What challenges did BNY Mellon face before AI implementation?

Managing extensive AI agent deployments for improved process automation in banking operations. Reducing the number of people client-facing employees must speak to when determining a suitable product recommendation.

What was the impact of the AI solution?

Reduced the number of people salespeople need to speak to from 10 different product managers, client people, and segment people to the agent system. Improved recommendation support and decision-making for sales staff. Created a reusable microservice-like agent that can continue to learn, reason and act.

Source

MicrosoftEvidence: Low40/100

BNY Mellon Bank Manages Armies of AI Agents for Financial Services Automation

Use case typeSales enablementUpdated Feb 25, 2025

BNY Mellon is updating its AI tool Eliza into a multi-agent resource for sales representatives and customer engagement. The bank uses a multi-agent architecture with about 13 agents that negotiate with each other to produce product recommendations based on client, segment, and product data. BNY built a framework around the agentic system and used AutoGen and LangChain to ground responses and make the agents more deterministic.

Organization: BNY Mellon
Industry: Finance
Location: United States
Published: February 2025

Reported outcomes

Strategic outcomes

Speed & agilityReduced internal consultation for recommendationsBetter decisions & insightImproved recommendation support for sales staffInnovation & cultureCreated a reusable agentic microservice

Why do we believe this deployment?Customer identity, provider attribution, maturity, and source checks

Customer: BNY Mellon
Provider: Microsoft
Maturity: Unknown
Linked source: VentureBeat

No explicit deployment-stage evidence found.

Customer identity supportedSource describes one deploymentMaturity evidence evaluated

Primary read

Use case focus

Showing 3 of 3

1Sales enablement
2Decision support
3Agent orchestration

Managing extensive AI agent deployments for improved process automation in banking operations.
Reducing the number of people client-facing employees must speak to when determining a suitable product recommendation.

Created a lead recommendation agent and expanded it into a multi-agent architecture inside Eliza.
Agents access client information, product information, and structured and unstructured data, and negotiate to produce recommendations.
Built guardrails and a blueprint for responses using open-source AutoGen and LangChain, with AI engineers working closely with full-stack engineers from mission-critical systems.

Technologies

Microsoft Azure Microsoft 365 Copilot AutoGen LangChain

Reduced the number of people salespeople need to speak to from 10 different product managers, client people, and segment people to the agent system.
Improved recommendation support and decision-making for sales staff.
Created a reusable microservice-like agent that can continue to learn, reason and act.

Architecture

BNY created a multi-agent architecture inside its Eliza tool, with about 13 agents that negotiate with each other to determine product recommendations based on client, segment, and product data. The bank built a framework around the agents, used AutoGen to provide guardrails and deterministic grounding, and looked at LangChain to architect the system. AI engineers worked closely with full-stack engineers from mission-critical systems, and the bank describes the agents as a reusable microservice-like capability that can learn, reason and act.

Sources & evidence1

Evidence: Low40/100Evidence strength