PepsiCo’s blueprint for agentic AI | BRK224
Rishabh Saha shares how Microsoft and PepsiCo engineers modernized PepsiCo’s data foundation for agentic applications, using Azure SQL, Cosmos DB, PostgreSQL, and Azure Databricks. The session outlines a practical build path for agentic RAG, including Azure SQL vector indexing and semantic search to speed up repeatable app patterns.
Overview
This Microsoft Build 2026 breakout (BRK224) covers a practical approach PepsiCo used to prepare its data layer for agentic applications.
What the session covers
- Modernizing the data foundation for agentic apps using:
- Azure SQL
- Azure Cosmos DB
- PostgreSQL
- Azure Databricks
- A build path for agentic RAG architecture
- Using Azure SQL capabilities to support agentic retrieval patterns:
- Vector indexing
- Semantic search
- Patterns aimed at reducing application development cycles by making data-layer approaches more repeatable.
Session chapters (from the video description)
- Recent Azure database announcements including Cosmos DB and Horizon DB
- Challenges in preparing customer conversations and hurdles before strategizing
- Deep dive into a Data Analyst Agent architecture (authentication, query generation, enrichment)
- Reach and Persist layer with knowledge storage
- Demo segment
- Tracking Agent concepts (learning, knowledge, fact layers)
- Simplifying agentic retrieval implementation using Foundry IQ at enterprise scale
- Compounding intelligence/efficiency gains from data sharing
- Lessons learned: scope management, data quality, domain expertise