Mastercard’s new in-house AI capabilities include a chat-based assistant, source management, and learning improvements with human feedback.
Developed in collaboration with Databricks, Mastercard’s new AI solutions help to build and deploy knowledge agent tools. The first application of this technology is a digital assistant designed to streamline customer onboarding, enabling faster, more efficient access to new payment options.
The program ensures Mastercard’s AI agents are continuously trained on proprietary data within strict governance standards, eliminating the need for third-party solutions. The onboarding assistant automates tasks and answers key customer questions using a large language model with Retrieval Augmented Generation (RAG), which pulls accurate information from Mastercard’s documentation.
The tool uses a human-in-the-loop approach, continuously integrating expert feedback to improve learning and response accuracy. This approach involves human intervention or correction during machine learning processes to provide feedback and review existing outcomes. A combination of human expertise with automated systems allows humans to refine, supervise, and improve AI outputs. It ensures greater accuracy, ongoing learning, and adaptability in the AI system, especially in complex or high-stakes tasks where machine-only processing may not be sufficiently reliable.
Eventually, Mastercard plans to expand the range of applications built on the new infrastructure beyond the digital assistant tool.
The company leverages AI as one of the critical technologies to achieve its mission. Thus, Mastercard uses machine learning to analyse over 143 billion transactions annually, detecting security and cyber threats to protect individuals and businesses. For example, the company identifies compromised payment cards with a combination of generative artificial intelligence and graph technology and utilises AI to combat crypto fraud.
AI also supports its ecosystem members with insights and enhances access to credit, helping financially underserved individuals build credit history.
