BlackRock’s Portfolio Management Group is transforming how investment teams conduct research by combining deep investment expertise with shared AI capabilities. AIROS, the AI Research Operating System, is the platform architecture behind that transformation. It separates the data, intelligence/context, and application/agent layers so teams can develop independently while sharing enterprise integration, security, and controls. The central design question is: how do you standardize enough to scale safely without centralizing every workflow and decision?

We will show how the architecture works through two research use cases. Continuous thesis monitoring shows how AIROS processes 150,000 documents each week to keep investment context current across 10,000 companies. The path from a user-built application to an investor-facing App Store shows how a controlled environment has enabled domain experts to create more than 1,000 dashboards and applications, including more than 30 promoted into the shared catalog.

Main Takeaways

• Separate the data, intelligence/context, and application/agent layers so each can evolve without requiring the entire system to change.
• Turn raw documents into fresh, permission-aware context as a shared capability instead of rebuilding data access in every application.
• Give domain experts a secure, zero-setup environment to build agents, then use their working prototypes to show engineers which data, tools, and controls are needed.
• Promote useful personal skills into tested, versioned capabilities that other teams can easily find and reuse.
• Move proven agents and applications from sandbox to production through a standard path with access controls, evaluation, monitoring, and clear ownership.