authorities, with over 1.15 million incidents recorded across England in 2023/24 alone. This project will develop a deep learning computer vision pipeline that leverages cameras already fitted to Refuse Collection Vehicles to detect fly-tipped waste during routine rounds. The system will automatically extract the location, timestamp, waste description, and photographic evidence needed to generate a report, with human-in-the-loop verification before any action is taken. By converting an existing but underutilised data asset into a proactive detection tool, the project aims to reduce the reporting burden on the public and accelerate the resolution of incidents.
Bidding Behaviour and Outcomes in Choice-Based Lettings
Principal Investigator: Dr Jerry Chen, University of Cambridge
Council Partner: London Borough of Camden
England’s choice-based lettings system gives eligible households the opportunity to bid for social housing, but the complexity of points-based allocation means that many residents make repeated, low-probability bids with limited understanding of their likely outcomes. This project, led by Dr Jerry Chen at the University of Cambridge in partnership with the London Borough of Camden, will use nearly a decade of de-identified administrative data to develop interpretable machine learning models that predict context-sensitive points thresholds and identify behavioural patterns contributing to system inefficiencies. The project will co-design officer-facing and resident-facing prototype tools, and will host a sector-wide symposium to share findings and support replication — contributing to a more transparent, equitable, and well-understood social housing allocation system.
Human-Oriented AI: Design Framework for Reaching Vulnerable Tenants
Principal Investigator: Viviana Bastidas Melo, University of Cambridge
Collaborator: Professor Jennifer Schooling, ARU
Council Partners: Cambridge City Council and South Cambridgeshire District Council
As local authorities develop AI-powered tools to support vulnerable tenants, there is a risk that technical development moves faster than the governance, ethical, and public value frameworks needed to ensure those tools are legitimate and trustworthy. This six-month project, led by Viviana Bastidas Melo at the University of Cambridge in collaboration with Professor Jennifer Schooling at ARU, will apply a Human-OrientedAI socio-technical framework to the early warning systems being developed by Cambridge City Council and South Cambridgeshire District Council. The project will produce an Architecture Vision, Architecture Modelling Scenarios, and an Architecture Roadmap — practical artefacts that embed governance, ethics, and public value requirements alongside technical design, and that can serve as a reusable blueprint for responsible AI adoption across the housing sector.
Scaling impact across local government
Over the next 12 months, project teams will develop and test their proof-of-concept solutions in close collaboration with council partners, supported by a structured programme of technical guidance. Findings will be shared through ongoing engagement with councils, residents, and policymakers, culminating in a final showcase to highlight lessons learned and opportunities for wider adoption across the local government sector.
“This programme is about moving beyond experimentation to understanding what works in practice,” said Professor Neil Lawrence, DeepMind Professor of Machine Learning and Chair of ai@cam. “By working directly with councils and embedding public input throughout, we are creating the conditions to develop AI systems that respond to real operational challenges and deliver meaningful improvements for communities.”
The Local Government AI Accelerator forms part of ai@cam’s broader mission to drive a new wave of AI innovation that delivers tangible public value through collaboration between academia, government, and society.