Following this announcement, by leveraging AI solutions developed with OpenAI, Nubank will optimise both customer experiences and internal efficiency.
The initiative will offer internal enterprise knowledge search (the feature will provide employees with instant access to company insights and drive faster decisions), call center copilot, AI-powered assistance that will resolve 55% of Tier 1 inquiries and reduce chat response times by 70%, as well as fraud quality assurance, which uses GPT‑4o vision to analyse transactions and documents, streamlining fraud detection, and delivering high quality of service.
More information on the Nubank x OpenAI partnership
Powered by GPT‑4o and GPT‑4o mini, the solution was developed in order to integrate Retrieval-Augmented Generation (RAG) techniques to fetch and prioritise relevant documents, as well as deliver accurate answers without any need to navigate siloed and incomplete information repositories. The solution also leverages fine-tuned models trained on Nubank’s domain-specific content and semantic search in order to prioritize the most relevant results for user queries.
Nubank and OpenAI developed a Call Center Copilot in order to assist agents in real time by integrating the bank’s knowledge base and chat history, offering real-time multimodal capabilities like text, speech, and fast yet context-aware response times–essential for good customer support. With the help of the copilot, agent support remains available 24 hours a day, and over 45% of Nubank agents use the copilot’s key features, including the next-reply suggestions, chat summarisation, step-by-step guidance, and chat powered by AI.
In addition, Nubank also built an AI Assistant powered by GPT‑4o to answer customers directly. The Assistant was developed in order to handle up to 5 automated interactions before escalating. These solutions help Nubank resolve queries 2.3x faster with higher accuracy, maintaining exceptional service, and strong Transactional Net Promoter Scores (tNPS) as well.
GPT‑4o vision combines natural language processing with image recognition in order to enable the analysis of both text and visual data. Nubank is using this technology to scrutinise transaction records, customer communications, and submitted documents, as well as identify patterns and anomalies indicative of potential fraud.