Industry Focus : Fin-tech

The Issue : Detection of Unspecified transactions​

The Solution : A Machine Learning powered bot to detect filled transaction forms​

The  Summary : Every year, the fin-tech industry loses a lot of money in the process of executing transactions that does not contain enough information for international transfers. Our client wanted us create a solution that helps customers execute Forex transactions involving 3 banks or more without unforced errors so that they can save money in transaction fees as well as improve customer satisfaction rates.​

The Result : After exploring the initial problem statement and reviewing the use cases of failed transactions, we detected over 350,000 EUR lost over the last 10 years due to transaction fees on incomplete transfer orders, and this indeed reflected in a lot of negative customer feedback. We detected that the form used for executing these orders was not optimized in a way where the backend was neither smart nor taking into account any human errors that would otherwise occur. We developed a custom chat bot powered by our own machine learning algorithm that interacts directly with the customer. Based on the customer profile and previous transaction history, the chatbot provides a set of reasons, or requests for specific information pertaining to that transaction rather than generalizing the entire flow in a single form. This allows our client to treat each customer transaction individually. Also, if any aberrant information is detected in the conversation with the customer through the chat-bot, that order is automatically kept on hold waiting for an expert to directly verify and contact the customer before executing the order. Using this solution, we were able to bring down the failure rate by 70% in the first quarter and upto 93% in the first year of operation. Our chatbot keeps learning from scenarios that failures had occurred and looks to improve it’s segmentation process. Within 3 years time, we expect the solution to reach 99% accuracy.