All Intune, AI and Azure Blog Posts

Guides, code and lessons from working with Intune, AI agents and Azure. Find your topic with search or the filters below.

Sep 24, 2026AI

The Enterprise Agent Control Plane I Want

I do not want another dashboard for enterprise AI agents. I want an enterprise agent control plane that can answer who owns an agent, what it can call, how it behaves and how to stop it. In this blog post I explain the five layers I would connect: identity, catalog, policy, telemetry and lifecycle. Microsoft Foundry now has a Control Plane experience for visibility and lifecycle operations across agents. That is an important platform building block. My broader point is that an enterprise control plane is also an operating model. It has to connect the AI platform with Microsoft Entra ID, Azure API Management, monitoring, data governance and the teams that own the affected systems.

Read article

Sep 22, 2026AI

Microsoft Foundry Python Agents: Build a Policy Assistant

Microsoft Foundry Python agents become easier to understand when you separate three steps: call a model, create an agent, and run that agent against a real question. In this episode, I walk through those steps in VS Code using an IT policy assistant as the example. The useful part is not getting another chatbot to answer “What is MFA?” It is moving from that simple connectivity check to an agent that can look up company policy and return sources. Watch the complete walkthrough below; the written notes explain the boundaries between the scripts and the checks I would make before using the same pattern in an application. Related Python examples: You can find model calls, agents and Azure AI Search examples in my Microsoft Foundry examples repository on GitHub. This is a broader collection, not an exact copy of the three scripts in this video. Some agent examples use the classic Agent Service API, so follow the repository README for the matching SDK and setup. Video language: German · Duration: 10:05 · Released 22 September 2026.

Read article

Sep 17, 2026AI

Five Gates Before an MCP Tool Reaches Production

Connecting a Model Context Protocol server to an agent is easy. The difficult part starts when that connection can reach a real business system. For MCP tool production, I want five gates to pass before the first user can rely on it: identity, scope, limits, approval and evidence. In this blog post I explain what I check at each gate and where Azure API Management can help. This is not a generic zero-trust checklist. It is the practical review I would use for an MCP tool that reads or changes enterprise data. The goal is simple: every call should have an accountable caller, a narrow permission, a controlled impact and enough evidence to explain what happened.

Read article

Copilot vs Cowork vs agents: Jannik Reinhard with the new Copilot Studio experience, licenses, costs and governance

Sep 16, 2026AI

Copilot vs Cowork vs Agents: Licenses, Costs and Governance

You want AI to prepare a rollout meeting. Should you ask Copilot, delegate the work to Cowork, or build an agent? All three might help. But you would be buying, operating and governing three different things. One approach may need only a prompt. Another needs a consumption budget. A third becomes a service that somebody has to maintain. In this Copilot vs Cowork vs agents guide, I explain how I would make that decision. We will look at each option, its licenses and charges, and the governance boundaries. I also walk through the new Copilot Studio interface with real product screenshots. Scope and date: commercial Microsoft 365, checked on 16 September 2026. This is not a comparison with consumer Copilot, Claude Cowork or the GitHub coding subscription. Prices below are US public list prices, excluding tax, base subscriptions and contractual discounts. Availability varies by region, cloud, license and rollout.

Read article

Sep 10, 2026AI

Azure API Management MCP: The Control Plane for Agent Tools

MCP makes it easy to connect an AI agent to tools. That is useful for a demo, but an enterprise needs more than a connection string. It needs identity, a controlled tool surface, rate limits, monitoring and a clear owner. In this blog post I show how I would use Azure API Management MCP as the control plane between agents and enterprise tools.

Read article

Sep 3, 2026Intune

Detect and Block Shadow AI with Intune: OpenClaw in Practice

Local AI agents are moving from developer experiments to normal Windows endpoints. They can read files, call tools and act with the permissions of the signed-in user. That makes them useful, but it also creates a new blind spot for endpoint teams. In this blog post I show how the new Intune Shadow AI controls can discover local agents such as OpenClaw, give you useful inventory data and help you decide what to control.

