Microsoft AI Weekly #12: Copilot, GitHub, Azure AI & more

Oct 13, 2026

👋 Hi folks!

If last week was big, this one was busy. It was all about agents "growing up” in the Microsoft AI world.

Microsoft dropped new Agent Experience playbooks, new Copilot super app features, local model tooling, and Microsoft‑Decision‑1, a fast decision‑making model built for agent workflows.

Everywhere you look, the stack is getting more agentic, smarter context, better tools, and more ways to actually ship real AI workflows.

Buckle up, there’s a lot to skim this week, and don’t miss Joanne’s Corner at the end!

⚡️ Microsoft

The latest from Microsoft on AI.

💼 Productivity & Work

Building trust in AI-powered support at Microsoft: Our Employee Self-Service agent story • David Hirning
Our Customer Zero guide: Enabling agent creation across Microsoft 365 Copilot, Copilot Studio, and Foundry • Alex Fleck
Boosting product manager productivity at Microsoft with AI • Alyse Maguire
6 New Microsoft Copilot Features You’ll Want to Try • Mike Tholfsen
The NEW Microsoft Copilot Super App • Mike Tholfsen

👩‍💻 Coding & Developer Tools

Introducing the Agent Experience (AX) Practitioner Playbook • Waldek Mastykarz
Bringing local models and sandboxed tools to Windows and GitHub Copilot • Patrick Nikoletich and Stuart Schaefer
From Spec-First to Enterprise-Ready: Extending GitHub Spec Kit • Apoorv Gupta
What is Agent Experience (AX)? • Waldek Mastykarz
ReviewBench: An open benchmark for AI code review • Michelle Zhou and Alejandro Carderera de Diego
VS Code Learn: Spring Boot with Copilot: Build, Debug, MCP & Playwright • Rory Preddy

🐙 GitHub Changelog

🏗️ AI Apps, Agents & Platform

AI transformation across the infrastructure lifecycle: From supply chain to fleet operations • Rani Borkar
Microsoft named a Leader in the 2026 Gartner® Magic Quadrant™ for Global Industrial AIoT Platforms • Michael MacKenzie
Microsoft-Decision-1: Our model for fast decision-making • Achint Srivastava
Beyond violation rates: Are your AI evals measuring the right things? • Xiawei Wang, Riccardo Fogliato, Abhinav Palia, Chang Liu, Yeming Tang, Heba Elfardy and Sandeep Atluri
Building an agentic system and learning loop around customer IP • Milena Marinova, Kanio Dimitrov, Don Hejna and Yousef Al-Kofahi
How to Use the NEW Agent Builder in Microsoft Copilot - Build an AI Agent in Minutes (2026) • Mike Tholfsen

🧩 Business Apps & Low Code

Build Azure canvases with the Canvas authoring plugin • Alex Weininger

🛡️ Security, Governance & Admin

Agent ID Effective Access: From Inventory to Permission Intelligence • David Alonso Dominguez
Correlation Isn’t the only Proof: Building a Provenance Graph for AI-Agent Investigations • David Alonso Dominguez

Copilot Studio

Are bigger agents better? How to design agents that scale • Gary Pretty
Copilot Studio Fundamentals | EP02 • April Dunnam
Creating a solution for your agent | EP03 • Microsoft Power Platform
Introduction to Agents | EP01 • April Dunnam

🌍 From the community

Hand-picked from MVPs and AI subject-matter experts around the globe.

👩‍💻 Coding & Developer Tools

Turn Meeting Transcripts into Project Plans with GitHub Copilot Agents • Microsoft Developer, Justin Garrett and Mahmoud ATALLAH

🏗️ AI Apps, Agents & Platform

When my agent is slow or wrong, where do I even look? • Microsoft Developer and April Gittens
Your Agent Is Only as Smart as Its Context • Microsoft Developer and Ayça Baş

🛡️ Security, Governance & Admin

AI Weekly: Named Copilot Models, GPT-6.1, and Advanced Reasoning • Josh Cook
A first test drive with Microsoft-Decision-1: Reviewing alert severity • Nicola Suter

Copilot Studio

Copilot Studio Hooks: Build Five Practical Customer Ops Automations • Josh Cook

Upcoming Events

.NET AI Community Standup: Build agentic experiences with the Blazor AI Components • dotnet

📢 Message Center

Microsoft 365 Message Center posts, via mc.merill.net.

💭 Joanne’s Corner

I love that Microsoft‑Decision‑1 isn’t trying to be the smartest model in the room — it’s trying to be the most decisive. And honestly, that’s refreshing. We’ve spent years obsessing over bigger models, longer context windows, and more parameters. Meanwhile, the thing that breaks most agent workflows is embarrassingly simple: the agent can’t decide what to do next.

Decision‑1 is basically Microsoft saying: “Fine, we’ll build a model that just makes good choices.” And that’s exactly what agent systems need. It’s fast, it’s structured, and it’s built for the boring‑but‑critical moments where an agent has to commit to an action. I think this model is going to unlock a whole wave of reliability improvements, the kind you only notice when things quietly stop breaking.

More thoughts next week, the agent era is just getting started.

Want to get featured on the weekly MSAI.ms? → Submit your content 😎

🪃 Acknowledgement of Country

MSAI.ms is written between Melbourne and Dubai. Part of this work is created on Wurundjeri land, and we acknowledge the traditional owners of country throughout Australia, recognising their continuing connection to land, water and community. We pay our respect to them and their cultures and to elders both past and present.

Discussion about this post

Ready for more?