Two Gates Cambridge scholars are involved in a local project that uses AI to better predict social housing issues for the most vulnerable.
interesting point [about this project]What is unique about this project is that it not only predicts observational data, but also data from real experience.
Ramit Debnath
Cambridge researchers, including Gates Cambridge Scholars adib hussein saeed [2025] and Ramit Debnath [2018], They are developing artificial intelligence tools that can tell UK parliaments which social housing tenants are most at risk before a potential crisis occurs.
The project, called PRISM (Predictive Risk Intelligence for Social Housing Maintenance), is a collaboration between the University of Cambridge and two local councils: Cambridge City Council and South Cambridgeshire District Council. This is supported by the Local Government AI Accelerator, a new initiative from ai@cam, the university’s flagship mission on artificial intelligence.
Instead of waiting for a tenant to report a leaky roof or damp bedroom walls, a computer model scans data from thousands of properties and flags the properties most likely to deteriorate and those most likely to suffer if they do.
Peter Campbell, director of housing at South Cambridgeshire District Council, which manages around 5,500 social housing units, said: “At the moment we are waiting for the situation to resolve before we take action.” “Very often when something breaks, it’s not just the object itself that is damaged, but the damage that is caused by that breakage. For example, it’s not just the roof that needs to be replaced; water can get in there and cause damage to other parts of the building.”
Together, the two councils manage thousands of rental properties across an unusually wide geography: the densely urban city of Cambridge and the more suburban and rural South Cambridgeshire, where it takes more than an hour to travel between the two addresses.
Campbell says better data can make teams much more efficient. In my previous role, I implemented route planning software for repair staff and watched them increase the number of visits they made per day from six to eight.
single risk score
The system being developed by Professors Ronita Vardhan and Debnath from the University of Cambridge’s School of Architecture and the Center for Human Inspiration and AI (CHIA) combines three data sources to create a single risk score for each property.
The first source is satellite data. Bardhan’s team has spent years developing AI algorithms that detect heat loss from buildings using thermal images captured by low-orbit satellites. The study created a building-level dataset covering the whole of England and Wales, mapping energy efficiency characteristics by property.
The second source of information is traditional housing data. These include building type, energy performance certificate rating, moisture and mold records, and repair history.
A third source of information is what researchers call ‘soft’ data, such as fuel poverty indicators, rent arrears and accumulated logs of tenant contact, which councils already have but rarely use at scale.
“Housing staff have a much more grounded idea of how they view vulnerabilities,” says Debnath, associate professor in the Department of Architecture and executive director of the Cambridge Center for Human-Inspired AI (CHIA). “They have information about fuel shortages, repair records, rental history, health checks, etc. What’s unique and interesting about this project is that they’re making predictions based not only on observational data, but also on data from real-world experience.”
The researchers say the result will be a dashboard showing a map of dangerous hotspots, flagging buildings in poor condition as well as highlighting where vulnerable people live.
Campbell gives specific examples of what dashboards do. Imagine you have two identical properties, both with cracks in their exterior walls. “In one building, the family goes to work all day, so the impact of heat loss due to cracks on the family is minimal,” he said. “The same property next door is home to a single person with a disability who is homebound, and heat loss can have an even greater impact. With this tool, we can target those who are most in need. It’s not just about the property, it’s about the people who live there.”
Changing approaches to social housing
The project reflects a broader change in the way social housing is regulated in the UK, with the government’s expectation to make better use of data to plan services.
One area of particular concern is reaching out to tenants who, for whatever reason, have little contact with the council: those with mental health issues, older residents who rarely seek help, or those who hide problems rather than report them.
Keep Humans Updated
The researchers emphasize that PRISM is not designed to make automated decisions about people’s housing or welfare. All alarms generated by the model are reviewed by housing personnel rather than being processed directly by the machine.
This project is designed as a 12-month proof of concept. If successful, the councils say they hope it will serve as a template for social housing authorities in other parts of the UK. Researchers are already building a roadmap for other councils to replicate.
*Photo by BEN ELLIOTT on Unsplash
