How Bridgewater Turned 100 Million Documents Into an Edge - Sam Green & Aaron Linsky, Bridgewater
Bridgewater built Vault to make 100M unstructured documents usable for AI-driven investment research.
“real success comes when you can lay the foundation to make agents proactive participants in your investment process, not just reactive tools for people”
Bridgewater Associates detailed Vault, an internal data platform that ingests and standardizes 100 million documents—news, transcripts, policy docs, and internal memos—for use by both human researchers and AI agents. Their core architectural insight is that humans and agents have fundamentally different data needs, yet the highest-value platform serves both simultaneously. The talk signals a broader institutional shift at Bridgewater toward agentic, proactive AI (embodied in their investment assistant 'Ayla') rather than purely query-driven tooling.