Capital One builds agentic AI to supercharge auto sales

At the VB Transform conference, Capital One showcased its cutting-edge agentic platform designed to enhance its auto business operations. Milind Naphade, Senior Vice President of Technology and Head of AI Foundations at the financial giant, emphasized the goal of creating AI agents that emulate human problem-solving capabilities when assisting customers. Naphade revealed that the initiative began 15 months ago, well before the term 'agentic' gained popularity. He explained that understanding how human agents interact with customers was pivotal in designing these AI systems. Capital One's internal operations also served as a blueprint for the agentic framework. "We drew inspiration from our own processes at Capital One," Naphade remarked. He highlighted the importance of risk management and the need for continuous evaluation, which are critical elements in the finance sector. The newly developed AI agents are not only designed to assist customers but also to monitor their own performance. An evaluator agent, trained on Capital One’s policies, can flag issues in real-time, ensuring that the customer service experience remains seamless and effective. Naphade likened the system to a collaborative team of experts, each contributing their unique skills to tackle customer challenges. The potential of these AI agents to support human staff in resolving customer inquiries and enhancing service delivery is already being recognized across the financial industry. Other institutions, like BNY, have also integrated similar technologies this year. With the deployment of these agents, Capital One's auto division is now helping dealership clients connect with customers seeking both vehicles and financing options, resulting in a remarkable 55% increase in engagement and quality leads. Naphade noted that the conversational nature of the AI interactions has significantly improved customer engagement. He mentioned the convenience of having these agents available around the clock, providing assistance even in the middle of the night. Looking ahead, Capital One is eager to extend this technology into its travel sector, particularly for direct customer interactions. Although the bank recently opened a new lounge at JFK Airport and offers a popular travel rewards card, Naphade stressed the necessity of thorough internal testing before rolling out similar AI solutions. Despite having vast amounts of data available for training its AI systems, Capital One is focused on ensuring that the context is appropriately integrated into its agents. Naphade and his team of researchers and engineers are exploring various model architectures to optimize performance. Through a series of experiments and evaluations, the team has navigated numerous challenges in developing these AI applications. Naphade pointed out the lack of industry precedents, making it difficult to benchmark their efforts against others. The journey has been one of continuous learning and adaptation, laying the groundwork for a future where AI enhances customer experience in the automotive sector and beyond.