Case Study: Automated Fraud Detection

Overview

A major credit card company wanted to improve their fraud detection process and reduce the time it takes to deal with false positives and improve customer satisfaction.

Previously, a complicated fraud-detection process required an offshore team to continuously monitor data feeds and handle customer cases, often with a lack of context to inform their calls. In 20% of cases, the offshore team phoned the customer directly to request additional information, which resulted in expensive security steps, language barriers, and, in some cases, lost customers.

How did Heron approach this?

We developed an automated decisioning model which was integrated with the client’s current RPA system in order to create an end-to-end system capable of reviewing and flagging fraudulent transactions. The tool incorporated a full audit trail to empower fraud agents to have meaningful interactions with customers when dealing with fraud cases.

Through modelling the best practice approach to identifying fraud, the automated fraud detection tool could process over half-a-million transactions per minute, leading to a reduction of back office costs by 60%.

The combination of NLP and automated decision-making is able to process both simple queries and more complex cases, and defers to human intervention where necessary. The sales team can now focus on making sales and increasing revenue for the company.

Technologies Used

  • NLP/NLU Chat interface
  • Automated decision making

The Outcome

The new AI-powered tool allows the process to be brought onshore, where a live agent can consult with chatbot in-built with an audit trail, meaning they can explain in their phone calls to customers why the transaction was flagged and blocked. The outcome is an employee better informed by a chatbot, and a less frustrated customer.