A story about Automation and better Customer Service

Challenge

A global Spanish bank renowned for its commitment to innovation and customer service, wanted to incorporate technology to enhance client experience and minimize the need for human interaction in certain processes. Based on their requirements and vision, we developed a bot designed to address various customer needs.

The bot was initially created to answer frequently asked questions (FAQs), but it was envisioned to evolve into a personal assistant capable of handling individualized transactional operations for each customer. By implementing this technology, the bank aimed to streamline problem resolution and eliminate delays associated with human interaction.

Proposed Solution

We combined natural language processing (NLP) technologies with database systems using the Genesis Cloud platform. For NLP, we utilized Google Cloud Services, including Data Flow and Google Login, among others. By integrating these technologies, the data processing tool accurately interpreted client requests and sent commands to the Genesis Pure Cloud platform to retrieve the necessary information. Once gathered, the bot responded to the client in natural language, such as "These are your transactions from the last month..."

The main objective of the project was to eliminate the need for clients to visit a branch or wait for human customer service to resolve queries, ensuring that issues were addressed and transactions completed as swiftly as possible. To achieve this, we built the infrastructure and architecture using Genesis Cloud and Google Cloud services. We developed client journeys through flowcharts, UX design, and relational experience elements to streamline the process. Once established, we implemented techniques in the PI system and natural language processing to accurately interpret user inputs and convert them into machine-readable commands for information retrieval. For testing and optimization, we utilized Google Cloud infrastructure to monitor live indicators such as interactions per second, success rates, and resolution frequencies. Additionally, algorithms were employed to identify topics the bot couldn't handle, allowing for continuous improvement.

This functionality is accessible through the bank's app, allowing users to perform personalized operations like checking account status, taking out a loan, requesting a credit card, and deferring purchases.

Results

Before our solution, the Human Takeover Rate (HTR) was 30%, meaning that 30 out of 100 requests required human intervention. After implementing our solution, this metric dropped to 4%. The Confusion Rate, which measures instances when the bot couldn't understand a request and had to escalate it to a human, also decreased to 8%. Our tool now handles 250,000 interactions per month, including loan applications and credit card acceptances, significantly enhancing efficiency and customer satisfaction.