From rows to reasoning: Designing databases for AI apps and agents | BRK223

Charles Feddersen and Abe Omorogbe explain how AI apps and agents change database design, focusing on reasoning over operational data instead of only transactions. They demo new capabilities across Azure SQL Database, Azure Cosmos DB, and Azure HorizonDB (cloud-native PostgreSQL) to simplify architectures and reduce latency.

Overview

This Microsoft Build 2026 breakout (BRK223) focuses on database and data-platform patterns for AI applications and agentic systems, where the data layer needs to support reasoning workflows (for example, agent memory and retrieval) rather than only classic OLTP transactions.

The session highlights recent innovations in:

• Azure Cosmos DB (including an agent memory toolkit and a Python-based demo workflow)
• Azure SQL Database (including a demo using an Azure SQL container)
• Azure HorizonDB, described as Azure’s new cloud-native PostgreSQL service, with a deep dive into architecture, AI functions, and durable pipelines

What the session covers (from the published agenda)

Cosmos DB: agent memory and a Python demo

• Introduction to an Agent Memory Toolkit concept for Cosmos DB
• Demo using the Cosmos DB Python SDK
• A workflow that covers memory storage and processing

Azure SQL: local + cloud integration and a container-based demo

• Transition from Cosmos DB to SQL scenarios
• Discussion of integrating local and cloud database environments
• Demo building a “live site support system” using an Azure SQL container

HorizonDB (PostgreSQL): architecture and in-database AI

• Introduction to Azure HorizonDB and Microsoft’s PostgreSQL direction
• Deep dive into:
  • Service architecture
  • AI functions
  • Durable pipelines
• Tooling notes including:
  • VS Code integration
  • AI-assisted query debugging

Resources

• Session resources: https://aka.ms/build26/BRK223
• Microsoft Build sessions catalog: https://build.microsoft.com