Useful information
Prime News delivers timely, accurate news and insights on global events, politics, business, and technology
Join the reliable event by business leaders for almost two decades. VB Transform brings together people who build a strategy of real business. Get more information
Inspiration can come from different places, even for the architecture and design of agent systems.
In VB transformation, Capital one He explained how he built his agent platform for his automatic business. Milind Naphade, senior vice president of technology and head of foundations of AI in Capital One, said during VB that the company wanted its agents to function in a similar way to human agents, since their problem solves together with customers.
Naphade said Capital One began designing his agent offers 15 months ago, “before the agent became a fashion word.” For Capital One, it was crucial that, when building their agents systems, learn how their human agents request information from customers to identify their problems.
Capital One also sought another source of organizational structure for its agents: in itself.
“We are inspired by how Capital One works in itself,” said Naphade. “Within Capital One, as I am sure of other financial services, you must administer the risk, and then there are other entities that also observes, evaluates, questions and audit.”
>> See all our transformation coverage 2025 here <This same structure applies to agents that Capital One wants to monitor. They created an agent that evaluates existing agents, which was trained in the policies and regulations of Capital One. This evaluation agent can reverse the process if it detects a problem. Naphade said that thinking about it as “a team of experts where each of them has a different experience and joins to solve a problem.”
Financial Services Organizations recognize the potential of agents to provide their human agents with information to solve customer problems, manage customer service and attract more people to their products. Other banks like Bny They have deployed agents this year.
Car concessionaire agents
Capital One deployed agents in their automotive business to help bank’s concessionaires clients help their clients find the right loan for cars and cars. Consumers can look at the inventories of dealership vehicles that are ready for the test units. Naphade said his concessionaire customers reported an improvement of 55% in metrics, such as commitment and potential customers of serious sales.
“They are able to generate much better potential customers through this more conversational natural conversation,” he said. “They can have agents 24 hours a day, 7 days a week, and if the car breaks at midnight, the chat is there for you.”
Naphade said Capital One would love to take this type of agent to his travel business, especially for his customer -oriented commitments. Capital One, who opened a new room at the JFK airport in New York, offers a very popular credit card for travel points. However, Naphade said the bank needs to perform extensive internal tests.
Data and models for banking agents
Like many companies, Capital One has many data for its AI systems, but you have to discover the best way to take that context to its agents. You also have to experiment with the best model architecture for your agents.
The team of researchers, engineers and data scientists applied to Naphade and Capital One used methods such as distillation of the most efficient architectures model.
“The comprehensive agent is most of our cost because that is the one who has to disambiguate,” he said. “It is a bigger model, so we try to distribute it and get a lot of money for our dollar. Then there is also a prediction of multiple token and pre-relation added, many interesting ways in which we can optimize this.”
In terms of data, Naphade said his team had undergone several “iterations of experimentation, proof, evaluation, human in the loop and all the right railings” before releasing their AI applications.
“But one of the biggest challenges we faced was that we had no precedent. We couldn’t go and say, Oh, someone else did this way, so we couldn’t ask how it worked for them. Naphade said.
Editor’s note: As gratitude to our readers, we have opened the early record for VB transform 2026, only $ 200. This is where the ambition of AI is found with operational reality, and you will want to be in the room. Reserve your place now.
