Is AI a quickstep to better services? How social landlords see the risks and opportunities

So, how and why are housing organisations using AI – and what are the dangers they need to consider?

Rise of the chatbot

There are different forms of AI. However, following the launch of ChatGPT in 2022, the one with which we are probably most familiar is generative AI. This can create new content. ChatGPT, in the chatbot’s own words, is
“a type of artificial intelligence trained to understand and generate human-like text”.

Last year, Leeds-based landlord Yorkshire Housing started experimenting with using generative AI to analyse its tenant satisfaction measures survey results. An employee at the 20,000-home provider fed text feedback captured from customers into Microsoft Copilot, a chatbot tool that helps with work tasks. It asked the AI assistant to summarise key themes and sentiment.

The main benefit of using AI in this way is that it can analyse large volumes of text quickly and save staff time, says Mark Shephard, the association’s head of data, performance and information security. “As well, it’s a cold machine so it takes an objective view and the risk when you bring that human element to [it is] that you get a more subjective interpretation of what’s being said,” he says, adding that there is still skill involved in asking the right questions to get the desired response.

His team tested outcomes by reading through samples and comparing the computer-generated and human interpretations, which Mr Shephard says reduces the risks involved in using AI in this way.

However, it is important to note that AI is not without bias as it is trained on data where decisions often have already been made. This can become an issue when categorising people by gender, for example.

Melanie Smallman, professor of science and technology studies at UCL, says assumptions of what categories are within the data can “perpetuate certain forms of inequality and disadvantage”. She adds that there are concerns that ethnic minority populations are underrepresented in certain health data, but the same populations are overrepresented in other data collected by the police.

Yorkshire Housing has also used AI to summarise text from customer complaints and pull out the root causes. While some of this work has reinforced what the organisation already knew, Mr Shephard says it has sometimes provided “a more nuanced view on something… that gives us a bit more of an angle or insight to what the problem is”.

In both cases, the results have fed into action plans. Mr Shephard says the organisation wants to get to the point where the process is fully automated, with early tests suggesting it would cost 0.01p to process the text from one complaint. He adds that the landlord, which wants to develop a centre of excellence for AI to train staff, is looking to roll out use of Copilot within the business to aid “operational efficiency” with administrative functions.

Longhurst Group is using generative AI in a different way – for music therapy sessions with customers. “They give me their lyrics about their experiences of things like homelessness and I use the AI to turn them into songs in their chosen genre,” community support worker Lee Hamilton said in the 24,000-home association’s annual report. Longhurst has since merged with Grand Union to form Amplius.

Mr Shephard says that while Yorkshire Housing has been concentrating on generative AI, it wants to move “from a reactive to a pre-emptive customer service offer” and use AI to help it identify future issues. Predictive AI uses machine learning and statistical analysis to identify patterns in historic data to forecast future events. The landlord is taking a small-scale “experimental approach” with this in relation to arrears and complaints, which Mr Shephard says allows it to assess the risks and feasibility, and where the technology can bring benefit and return on investment.

Yorkshire Housing is not alone in testing the water. Riverside Group has “taken tentative, controlled and risk-based steps” to explore the use of artificial intelligence and predictive analysis, according to its 2023-24 annual report. However, a spokesperson says it is not possible to provide more details at this stage.

Aster Group, a 37,000-home landlord in London and the South of England, has built a predictive AI tool in-house. “We’re currently trialling machine learning tools that model and track patterns in our maintenance and repair work, and predict pain points before they become an issue for our customers,” said Andy Isted, the association’s director of transformation, technology and data, in a statement. “The ability to quantify massive amounts of anonymised repairs data will be game-changing for future planning by spotting trends so we can better forecast investment and find opportunities to make cost savings.”

Peabody is awaiting the outcome of a pilot held in March 2024 that aims to provide better data to inform retrofit projects. As part of a £240,000 Innovate UK grant, the 108,000-home association worked with Kestrix, a company that specialises in mapping heat loss, and contractor United Living to assess heat loss from about 100 houses in Potters Bar, Hertfordshire. This involved flying a drone overhead with an attached infrared camera. A second pilot last November studied 20 homes in Broxbourne.

