### BNY Mellon’s Strategic Implementation of Multi-Agent AI Systems **Date: February 25, 2025** BNY Mellon, a leading financial services company founded by Alexander Hamilton, is at the forefront of integrating advanced AI technologies into its operations. The bank is updating its AI tool, Eliza, to develop a multi-agent system that aims to provide valuable assistance to its sales representatives and enhance customer engagement. This move reflects the growing trend in the financial sector to leverage AI for more efficient and personalized services, despite the industry's cautious approach due to regulatory constraints and the need for data security. #### Key Events and Initiatives 1. **Development of Eliza 1.0 and 2.0:** - **Eliza 1.0:** Launched in 2024, Eliza is an AI tool that allows BNY Mellon employees to access a marketplace of AI applications, approved datasets, and insights. - **Eliza 2.0:** Currently in development, Eliza 2.0 aims to be more intelligent and autonomous. It will focus on improving the learning and reasoning capabilities of AI agents, while ensuring robust risk management, explainability, and transparency. 2. **Multi-Agent Architecture:** - BNY Mellon has created a multi-agent architecture to assist its sales team in making suitable product recommendations. This system includes about 13 agents that "negotiate" with each other to determine the best solutions for clients based on their needs and the bank's product offerings. - **Client Agent:** Contains comprehensive information about clients. - **Product Agent:** Provides details on all the bank's products, from liquidity to collateral, payments, and treasury services. - **Segment Agents:** Specialize in structured and unstructured data, tailored to different market segments. 3. **Enhanced Client Interaction:** - The multi-agent system reduces the need for salespeople to consult multiple product managers and client representatives. Instead, the agents handle the information gathering and recommendation process, allowing sales teams to focus on client interactions and specific client needs. - For example, the system can answer detailed questions such as whether the bank supports foreign currencies like the Malaysian ringgit for launching a credit card in that country. 4. **Technological Foundations:** - **Microsoft’s Autogen:** BNY Mellon chose Autogen, an open-source AI framework, to develop its AI agents. Autogen provides solid guardrails to ensure that the agents' responses are grounded and deterministic. - **LangChain:** The bank also explored LangChain to architect the system, focusing on creating a framework that allows the agents to learn, reason, and act effectively. 5. **Collaborative Development:** - BNY Mellon's AI engineers worked closely with full-stack engineers who have extensive experience in building mission-critical platforms. This collaboration was crucial in developing AI agents with specialized expertise and minimizing "hallucination" (AI-generated errors or inconsistencies). - The bank's deep bench of knowledge in areas such as clearance and collateral platforms provided a solid foundation for the AI engineers to componentize and build reusable AI systems. #### Industry Context and Challenges - **Regulatory Environment:** The financial services industry is highly regulated, which has historically made companies cautious about adopting new technologies. However, the absence of clear regulatory frameworks for generative AI has further slowed the adoption process. - **Data Management:** Financial companies manage vast amounts of data, making AI a powerful tool for data analysis and decision-making. The challenge lies in ensuring that AI systems can handle and process this data securely and accurately. - **Automation Trends:** While the industry is familiar with automation, the use of AI agents is a relatively new and evolving trend. Banks like JP Morgan and Bank of America have recently debuted AI-powered assistants, signaling a shift toward more intelligent and autonomous systems. #### Future Directions - **Autonomy and Intelligence:** BNY Mellon is committed to building more advanced AI agents that can not only learn and reason but also take actions based on client needs. The goal is to create a more autonomous system that can provide actionable insights, such as identifying potential issues or opportunities. - **Risk Management:** As the bank moves toward more autonomous AI agents, it will prioritize setting up robust risk management frameworks. This includes ensuring that the AI systems are transparent, explainable, and linked to existing risk management protocols. - **Scalability and Reusability:** The bank aims to build AI agents as microservices that can be scaled and reused across different lines of business. This approach ensures that the agents can continue to learn and adapt, providing consistent and high-quality service. #### Conclusion BNY Mellon's strategic implementation of a multi-agent AI system through its Eliza tool demonstrates a forward-thinking approach to leveraging AI in the financial sector. By combining the strengths of open-source frameworks like Autogen and LangChain with the bank's deep expertise in mission-critical systems, BNY Mellon is creating a more efficient, personalized, and intelligent service for its clients. As the bank continues to develop and refine its AI capabilities, it is setting a precedent for how financial institutions can responsibly and effectively integrate AI into their operations.