Mastercard has launched an AI-powered assistant to streamline onboarding and help customers access new payment options faster.
The assistant leverages a large language model with Retrieval Augmented Generation (RAG) and fine-tuning to automate routine tasks and address key customer questions during onboarding.
Using a human-in-the-loop system, the tool ensures that human oversight remains integral by incorporating feedback from subject matter experts to support continuous learning and maintain response accuracy. Mastercard plans to develop several applications using this same infrastructure.
Developed in collaboration with Databricks on its Data Intelligence Platform, this AI capability features a chat-based assistant interface, source management tools and a learning system driven by human feedback. Mastercard emphasised that this product operates under the company’s AI and data governance standards.
By creating this solution internally, Mastercard aims to avoid the need to vet and procure third-party services, streamlining operations and reducing friction. This approach seeks to ensure Mastercard maintains full control over the assistant’s accuracy and data integrity.
AI is now widely utilised across the payments sector, commonly applied in areas like fraud detection and prevention, AI assistants such as this one, and analysing spending patterns.
Currently, Mastercard uses machine learning models to analyse more than 143 billion transactions each year to identify potential security and cyber threats. Additionally, the company leverages AI to provide crucial insights to ecosystem members and improve access to credit and financial management.
The UK has recently been grappling with a rise in authorised push payment (APP) fraud. In response to what some experts are describing as a crisis, Mastercard updated its AI-powered Consumer Fraud Risk (CFR) solution in September.
