Intelligent housing repair software made simple.

Social Housing Repairs and AI – Chatting with ChatGPT

There is a lot of buzz around AI and how it might help social housing organisations in all manner of ways. But has anyone thought to ask an AI system itself about how it thinks it might be able to assist? We thought we’d ask ChatGPT if it understood the challenges of diagnosing social housing repairs.

The results were… interesting:

To begin with, we just thought we’d ask if it could help with a housing repair – with no context:

Can you help with my repair?

Omfax: My basin is loose

I assume you are referring to a loose basin in your bathroom or kitchen. If your basin is loose, it’s important to address the issue promptly to prevent further damage or potential injury. Here are a few steps you can take to tighten a loose basin:

• Turn off the water supply: Before you start working on your basin, make sure to turn off the water supply to avoid any water damage or potential injuries.

• Remove the basin: You may need to remove the basin from its mountings to access the fixings that are holding it in place. Depending on the type of basin, you may need to remove the taps, waste fittings, or any other fixtures that may be in the way.

• Tighten the fixings: Once you have access to the fixings, use a wrench or spanner to tighten them. If the fixings are old or rusted, you may need to replace them with new ones.

• Reattach the basin: After you have tightened the fixings, reattach the basin to its mountings and reconnect any fixtures that you removed.

• Test the basin: Turn on the water supply and test the basin to make sure that it’s no longer loose and is working correctly.

If you’re not comfortable with DIY repairs, it’s always best to consult a professional plumber to fix the issue for you.

Omfax: Sure, I know that, but is there a schedule of rates code that represents the work that a contractor might need to carry out?

Yes, there are various Schedule of Rates codes that can be used to represent the work that a contractor may need to carry out when repairing or installing a loose basin.

The most common Schedule of Rates codes that can be used in this scenario include:

• P1103: Removal and refixing of basin or sink

• P1105: Replacement of fixing bolts or brackets for basin or sink

• P1202: Installation of basin or sink, including waste connection and taps

• P1204: Installation of wall-mounted basin or sink, including brackets and fixings

These codes may vary depending on the specific Schedule of Rates that you are using, and the rates associated with each code may also vary depending on your location and the type of work required.

It’s always a good idea to consult with a professional contractor or a pricing expert to ensure that you are using the appropriate Schedule of Rates codes and that the rates associated with each code are fair and reasonable.

We have to admit that this is impressive. Not only did ChatGPT understand what we meant by Schedule of Rates codes, it offered (unprompted) a list of SOR codes relevant to repairing our loose basin.

Can AI help with repairs in social housing?

So, we thought we’d ask it to stand back from the loose basin repair and consider the wider issues:

Omfax: How can AI accurately link outcomes of a diagnostic to codes, for example, in a social housing context? Schedule of rates codes link to very specific repairs but these codes change over time, as well as new codes being introduced. Similarly, priority codes are quite dynamic, based on many factors, for example time of year, vulnerability or age of tenant and other factors that may impact the urgency.

To accurately link outcomes of a diagnostic to codes in a social housing context, AI systems can use a combination of techniques, such as natural language processing (NLP), machine learning, and data analytics. Here are some steps an AI system might take:

• Collect and organise data: The AI system will need access to a large dataset of social housing repair codes, priority codes and diagnostic outcomes. The data will need to be organised in a way that allows the system to make connections between codes and outcomes.

• Train the system: The AI system can be trained on the dataset using machine learning algorithms, which can identify patterns in the data and make predictions about which codes are likely to be associated with specific diagnostic outcomes.

• Use natural language processing: To make sense of the data, the AI system can use natural language processing to understand the meaning of repair codes and diagnostic outcomes. This can help the system accurately identify which codes are most relevant to a particular diagnostic outcome.

• Adapt to changes: To keep up with changes in the repair codes and priority codes, the AI system can be programmed to learn and adapt over time. For example, the system can be updated with new data and trained on new patterns to improve accuracy.

• Consider contextual factors: To factor in contextual factors that may impact the urgency of a repair, the AI system can be programmed to take into account additional data, such as the time of year, tenant vulnerability, or age. This can help the system prioritise repairs and ensure that urgent repairs are addressed promptly.

Overall, an AI system that uses NLP, machine learning and data analytics, can help accurately link diagnostic outcomes to repair and priority codes in a social housing context. However, it is important to ensure that the system is regularly updated and monitored, to ensure that it continues to perform accurately over time.

Can AI handle the challenges around accurate repairs diagnosis?

Omfax: But, if a particular repair, priority or other code changes overnight, I assume retraining AI won’t be immediate, so could the wrong codes be used?

You are correct that if a code changes overnight, it may take some time before an AI system can be retrained to accurately recognise and link that code to a specific repair or priority outcome. Therefore, it’s important to have a system in place to update codes and ensure that the AI system is aware of any changes as quickly as possible. Here are some potential strategies for dealing with code changes:

• Monitor code changes: Establish a process for monitoring changes to codes and ensure that the AI system is updated accordingly. This could involve designating a team or individual to monitor and update codes in real-time.

• Build in redundancy: To minimise the risk of the wrong codes being used, it may be helpful to build in redundancy to the AI system. For example, the system could be programmed to use multiple codes or prioritise certain codes over others, if there is ambiguity or uncertainty.

• Develop a feedback loop: Establish a process for collecting feedback from tenants and maintenance staff, to identify when the wrong codes are being used. This feedback can then be used to improve the accuracy of the AI system over time.

• Use human oversight: Consider incorporating human oversight into the AI system, such as having a maintenance supervisor review the codes generated by the AI system before assigning repairs. This can help catch any errors or discrepancies before they become a problem.

Overall, while retraining an AI system may take some time, there are strategies that can be put in place to minimise the risk of the wrong codes being used. By monitoring code changes, building in redundancy, collecting feedback, and using human oversight, an AI system can be designed to adapt and improve over time.

Omfax: So, it is possible that an extremely urgent code change (or indeed a raft of changes) may not take effect immediately?

Yes, it is possible that an extremely urgent code change may not take effect immediately, especially if the AI system has not yet been updated with the new codes. Depending on the complexity of the system, the number of codes that have changed, and the availability of resources, it may take some time to update and retrain the AI system with the new information.

In situations where urgent code changes need to be made, it may be necessary to take other steps to address the issue in the short term. For example, maintenance staff could be informed of the changes and instructed to manually assign repairs based on the new codes until the AI system can be updated. Alternatively, a temporary fix could be put in place to prevent any critical repairs from being missed while the AI system is updated.

It’s important to have a contingency plan in place for situations where urgent code changes are required, to ensure that repairs are addressed promptly and accurately. This may involve having a team responsible for monitoring and updating codes, establishing clear communication channels, and having backup systems or processes in place to address any issues that may arise.

Does AI understand it’s own limits?

So, it’s pleasing to note that even ChatGPT recognises the challenges. i.e. that an AI system could diagnose the wrong SOR code for a repair, if it has been recently changed, or if the system has not been re-trained on the new code.

ChatGPT’s suggestion of using human checking of an AI systems diagnosis of particular SOR code for repairs, rather misses the point. The advantage of any system diagnosing housing repairs is that it provides the right SOR codes, priority level and associated information, reliably and automatically. Relying on human intervention to ‘check its work’ is really a non-starter.

Consequently, we can’t see any purely AI-based system working well to diagnose social housing repairs any time soon. But AI can certainly help with the natural language part of the user interaction and hand off to either advisor-assisted repairs diagnosis, or online tenant self-service repairs.