You are viewing 1 of your 1 free articles
The London borough using AI to solve temporary accommodation allocations
Newham has the highest number of households in temporary accommodation in the country. Its solution? Becoming one of the first councils to appoint a head of AI. Ella Jessel reports
Newham has the highest number of households in temporary accommodation in the country. Its solution? Becoming one of the first councils to appoint a head of AI #UKhousingElla Jessel speaks to Newham Council’s head of AI to find out more about the role, and what other councils could learn #UKhousingWhen Nathan Nagaiah, head of artificial intelligence (AI) at Newham Council, asked officers for their biggest problem there was a clear winner: temporary accommodation.
The east London borough has around 7,000 residents in temporary accommodation, the highest figure in England. The crisis is financially crippling: temporary accommodation is expected to cost the borough £140m by 2028-29, creating a budget gap of £53m.
High demand has led to a huge bottleneck in homelessness applications, and the council is turning to AI for help. It has developed a prototype tool it hopes could drastically speed up allocations.
“What once took local government caseworkers in temporary accommodation over three hours of manual work can now be completed in just under three seconds,” says Mr Nagaiah, who hopes the tool will move to live testing in July.
Newham is far from the only authority looking for ways to cope with sky-high temporary accommodation numbers. Could Newham’s AI model save costs and help households experiencing homelessness, and what are the risks involved? Inside Housing sat down with Mr Nagaiah earlier this year to discuss the tool, and his journey from maths teacher to AI lead.
From classroom to council
Mr Nagaiah arrives at our meeting on a damp morning in January armed with a large umbrella, and we settle into plush sofas at a central London bistro.
Originally from Kuala Lumpur, Malaysia, Mr Nagaiah came to the UK to study at UCL and has stayed ever since, he tells me. After graduating in 2009, he became a maths teacher at schools in Essex and London, and then at colleges, rising to become principal of South Thames College Group between 2021 and 2022. In the past five years his career has veered into local government, coinciding with the explosion of AI.
It was while working as the college principal that he first became interested in whether data could tell him more about his students. “We were looking at students who were underperforming and using technology to track them. We found out which postcodes they come from, what their family circumstances are, and how that was contributing to their educational outcome,” he says.
“I started to get my head further into the data and, being a mathematician, it actually comes naturally.”
As principal, he had some of his first encounters with homelessness: “There were students who had bailiffs coming knocking at their door, so they couldn’t turn up to the school.” During the Covid pandemic, Mr Nagaiah was appointed to the UK Cyber Security Council to chair the cyber security programme. In this role he began to explore automation and how to predict cyberattacks.
After a stint at Brent Council as programme lead for digitisation, Mr Nagaiah joined Newham in 2022, initially as data economy sector lead. He carries out a smattering of advisory roles too, including for the Ministry of Housing, Communities and Local Government, as well as a “few boards” and a part-time teaching role at the University of Oxford.
When ChatGPT launched a year later, Mr Nagaiah was perfectly placed to explore its potential use in the public sector: “It was only when ChatGPT came along that AI started to get big; but AI is not new, it’s been around since the 1960s. I started looking at how we could do automations and use artificial intelligence using all the data we have in local government.”
Capitalising on the wave of public interest, he then helped found the UK Centre for AI in the Public Sector in 2024, with Newham Council in partnership with the University of East London (UEL). The centre is dedicated to pioneering AI research, training and solutions for the public sector, and its mission is to tackle some of the everyday challenges faced by local authorities, such as housing.
“We’re not here to solve the world’s problems,” Mr Nagaiah says. “We are employed locally, so we have to solve the local problems and then we can look regionally, and then go national.”
The Domus project
The centre’s flagship project is Domus, the Latin word for ‘home’. Put simply, the tool is designed to speed up the allocation of temporary accommodation. “We spoke to council officers, and they said our single biggest cost pressure and single biggest problem in the local authority is temporary accommodation,” Mr Nagaiah says.
“Within temporary accommodation there are so many different problems, from homelessness, people not being able to pay the rent, maintenance and properties,” he says. He zeroed in on allocations after identifying it as an area where officers were spending thousands of hours matching households with homes.
Under the current system, staff pore over spreadsheets to create a list of 20 candidates who best match each property. Criteria include how many bedrooms applicants require, if they need medical facilities nearby and whether their children have special educational needs. Affordability is another key criteria, with officers needing to know how much applicants would be able to ‘top up’ rent, given properties are rarely within Local Housing Allowance rates.
This painstaking matching process slows down placements and can also lead to extended stays in temporary accommodation, pushing costs up further. “It takes around three hours per property, per caseworker, and there are thousands of people on the waiting list,” Mr Nagaiah says.
“We then have to contact all 20 people we have selected for viewings. By the time we do, the property is often gone. Imagine the complexity that comes with it,” he says, grimacing in exasperation.