“[Customers] give me their lyrics about their experiences of things like homelessness, and I use the AI to turn them into songs in their chosen genre”

The project is about developing Rapid Thermal Performance Assessment Algorithms (RaThPAs), which use AI and thermal imaging to analyse building performance. These algorithms predict thermal transmittance, identify insulation issues and estimate heating demand intensity. If the tool works, Richard Ellis, director of sustainability at Peabody, says it will reduce the time it takes to survey properties and provide an additional data point, alongside Energy Performance Certificates and Parity Projects analytics, to help the association prioritise where to invest to improve energy efficiency.

“We’ve got a limited amount of money to [retrofit homes], and we need to make sure that we are hitting the right properties with the right interventions at the right time so that we can reduce our residents’ energy bills and reduce carbon,” says Mr Ellis.

Peabody could roll out the approach if the cost of this type of survey comes down to £15-£20 per home, and is in talks with Kestrix about the possibility of doing fly-bys before and after a retrofit project has been completed so it can use the tool as a “quality control measure”. Mr Ellis says the association is mindful of residents’ privacy, with the software scrubbing out people’s faces.

Computer says no

WHG is using AI to predict arrears, in an approach that Fay Shanahan, corporate director of operations and IT at the 22,000-home landlord, says demonstrates how AI can deliver “early interventions and better outcomes”.

“However, we’re mindful of the significant risks associated with AI technology, especially around data privacy, security and ethical considerations,” she says. “Our approach is therefore measured and deliberate, ensuring we build on a strong foundation of modern infrastructure and robust data management practices.”

Concerns about privacy, data use and copyright in relation to AI are well documented, but Professor Smallman says associations also need to consider the “appropriateness of this technology” coupled with cost, including maintenance. “Tech companies are really good at selling these technologies and telling people they’re going to fix X and Y, but are they really solving the problems at hand, or is the problem a lack of resource because… AI is not going to help that?” she adds.

She says boards need to ensure transparency and that there is room for tenants to raise concerns, “so people don’t come up against the ‘computer says no’ situation”.

“Social housing isn’t a hard-nosed business, it’s dealing with some of the most vulnerable people in our communities, and sometimes care and kindness and discretion are what’s needed,” she says.

Therefore associations need to think hard about creating a system where that discretion can still “be exercised at the point of service delivery so that people don’t find themselves in a really harsh, unyielding and unkind world”, she adds.

Security breaches

Platform’s Mr Cocker agrees that transparency about where and why a provider is using AI is key, as well as “explainability” – “being able to articulate to people who might not be technical what the artificial intelligence is actually doing and why it’s doing it”, he says.

The government is currently developing the Algorithmic Transparency Recording Standard, designed to help public sector bodies openly publish information about the algorithmic tools they use. Mr Cocker says there is an argument to have an AI register in every association; Platform is working on its own so it has “very clear visibility” of where AI is being used.

The association also has an AI ethical framework, and one of its pillars for the responsible use of AI is security. The number of cyber incidents is rising and “the complexity of the threat landscape is also almost certain to intensify with the use of AI technology”, according to the National Cyber Security Centre’s 2024 annual review. Housing association Bromford, which recently merged with Flagship, published AI user guidelines to mitigate against a breach.

Inside Housing’s annual Risk Register analysis found 85% of landlords identify a cyber threat as a strategic risk. Futures Housing Group has implemented controls to address this concern with software that uses AI. The landlord protects devices with an endpoint security solution that uses AI to detect and stop cyberattacks, and mitigates the risk to its server with antivirus software. For security reasons, the organisation declined to provide more details of how the software works.

Many landlords are developing plans in response to the growth of AI and our snapshot shows there is much for the sector to navigate. However, Mr Shephard of Yorkshire Housing believes AI has a role to play at housing associations.

“It’s everywhere now anyway, so to ignore it is dangerous,” he says. “As with the way that the internet grew, and smartphones and technology, there starts to become an expectation from people. They start to go, ‘Well, why aren’t you using this? I have this service when I use Amazon, when I go to Google… so why don’t you?’ So it’s important that we embrace it, but do that in a safe and controlled way.”