Diagnosing Repairs from Pictures
AI and Machine Learning
The next generation of housing technology is here. We are finalising our technology that allows customer pictures to be analysed by computers to diagnose housing repairs to a higher level of success than before.
Repairs operatives often turn up to fix domestic household repairs, but the repair has been diagnosed incorrectly or the operative has the wrong parts to fix the issue. This results in the need for an extra visit.
After interviewing operatives, tenants and call centre agents and conducting more than 200 call listening, we found out that one of the steps that was bringing more problems to the journey was the “Diagnose” step. The repairs were wrongly diagnosed, consequently the authority was sending the wrong operative or sending the operative without the correct parts to fix the problem, causing more revisits, unsolved repairs and angry tenants. The reasons for misdiagnosing were sometimes human mistakes and others, technology problems.
The agent is responsible for diagnosing the repair from what the tenant tells him. In order to report, diagnose and book an appointment for the repair, the agent usually takes an average of 6 min, but sometimes because of technology constraints and communication gaps, the call can take more than 15 mins.
Depending on the location, operatives usually visit between 5 and 7 properties per day and the incidence of diagnosing the wrong part occurs around 20% of the time.
How we will solve the problem?
Automate Diagnostic Tool
Diagnosing the nature of the repair will require using machine learning to compare hundreds of thousands of images and the actual part used.
- captures images from an external source,
- diagnoses the nature of the repair and the required part
- compares the part to the one the operative has in their van
- Reduction of human errors
- Increase of the “first time fix”
- Reduce communication gaps
- Reduce call centre costs:
- Less talk times
- Less waiting times
- Less number of calls