Mr Nagaiah asked his data engineers for a solution that would use automation to find the 20 people at the top of the list and spent months developing a prototype. The tool works by integrating data science, machine learning and automation, and according to a document he sends me, uses “structured feature codes and automating the search and recommendation process through the use of a local large language model”.
Thanks to his college teaching days, Mr Nagaiah is able to translate the jargon: “AI will present the 20 people that are suitable. Against each of these households it will say what the percentage match for each of them is.”
All match decisions are reviewed by officers, meaning the suggestions remain “advisory”. To protect against bias, the first step of the process is to cleanse the data and remove any information not relevant to property allocation. “So, name, gender, ethnicity, sexual orientation. It’s completely anonymous, and the machine will not know who it is allocating to,” he says.
Trials showed the prototype was 89.2% accurate, and the team is now trying to train it to be 95% accurate before heading to live trials in July. It will then need full council approval before any wider roll-out.
I ask how accurate the current human and spreadsheet approach is. “Less than 60%,” he says. The idea is that the automation will free up officers so they can spend more time helping residents.
“People fear AI will take their jobs. But the time we save from the manual tasking means we can spend more time with residents, asking why they are becoming homeless, and trying to help them find employment.”
Some of the challenges for Domus are price fluctuations in the property market that the model can’t predict, and on the technological side, the system updates required in maintaining the large dataset. Another challenge is data quality. “To understand AI you need to understand data. AI can solve problems, but you need to know what data you have, before you can see whether there’s a solution.
“I think the lack of high-quality data within local government is a problem.”
Mr Nagaiah is leading on a new data strategy across the whole council, but in the housing department it is an area that has come under scrutiny. In 2024, the council was given a C4 rating by the Regulator of Social Housing, and data was one of its major failings.
A council-commissioned report, published at the end of last year, found data collection and handling was “arguably the most significant driver” of the C4 rating. “It’s not a good report,” Mr Nagaiah admits, but he insists the council is on the right path. “As a local authority we are making significant progress.”
Educating the public
Nationally, councils are starting to explore the use of AI solutions, including Hertfordshire County Council’s pothole robot, and Hammersmith and Fulham deploying AI to find damp in its housing stock. A report by the County Councils Network last year found 85% of respondents said they have been using AI, with the remaining 15% considering adopting or trialling this technology.
But the report also highlighted major barriers, from staff capacity to skills. According to Mr Nagaiah, these challenges, alongside a lack of investment, drove Newham to link up with UEL to create the UK Centre for AI in the Public Sector. “You can’t innovate without funding, and we need collaborators and funding streams.”
While the main thrust of Mr Nagaiah’s role is to develop AI solutions, it also involves educating the public. “There’s a lot of public misunderstanding about AI. A lot of public think that AI is just ChatGPT. We have to actually train the public, to say, ‘Look, this is what AI is, and this is how AI is being deployed to improve the efficiency of delivery in the council.’”
The council recently debuted its second AI solution, a translation tool for residents who do not speak English. It translates in real-time into 92 languages, allowing council staff to talk to residents via a headset and improve access to services. Newham has also built an in-house procurement tool called Agent Ali, designed to help council staff with procurement information.
It’s not only staff who are using AI – huge numbers of resident complaints to councils are now written using ChatGPT. Then, Mr Nagaiah says, some council officers also run their responses through the bot, resulting in “two machines talking to each other”.
He adopts a robotic voice: “The response will say: ‘Dear Ella, thank you for writing to us today... etc.’ It’s better to build a custom AI agent that will actually reply with the correct information, issue a case reference number and refer to council policies.”
Mr Nagaiah is keen to assuage fears of AI as a threatening force and show how it can be harnessed for the benefits of both councils and residents. “We want to show that AI is for the public, and we want the public to be involved at every stage.”
Other recent longform articles on AI
The sector needs to up the pace on harnessing artificial intelligence
As expectations rise, AI and data can play a crucial role in driving improvements in customer experience, writes Fiona Harris, chief information and data officer at Vivid
Social housing AI platform appoints new chief executive ahead of launch
An artificial intelligence tool designed for social housing professionals will launch with a new boss after a year’s worth of input from large social landlords
Is AI a quickstep to better services? How social landlords see the risks and opportunities
From tracking data on retrofit to scanning for ‘silent’ tenants, housing associations are using AI tools in their day-to-day work. Kate Youde explores landlords’ most recent annual reports to find out how the technology is changing the sector
Which housing associations are leading the way on AI?
Our new report takes stock of where the sector is at currently on artificial intelligence and shares practical examples, writes Lizzi Hearn, external affairs manager at the National Housing Federation
Could AI help assess housing need and deliver more social homes?
Housing policies analysed by Explainable Artificial Intelligence (XAI) models could help policymakers improve social housing in the UK and “avoid bad policy design”, according to a leading computational economist. Ben Lee finds out how
Are you subscribed to Inside Housing?
Inside Housing is the most comprehensive information service for UK housing professionals. Not subscribed yet?
Related stories
