Security isn’t a feature you bolt on at the end. This capstone post ties together every security capability from the series into a six-layer reference architecture for securing Azure AI applications, with a practitioner’s checklist you can take into your next project.More
Securing the Agentic Era: Microsoft Entra Agent ID and Zero Trust for AI
As organisations deploy dozens of AI agents, the question “who authorised that agent to do that?” becomes critical. Microsoft Entra Agent ID brings the same Zero Trust identity controls we use for humans to AI agents, and Microsoft Agent 365 wraps it all in an enterprise governance control plane.More
Azure AI Content Safety for Agents: Task Adherence, Prompt Shields, and PII Filters
Azure AI Content Safety now goes beyond filtering harmful text. With Task Adherence, Prompt Shields Spotlighting, and PII detection, you can build a layered safety architecture purpose-built for the agentic era.More
Agentic RAG on Azure: AI Search Knowledge Bases and Foundry IQ
Traditional RAG fires a single query at a vector index and hopes for the best. Agentic RAG breaks complex questions into sub-queries, runs them in parallel, synthesises answers with citations, and enforces document-level permissions. Here’s how Azure AI Search’s new knowledge bases and Foundry IQ make it work.More
Azure Content Understanding GA: One Service to Replace Them All
Azure Content Understanding went GA at Ignite 2025, unifying document, image, audio, and video processing into a single API. Here’s what it replaces, how the new analyser model works, and why this changes the game for document-heavy workflows.More
Model Router and Provisioned Spillover: Cost Optimisation Patterns for Azure OpenAI
Azure OpenAI at scale gets expensive fast. Model Router, Provisioned Spillover, and Stored Completions change the cost optimisation playbook, and in this post I walk through how to combine all three into a tiered deployment architecture that saves money without sacrificing quality.More
Multi-Agent Architectures on Azure: Connected Agents, MCP, and Hosted Agents
Azure now offers three distinct multi-agent patterns on one platform. Here’s a practical breakdown of Connected Agents, MCP tool integration, and Hosted Agents, plus when to use each one.More
Computer Use Comes to Azure: Browser Automation, Desktop Agents, and the RPA Replacement
Microsoft shipped three layers of “computer use” to Azure in 2025: a vision model that reads screenshots and clicks buttons, a Browser Automation tool running Playwright in isolated sessions, and a full Computer Use tool for desktop apps. Here’s how each layer works, when to use which, and why your RPA vendor should be nervous.More
Evaluating and Red-Teaming Your AI Agents on Azure
Shipping an AI agent without evaluation is like deploying code without tests. Here’s how to use Azure’s new agentic evaluators, the AI Red Teaming Agent, and Defender for Foundry to build a proper CI/CD safety pipeline for your agents.More
Building Your First Agent on Microsoft Foundry: From Zero to Production
The Foundry Agent Service went GA at Build 2025. Here’s a hands-on walkthrough of building, deploying, and monitoring your first AI agent with tools, tracing, and the .NET SDK.More
James Westall 