Artificial intelligence is currently the buzzword in social housing. With increased regulation, rising workloads and constant pressure on budgets, housing teams have a lot to deal with.
But AI isn’t a magic wand; it won’t instantly solve problems or improve standards.
Generic AI isn’t enough
The thing about generic AI is that it doesn’t ‘get’ the nuances of housing. It’s certainly useful for tasks such as summarising documents or organising calendars but it won’t understand the language of housing, the meaning of tenancy data or the risks associated with getting it wrong.
Think about the word ‘void’. To a general AI tool, it could mean just an empty space. But to someone working in housing, that means an empty property, something that needs attention to get a family moved in and rent coming in again.
AI could be transformational for this sector, but only if it’s trained to act like a housing expert.
Here are three ways AI can make a real difference:
1. Speaking our language
One of the big challenges in social housing is that people say one thing in a million different ways. One resident might say they have a ‘bit of damp’ in their home while another might talk about ‘significant mould’. Then the housing officer might log the complaint as DMC (damp, mould & condensation). A standard AI could easily get confused with these different expressions and terminologies.
It’s essential to make sure the AI you use really understands the different ways people may communicate. It needs to be taught the specific words, phrases and even slang that residents will use, whether they are talking about a broken window or showing signs that they are struggling financially. The aim is that AI can understand what someone means, however they say it, so nothing important gets missed.
We are building a comprehensive housing dictionary for AI so that it understands not only the words but also what they actually mean in a housing context and how serious they might be.
This is really valuable. For example, a resident might email to say they are feeling exasperated, at the end of their tether or extremely vulnerable, and the system will assign them a sentiment rating out of 10 so that the housing officer can alerted that they need to contact them immediately. Another resident might call and state they are confident, don’t need help and are happy with everything, in which case the system will give them a sentiment rating of one, leaving the housing officer to focus on higher priority cases.
2. Better data, not perfect data
When you work in housing, you know the headache of information spread across multiple systems. You might have tenancies in one place, rent collections in another and repairs somewhere else.
A well-trained AI system doesn’t throw its metaphorical hands up in frustration at this; instead, it should thrive on connecting these dots.
Housing providers often assume they need to sort out poor data before they can use AI. However, AI can help improve data quality because it can identify missing, mismatched or inaccurate information and learn from patterns.
For example, it can check if a gas safety certificate is missing or if something a resident reported doesn’t match the repair that was logged. AI can look at information from different places and highlight things that look wrong and even suggest how to fix them.
For better data and data insights, AI is only as good as the training it receives. While AI training models require millions of data points to learn properly, social housing is a small sector and most housing providers have fewer than 100,000 properties. This creates a challenge.
The solution lies in having both domain expertise and data science knowledge to identify which information matters most. By focusing on the most statistically significant data features, we can make AI systems valuable even when working with limited data volumes or data of varying quality.
3. Predicting the unpredictable
Where AI can really shine is by predicting problems before they happen. This means housing teams can get in early to help people and prevent things from getting worse. For instance, we worked with Southwark Council and Wolverhampton Homes to get AI doing exactly that.
For Southwark Council, we trained AI to analyse how people manage their rent and spot those who might be at risk of falling behind, even if they seem okay at that moment. The AI looks at different clues to give a risk score, so the council can focus its support on where it’s really needed. This is smarter than just chasing people who owe the most rent because someone with higher arrears might still be able to pay, while someone with a smaller debt might be really struggling.
Similarly, Wolverhampton Homes is using AI to predict which homes are at risk of damp and mould. By learning from past repairs, complaints, the condition of properties and even the weather, the AI can highlight the at-risk homes. This means they can carry out inspections and repairs before those homes become unsafe.
The success of these projects is down to using AI that has been specifically built for social housing. They have been created with housing experts and trained on years of sector-specific information.
The systems are all connected, so the AI can learn from what’s happening right now, such as a resident’s phone call or a new complaint. This means it gets better and better at understanding different situations and at giving helpful insights, alongside the final decision of what to do always being made by a human.
What’s also important is to use ‘explainable AI’. This means that the AI doesn’t just give an answer, it also tells you why it thinks something and how sure it is. This makes a big difference because housing teams can see the reasons behind an AI’s suggestions and decide if it makes sense rather than blindly following the advice of a computer. If the AI isn’t confident or the reasons are weak, they know not to act on the AI’s advice.
What’s next for housing AI?
Looking ahead, AI has the potential to do some amazing things in social housing such as analysing photos of repairs and explaining exactly what needs fixing. A resident could send a picture of a leaky tap to their housing provider and the AI could suggest to the housing officer that a new washer is needed.
As the regulatory burden grows and budgets remain tight, AI that really understands housing could be hugely beneficial. But it needs to speak the same language and understand the sector’s specific challenges to make a real difference. It’s important to remember that AI will only be as clever as we teach it to be.
Trevor Hampton is the director of housing solutions at NEC Software Solutions
