AI Agents on Brian Swiger

The End of AI Vibes: Measuring True Business ROI for Production Agents with Azure AI Foundry

Thu, 09 Jul 2026 17:47:00 -0400

Building AI agents is easy, but proving their business value is hard. This post explores Microsoft’s new ROI tracking capabilities introduced at Build 2026 (Session BRK252), detailing how organizations can move beyond basic token and latency metrics to map agent performance directly to financial and operational outcomes.

Shifting from Plumbing to Prompts: Introducing the Azure Functions Serverless Agents Runtime

Sun, 07 Jun 2026 22:14:00 -0400

Microsoft has launched the public preview of the Azure Functions serverless agents runtime, introducing a declarative, markdown-first approach to building event-driven AI agents. This runtime leverages the core scaling, billing, and integration strengths of Azure Functions while removing the infrastructure boilerplate usually associated with production agent development.

What's New at Microsoft Build 2026: Building, Deploying, and Operating Enterprise Agents at Scale

Wed, 03 Jun 2026 16:31:00 -0400

Microsoft Build 2026 introduced the Foundry Agentic Platform, a unified ecosystem that Bridges the gap between local agent prototypes and production-grade enterprise deployment. Key updates include the unification of Semantic Kernel and AutoGen into the stable Microsoft Agent Framework, managed hosted runtimes with framework interoperability, autopilot agents integrated into the corporate org chart via Entra ID, and automated closed-loop optimization using production traces.

The Age of Agentic Engineering: Orchestrating Copilot Teams with Squad and Layered SDKs

Mon, 01 Jun 2026 21:04:00 -0400

AI development is shifting from single-turn prompts to multi-agent orchestration. Discover how Microsoft’s Command Line platform highlights the future of software development through the Squad framework, durable memory architectures, and layered agentic SDKs designed for predictable enterprise workflows.

Azure AI Foundry Toolboxes: Define Your Agent Tools Once, Use Them Everywhere

Mon, 27 Apr 2026 17:43:00 -0400

As organizations scale their AI agent deployments, tool management becomes a bottleneck. Every agent wires its own tools, manages its own credentials, and duplicates integration code. Azure AI Foundry Toolboxes solve this by letting teams define tools once, manage authentication centrally, and expose everything through a single MCP-compatible endpoint that any agent framework can consume.

Now in public preview, Toolboxes support built-in tools like Web Search, Code Interpreter, and Azure AI Search, plus open protocols like MCP, A2A, and OpenAPI. Best of all, they are framework-agnostic: Microsoft Agent Framework, LangGraph, GitHub Copilot SDK, and custom agents can all consume the same toolbox.