Reducing “False Positive” Exceptions for an Energy Management Leader

The backstory

As part of service delivery, an energy management company was reviewing the energy invoices of their clients to find savings from erroneous bills and bring in accuracy. Despite rules being implemented to raise exceptions and indicate that an invoice might have a potential error, analysts were required to manually review these bills for errors. The customer was getting high numbers of false positives, putting strains on time, effort, money and meeting SLA requirements.

The solution

We collaborated with the customer to create an end-to-end solution that focused on keeping machine learning models at its core. We trained the model on past exception data to help its resolutions determine if an exception is being recorded accurately. We also ensured that the solution allowed room for automation to reduce the amount of manual effort in the process as the solution could reduce false exceptions by as much as 80%.