Summary of "How BNY Mellon using GenAI to help find the right information"

The video discusses how BNY Mellon is leveraging Generative AI (GenAI) to improve information retrieval for its large workforce of 50,000 employees across multiple countries and departments. The main focus is on a Large Language Model (LLM)-based virtual assistant designed to help employees quickly find precise answers from vast and complex internal data, moving away from traditional document search methods.

Main Financial Strategies and Business Trends:

Digital Transformation with AI: BNY Mellon integrates AI tools, specifically Generative AI and conversational AI, to enhance employee productivity and streamline access to internal knowledge.
Scaling AI Solutions in Large Organizations: Starting from small pilots within specific departments (e.g., People Experience, Risk), BNY Mellon expanded the AI assistant bank-wide.
Improving Knowledge Management: The project highlighted the need to improve how documents are written, tagged, and stored, emphasizing metadata and content curation for AI-readability.
Localization and Context Awareness: The AI assistant incorporates user context such as geographic location to provide relevant policy information, demonstrating personalized AI application in financial services.
Continuous Feedback and Improvement: The bank is developing AI-driven feedback loops to monitor and enhance the virtual assistant’s accuracy, relevance, and bias mitigation.
Hybrid Cloud and AI Infrastructure: Collaboration with Google Cloud’s Vertex AI ecosystem enabled rapid experimentation and deployment, reflecting a trend towards cloud-based AI solutions in finance.

Methodology / Step-by-Step Approach:

Pilot and Scale:
- Start with a small pilot in one department.
- Expand incrementally to other departments.
- Eventually roll out across the entire organization.
Data Handling and Document Processing:
- Initially apply chunking strategies to ingest documents.
- Adjust chunking and document processing techniques based on document complexity.
- Use specialized tools like Google Document AI for complex documents.
Contextualization:
- Incorporate user metadata (e.g., location) to tailor responses.
Feedback and Metrics:
- Collect manual feedback on assistant performance.
- Plan to implement AI-based monitoring for performance, context accuracy, and bias.
- Track knowledge source effectiveness and areas for improvement.
Content Improvement:
- Work with content owners to improve document tagging and metadata.
- Encourage better document curation to enhance AI understanding.

Presenters / Sources:

Boris Tarka – Customer Engineer and Technical Lead for AI/ML in Financial Services.
Anil Valala – Head of Intelligent Automation at Bank of New York Mellon.

This video highlights how a major financial institution operationalizes Generative AI for enterprise knowledge management, emphasizing practical challenges, iterative development, and strategic collaboration with cloud AI providers.