Like large US Bank BNY armies by AI agents

The financial services industry is one of the most regulated sectors. It also manages large amounts of data. The financial companies are aware of the need for caution Slowly added generative AI and AI agents for their service stables.

The industry is not a stranger for automation. However, the use of the term “agent” was muted. And understandably many have taken one in the industry very cautious attitude towards generative AIEspecially without regulatory framework. But now Banks like JP Morgan And Bank of America debuted AI assistants.

A bank at the top of the trend is Bny. The financial service company, Eliza (named after Hamilton’s wife), founded by Alexander Hamilton, develops it into a multi-agent resource. The bank sees AI agents as valuable support for its sales employees and is more committed to its customers.

A multi-agent approach

Sarthak Pattanaik, head of Bny’s artificial intelligence hub, said Venturebeat in an interview that the bank first found out how to connect its many units so that their information can easily access.

BNY created a leading recommendation agent for his various teams. But it did more. In fact, it uses an architecture with several agents to receive suitable recommendations to the sales team.

“We have an agent who has everything (the sales team) about our customers,” said Pattanaik. “We have another agent who talks about products, all products that the bank has … from liquidity to collateral to payments, the Ministry of Finance and so on. Ultimately, we try to solve a customer requirement through the functions we have, the product functions we have. “

Pattanaik added that his agents have reduced the number of people with whom many of their customer -oriented employees have to speak to determine a good recommendation for customers. “Instead of the seller who speak with 10 different product managers, 10 different customers, 10 different segment persons, everything is now carried out by this agent.”

The agent has his sales team answered very specific questions that customers may have. For example, does the bank support foreign currencies like the Malaysian ring git if a customer wants to start a credit card in the country?

As you built it

The multi-agent recommendation functions made their debut in the Bny Eliza tool.

There are approximately 13 agents who “negotiate” to find a good product recommendation depending on the marketing segment. Pattanaik explained that the agents range from functional agents such as customer agents to segment agents who touch on structured and unstructured data. Many of the agents within Eliza have a “feeling of reasoning”.

The bank understands that its agent ecosystem is not completely agent. As Pattanaik emphasized:

Pattanaik said the bank turned Microsoft’s autogenic bring his AI agents to life.

“We started with autogenic because it is open source,” he said. “We are generally a construction company. Wherever we can use open source, we do it. “

Pattanaik said that the bank’s autogen was provided by a number of solid guardrails with which many answers from the agents can become earth and more deterministic. The bank also looked at Langchain to architect the system.

BNY built up a framework for the agent system that gives the agent a blueprint for answering inquiries. To achieve this, the company’s AI engineers worked closely with other bank departments. Pattanaik underlined that BNY has been building mission -critical platforms for years and scaled products such as its approval and collateral platforms. This deep knowledge bank was the key to helping the AI ​​engineer responsible for the Agent platform to give agents the special expertise they needed.

“Lower hallucination is a feature that always helps, compared to AI engineers driving the engine,” said Pattanaik. “Our AI engineers worked very closely with the full stacking engineers who built the mission-critical systems to cause the problem. It is about component animals in such a way that it is reusable. ”

For example, if you create a lead recommendation agent in this way, they can be developed by Bny’s different business images. It acts as a microservice “that continues to learn, combines and acts”.

Expand Eliza

With the expansion of the agent, BNY, his flagship -KI -tool, Eliza, plans to further improve. BNY published the 2024 tool, although it has been under development since 2023. Eliza has BNY employees access a marketplace of AI apps, received approved data records and search for knowledge.

According to Pattanaik, Eliza already provides a blueprint for how BNY progresses with AI agents and offers users a more advanced and intelligent service. But the bank does not want to stagnate and wants the next iteration of Eliza to be more intelligent.

“What we built with Eliza 1.0 is a representation and the learning aspect of things,” said Pattanaik. “With 2.0 we will improve the process and also ask how we build a great agent. If you think about agents, it is something that can learn to learn and reason and eventually take a few measures during this break. This is not a break and so on. This is the direction in which we go to Bau 2.0, since many things in relation to the risk teachers, the explanation, transparency, the links, etc. have to be set up before we become completely autonomous. ”