Month: September 2026

🔹 Feature: Point-in-Time Restore for Windows 🔹 What It Does: Restore a Windows 11 PC to the exact state it was in at an earlier point in time using automatic local snapshots — recovery in minutes from a broken update, misconfiguration, or bad application install, without reinstalling Windows. 🕰️ Forgotten History, Still Relevant: 🔸 Classic System Restore has been sitting quietly in the Control Panel for two decades — most admins forgot it existed after “backup to the cloud” took over the conversation. 🔸 Point-in-Time Restore is the same idea, reinvented for the modern Windows era — MDM-manageable, more reliable, storage-aware, and full-system-state. The old idea, finally usable in an enterprise. What Is It Giving You: ✅ Full System State Restore in Minutes: Restore point includes the OS, apps, settings, and local files — nothing is scoped out. The device comes back exactly as it was, without a reinstall or an image redeployment. ✅ Automatic Local Snapshots via VSS: Restore points are captured in the background on a configurable schedule (default ~every 24 hours) using Volume Shadow Copy Service. Stored locally — no cloud upload, no bandwidth cost, no waiting for a backup server. ✅ Rolling 72-Hour Recovery Window: Retains … Point-in-Time Restore for WindowsRead more

🔹 Feature: Logic Apps (What’s New in Designer) 🔹 What It Does: Logic Apps Designer is now more productive. What’s New: ✅ Draft Mode With Auto-Save Every 10 Seconds: Workflows save as DRAFT first, auto-saving every 10 seconds. Refresh the browser and you land exactly where you left off — nothing goes live until you publish. ✅ Explicit Publish Button: Draft and production are now separate. Hit Publish to make your changes live. ✅ Document Your Workflows With Notes: Add inline notes to actions and branches — documentation travels with the workflow. ✅ Unified Monitoring and Run History: Design, monitor, and troubleshoot runs in one consolidated interface. ✅ New Timeline View for Better Debugging: Run history includes a horizontal Timeline showing action start times, duration, and idle gaps. ✅ Consolidated Toolbar + Unified Single View: Use a single top selector for Workflow, Code, and Run History. Vertical ellipsis (⋮) beside Publish holds the extra actions. ✅ Better for Large Workflows: Rebuilt rendering handles 50+ action workflows smoothly, with pan, zoom, and a mini-map for navigation. 🌐 https://techcommunity.microsoft.com/blog/integrationsonazureblog/moving-the-logic-apps-designer-forward/4469835

🔹 Feature: Copilot Cowork 🔹 What It Does: Delegate multi-step work to Copilot — Cowork plans it, executes it end-to-end in the cloud (even while your laptop is off), pauses for your approval when needed, waits, and then finishes the task and returns a completed result. The Shift That Actually Matters: 🔸 Copilot Chat = Assistant: You ask, it drafts, you act. Every step still requires you. 🔸 Copilot Cowork = Coworker: You delegate the outcome. It plans, retrieves context, calls tools, produces the artifacts, and stops only for genuine approval checkpoints. Then it waits. Once you approve, it resumes and completes the job. Example: “Prepare the Q3 exec review, send the draft to my VP for sign-off, and when they approve, publish it to the leadership channel and update the roadmap in ADO.” Chat drafts one email at a time. Cowork does the entire job — including the wait for the boss’s confirmation and the execution after approval. What Is It Giving You: ✅ End-to-End Task Completion: You define the work; Cowork executes start to finish and delivers finished artifacts — a deck, report, or updates in a system of record. ✅ Runs in the Cloud, Keeps Going: Tasks … Copilot CoworkRead more

Why move to v6 with NVMe The v6 generation (Dv6/Ev6 and variants) runs on newer Intel hardware and uses NVMe as the only disk interface, both for remote managed disks and for local storage. NVMe is a leaner protocol than emulated SCSI, which translates into lower storage latency and better IOPS and throughput per vCPU, backed by Azure Boost offloading. There is also a strategic reason: Azure continues to support SCSI on the VM sizes that offer it, but not every new VM series will include SCSI going forward. Preparing your Windows images and session hosts for NVMe now avoids surprises when your next resize or refresh lands on an NVMe-only size. Prepare Windows before the switch The classic advice is to make sure stornvme is a boot-start driver. That is necessary, but not sufficient. The real trap is the StartOverride registry key: when present, it overrides the Start value and prevents the driver from loading at boot. On Azure Gen2 VMs there is a second, less known culprit: vpci. The PCI bus on these VMs is not exposed through ACPI but over VMBUS, handled by vpci.sys. While the VM runs on the SCSI controller, no virtual PCI device exists, … Moving Azure Windows VMs from v5 (SCSI) to v6 (NVMe): what actually breaks and how to fix itRead more

