Vanguard builds Virtual Analyst conversational AI solution by establishing AI-ready data infrastructure with eight guiding principles

Vanguard developed Virtual Analyst, a conversational AI tool that allows financial analysts to query complex datasets through natural language instead of writing SQL queries, addressing a workflow where typical requests previously took several days to fulfill.

The solution uses Amazon Bedrock for foundation models and natural language understanding, Amazon Bedrock Guardrails to secure sensitive financial data, Amazon Redshift for data warehousing, AWS Glue for data cataloging, and other AWS services for compute and storage. Success required cross-functional collaboration among data engineers, business analysts, compliance officers, security teams, and business stakeholders to establish clear data ownership, semantic definitions, and quality standards.

Vanguard identified eight guiding principles for AI-ready data: establishing clear data product and operating models with defined service-level agreements; defining governance and security measures with role-based access controls; building a unified metadata catalog that combines technical and business context; and implementing a semantic layer to operationalize business metadata—principles that the organization found benefited use cases beyond the initial Virtual Analyst project.