How Capital One Built Multi-Agent AI Workflows to Power Enterprise Use Cases | VB Transform 2025

Capital One's production multi-agent AI system coordinates specialized agents for data retrieval, analysis, and action execution using a proprietary multi-agentic conversational AI workflow that prevents cascading failures between agents. Milind Naphade, SVP, Technology, AI Foundations, will detail their technical architecture choices—including agent communication protocols, context sharing mechanisms, and granular permission controls—and why custom orchestration was the chosen approach. This talk will also include approaches for balancing risk management and safety with innovation in agentic systems, as well as core considerations regarding data and model selection. Take away best practices and lessons learned from real-world experiments and applications for deploying and scaling an agentic workflow.