Overcoming Common Utility “Myths”
‘My utility is too small to get any value from AI/ML’
The truth is that AI/ML can provide good results and help utilities of any size make informed decisions. While it IS TRUE that AI/ML improves with more data, even small utilities can get positive results. VODA.ai has several customers with less than 500 miles of water mains. One is a small US-utility with less than 200 miles of mains and very few annual breaks. Over the last 24-months, they have experienced eight water main breaks.
VODA.ai was able to predict seven of the eight failures of a small US-utility.
Of these, VODA.ai was able to predict seven of the eight failures (87.5% of the breaks) in less than two percent of the worst pipes in the entire system. The utility uses the AI/ML predictions to better prepare for failures by targeting operational activities. Additionally, the utility has saved hundreds of thousands of dollars by using the AI/ML predictions to make informed paving and replacement decisions.
‘My GIS data is too poor to get good results from AI/ML’
The truth is that AI/ML can provide good results and help utilities even with poor GIS data quality. While it IS TRUE that AI/ML tools ARE more accurate with better data inputs, most AI/ML companies can assist utilities with data cleaning and improvement tools. Even when there is data that can’t be known or improved, the predictions generated by AI/ML are more accurate than traditional baseline methods.
VODA.ai was still able to predict 50% of the water main failures
VODA.ai worked with a mid-sized US-utility with around 1,400 miles of water main. At the start of the project, they provided a map of the water mains and used a magic-marker to circle large areas of the system and simply said, “we think these pipes are mostly AC”, and “we think these pipes are mostly PVC”, and so forth. Using these as inputs, VODA.ai was still able to predict 50% of the water main failures in less than eight percent of the pipes.
‘We can do it ourselves and get the same results’
The truth is that statistical methods, linear models, and human engineering judgement can’t compete with the data-driven pattern recognition of AI/ML, which is able to consider thousands of variables in a fraction of the time. A lot of innovative utilities today are using traditional models and statistical tools to calculate Likelihood of Failure (“LoF”) and the truth is that these utilities ARE seeing a lot of benefits using this approach. At VODA.ai we applaud these forward thinkers for their use of LoF in making asset management-based decisions and are excited to see their results.
VODA.ai has outperformed every single model.
In fact, these are our ideal clients as they already see the value in using LoF and other risk-based metrics in their approach. However, by using an AI/ML based tool like VODA.ai, the same utilities can reduce the workload on engineering staff by hundreds of man-hours AND improve their results. VODA.ai has worked with a number of these sophisticated utilities who wanted to compare their existing results with those from our AI/ML engine. The result? VODA.ai has outperformed every single statistical, desktop, traditional, or other model. Every. Single. Time.