When Capital One debuted a new agentic AI tool called Chat Concierge, the credit card giant decided to design it specifically for business customers in the auto dealership industry. The goal: make it easier for car buyers to ask questions about different vehicles and set up appointments with salespeople or schedule a test drive.
With nearly 16 million new vehicles sold annually in the U.S., this use case was right in the sweet spot of where Capital One focuses its agentic AI efforts. “We want to start off at the low end of the risk spectrum, but also find use cases with impact and enough complexity that we can learn from it,” says Prem Natarajan, head of enterprise AI at Capital One. He says that Chat Concierge has been embraced by dealers because it has dramatically increased customer engagement and is 55% more successful in converting leads into buyers.
But like most deployments of agentic AI, Chat Concierge did require a bit of massaging after it launched. Capital One kept a close eye on latency, which is the time delay between an AI system receiving an input and then generating a response. Since launch, Natarajan says, Capital One has reduced latency fivefold, an improvement he attributes to increased engagement and the company’s decision to build its own proprietary multi-agentic workflow. “We can really tune these things for latency, because we build our own stack,” adds Natarajan.
