Cambridge AI tool aims to flag failing council homes
TL;DR:
- University of Cambridge researchers are developing an AI tool to identify council homes most likely to deteriorate.
- It combines satellite heat-loss imagery, Energy Performance Certificates and anonymised council-tenant contact records into a single risk score.
- Welfare decisions would remain with trained council officers, the researchers stress.
Researchers at the University of Cambridge are building an AI tool, alongside Cambridge City Council and South Cambridgeshire District Council, designed to flag problems with council housing before they become critical. The system would scan data across thousands of properties and surface those most at risk of decline — and the residents most likely to be harmed as a result.
Prevention over reaction
The tool draws three data sources into one risk score per property: satellite thermal imagery that detects heat loss, conventional housing records such as Energy Performance Certificates, and anonymised logs of contact between councils and tenants. The results would feed a dashboard mapping “risk hotspots” for housing teams. Professor Ronita Bardhan, who is leading the work, described it as “a starting point” the team hopes can be replicated by councils across the country.
The appeal for stretched local authorities is a shift from reactive to preventive maintenance. Peter Campbell, head of housing at South Cambridgeshire, put the current reality bluntly: “we’re very much waiting for things to break before we act”, with damage often cascading into other parts of a home. Catching deterioration early could cut both costs and the disruption to tenants. Crucially, the researchers say welfare judgements would stay with trained officers rather than being handed to an algorithm — a design choice that matters given how often automated decision-making in the public sector draws scrutiny.
Looking forward
The project joins a growing set of UK public-sector tools that apply AI to long-standing operational problems, from planning decisions to council services. Its real test will be whether risk scores built largely on proxy data — heat signatures, certificates and call logs — reliably identify the homes and households that most need help, without flagging the wrong ones. If the Cambridge model proves itself, the bigger prize is a template other councils can adopt rather than rebuild from scratch.