Building AI-ready data: Vanguard’s Virtual Analyst journey

2026-04-29 · Source: Artificial Intelligence · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Data Science & Analytics, Cloud Computing & IT Infrastructure · Depth: Intermediate, medium

Summary

Vanguard, a global investment management firm, developed a "Virtual Analyst" solution to provide financial analysts with immediate access to complex datasets, reducing query response times from days to minutes. This initiative revealed that successful conversational AI deployment hinges on "AI-ready data infrastructure" rather than solely on advanced foundation models. The solution, powered by AWS services like Amazon Bedrock, Amazon Redshift, and AWS Glue, involved a cross-functional collaboration among data engineers, business analysts, compliance, security, and business stakeholders. Vanguard identified eight guiding principles for AI-ready data, including establishing clear data product and operating models, defining governance, building unified metadata catalogs, implementing a semantic layer, developing ground truth examples, automating data quality checks, establishing change control, and creating continuous evaluation mechanisms. This approach significantly improved data access, accuracy of AI-generated SQL queries, and reduced data team workload.

Key takeaway

For Directors of AI/ML evaluating conversational AI solutions, prioritize building robust AI-ready data foundations over solely focusing on foundation model capabilities. Your teams should establish clear data product ownership, implement unified metadata and semantic layers, and integrate continuous data quality and evaluation mechanisms. This strategic focus will ensure reliable, accurate AI outputs and significantly reduce time-to-insight for business users, transforming data access across your organization.

Key insights

AI-ready data infrastructure, not just advanced models, is crucial for successful enterprise conversational AI deployment.

Principles

• Assign clear ownership for data products.
• Unify technical and business metadata.
• Operationalize business metadata via a semantic layer.

Method

Vanguard's method involved cross-functional collaboration, defining eight guiding principles for AI-ready data, and leveraging AWS services to build a Virtual Analyst solution that translates natural language queries into SQL.

In practice

• Document top 20 metrics with definitions and calculations.
• Create 20-30 question-to-SQL ground truth examples.
• Store data definitions in version control like Git.

Topics

• Vanguard Virtual Analyst
• AI-Ready Data
• Conversational AI
• AWS Data Services
• Metadata Catalog

Best for: AI Architect, Data Engineer, Director of AI/ML

Related on AIssential

• Companies Winning with AI Built the Data Layer First — A robust data foundation is critical for unlocking scalable AI and advanced analytics capabilities.
• From Context to Semantics: How Metadata Powers Agentic AI — Metadata platforms are evolving into semantic context layers crucial for precise AI and agentic system outcomes.
• D&B's database of 642 million businesses was built for humans, not AI agents. So they rebuilt it. — Enterprise data systems built for humans require fundamental redesign for AI agents.
• Reflections on Anthropic’s Self-Service Data Analytics Playbook — Data infrastructure, not AI, is the bottleneck for self-service analytics, requiring robust data foundations and governance.
• AI-Ready Data vs. Analytics-Ready Data — Analytics-ready and AI-ready data serve distinct purposes and require independent optimization paths.
• Building Agent Ready Data Pipelines From Scratch — AI agents require data pipelines to provide rich metadata for accurate interpretation, beyond traditional data quality.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Artificial Intelligence.