Read article

Aug 27, 2026AI

Work IQ API: Query Microsoft 365 Data With AI

Your Microsoft 365 tenant already contains the context that many AI agents are missing: meetings, email, Teams conversations, documents, people and decisions. The hard part is not generating another answer. The hard part is giving an agent the right work context without building a second data platform or bypassing Microsoft 365 permissions. This is exactly where the Work IQ API becomes interesting. Microsoft currently provides Work IQ as a public preview across CLI, MCP, A2A and REST experiences. In this guide I focus on the path that is easiest to test: install the official package, connect the MCP server and ask useful questions against your own Microsoft 365 context. I also cover the part that matters in an enterprise: tenant enablement, user permissions, tool boundaries and a rollout that starts read-only.

Read article

Aug 26, 2026AI

Microsoft Foundry Tracing and Evaluation: Debug an Agent

Microsoft Foundry tracing and evaluation become much more useful when something goes wrong. In this episode, my agent has a web-search tool configured, but the first requests do not use it. That gives us a real debugging case instead of a perfectly rehearsed answer. I inspect the trace, run an evaluation, compare optimization candidates and then look at the telemetry in Application Insights. The important lesson is that an attractive overall score can still hide failure in the part of the task that matters most. Video language: German · Duration: 20:04. Video originally published 26 August 2026; this written companion was added on 22 September 2026.

Read article

Aug 25, 2026Automation

Desk Setup 2026: Every Product I Use and What I Would Skip

After my post about the Oakywood Standing Desk Pro, many of you asked what else is on my desk. This is my complete desk setup 2026: the hardware I use every day, why I chose it and what I would not buy again. This is not a list of products that look good in a photo. My desk is where I write, code, record videos, join calls and prepare community sessions. Every item must remove friction from that work. I will also explain where I would start if I had to build the setup again with a smaller budget. Sponsored-product disclosure: Oakywood provided the Standing Desk Pro and the Oakywood accessories shown in this article free of charge as part of a sponsored collaboration. I did not pay for these products. Oakywood did not approve or edit my verdicts, and all opinions and trade-offs are my own. I bought every non-Oakywood product in this article myself.

Read article

Aug 20, 2026AI

Fabric Data Agents: Chat With Your Data in OneLake

Every company has the same gap: the data sits in lakehouses and semantic models, and the people with questions cannot write SQL or DAX. In this blog post I explain what a Fabric data agent is and how you build one. A Fabric data agent is the conversational analytics piece of Microsoft Fabric — your colleagues ask questions in plain English, and the agent generates and runs the queries against your governed data in OneLake.

Read article

Aug 19, 2026AI

Microsoft Foundry Landing Zone: Governance in the Portal

A Microsoft Foundry landing zone starts with a question that the playground cannot answer: who owns the services around the agent? Someone needs to decide how identities, networks, shared connections, monitoring and deployment policies fit together. In this episode, I move between the Foundry and Azure portals to look at those decisions. It is an architecture and configuration walkthrough, not an end-to-end deployment of a new landing zone. Watch the episode first, then use the notes as a review guide for your own environment. Video language: German · Duration: 11:48. Video originally published 19 August 2026; this written companion was added on 22 September 2026.

Read article

Aug 18, 2026AI

Agent Skills vs MCP: When to Use Which for AI Agents

In February I wrote about why CLI tools are beating MCP for AI agents. It became my most read post so far, and one follow-up question came up again and again: "Okay, and where do Agent Skills fit in?" That is a fair question, because Skills vs MCP is not an either-or decision. They solve different problems, and they work well together. In this blog post I explain the difference, compare the context cost, and give you a simple decision guide for when to use which.

Microsoft Intune & AI Insights

All Intune, AI and Azure Blog Posts

The Enterprise Agent Control Plane I Want

Microsoft Foundry Python Agents: Build a Policy Assistant

Five Gates Before an MCP Tool Reaches Production

Copilot vs Cowork vs Agents: Licenses, Costs and Governance

Azure API Management MCP: The Control Plane for Agent Tools

Detect and Block Shadow AI with Intune: OpenClaw in Practice

Work IQ API: Query Microsoft 365 Data With AI

Microsoft Foundry Tracing and Evaluation: Debug an Agent

Desk Setup 2026: Every Product I Use and What I Would Skip

Fabric Data Agents: Chat With Your Data in OneLake

Microsoft Foundry Landing Zone: Governance in the Portal

Agent Skills vs MCP: When to Use Which for AI Agents

All articles by year

2026

2025

2024

2023

2022

2021

Instruction pages