🔹 Feature: Frontier Week Lab 🔹 What It Does: Official Microsoft hands-on lab that walks you through the full production lifecycle of AI agents on Microsoft Foundry — build, observe, evaluate, and orchestrate — using the Python SDK and the Foundry portal side by side. What Is It Giving You: ✅ Real Instrumentation, Not Toy Traces: OpenTelemetry-based GenAI tracing lands in Application Insights. Debug prompts, inspect tool calls, measure latency and token cost, replay failing conversations — the actual observability story you need in a regulated environment. ✅ Systematic Quality Evaluation: LLM-as-judge evaluations run against test datasets so you measure agent quality objectively across releases. Turns “the agent seems better after this change” into a chart. Aligned with what regulators are going to ask about AI systems, not what marketing decks show. ✅ Multi-Agent Orchestration Two Ways: Python SDK for programmatic pipelines AND the Foundry portal for visual workflow authoring. Same underlying runtime — pick the surface that fits your team. Prerequisites (worth knowing): 🔸 Azure subscription with Contributor + Foundry User role. 🔸 GitHub account (for Codespaces) OR local Python 3.10+ and az CLI. 🌐 https://github.com/microsoft/FrontierWeekHack 🌐 Docs site: https://microsoft.github.io/FrontierWeekHack/

🔹 Feature: On-Demand (Autoscale) Billing for Spark in Microsoft Fabric 🔹 What It Does: Pay-as-you-go billing model for Apache Spark in Fabric — Spark jobs run on dedicated serverless compute, billed independently, so they don’t consume CU from your Fabric capacity. Why It Matters — Fabric Capacity Plans Are Not Cheap: 🔸 Fabric capacity SKUs are billed 24/7, whether you’re running heavy Spark jobs or not. For teams whose Spark workload is bursty — a few notebook runs a day, weekend batch processing, monthly ETL — sizing capacity for the peak means paying for capacity you rarely use. 🔸 On-Demand Billing for Spark solves this: pay for Spark only when Spark actually runs. What Is It Giving You: ✅ Spark Stops Consuming Fabric Capacity CU: Spark jobs use dedicated serverless resources and leave Fabric capacity for Power BI, Warehouse, semantic models, Real-Time Intelligence, and OneLake — no more slowdowns from a heavy Spark run. ✅ Pay Only for Active Spark Compute: Billed at 0.5 CU-hour, only for active execution. No idle-time cost. ✅ Configurable Max CU Ceiling: Set a per-capacity CU limit for on-demand Spark. Keeps your top-end cost predictable. ✅ Predictable Throttling/Queueing: Hitting the max CU limit throttles interactive … On-Demand (Autoscale) Billing for Spark in Microsoft FabricRead more

🔹 Feature: Azure Infrastructure Resiliency Manager 🔹 What It Does: Unified control plane in the Azure portal for zone resiliency across your entire Azure estate — at-scale posture view, resiliency goals, simulated zone outages, and orchestrated recovery, all in one place. What Is It Giving You: ✅ At-Scale Resiliency Posture Across the Whole Azure Estate: Central overview of every resource categorized as Zone Resilient or Non-Zone Resilient. Drill into individual resources, see their zone posture, and get Azure Advisor recommendations for what to fix. Global, nonregional service — one view across every Azure region. ✅ Service Groups — Application-Centric Resiliency View: Group resources into logical applications that span subscriptions, resource groups, and tags. Get the “is THIS application zone-resilient?” answer instead of drowning in a per-resource dashboard. Exclude non-critical resources from evaluation, manually attest resources you’ve made resilient via a custom setup. ✅ Zonal Resiliency Goals + Targeted Recommendations: Set an explicit zonal resiliency goal per service group. Get targeted recommendations for how to get there, and track progress against the goal over time. Turns “we should be zone-redundant” into a measurable roadmap. ✅ Availability Zone Down Drills: Simulate a zone outage to test how your app actually behaves when … Azure Infrastructure Resiliency ManagerRead more

🔹 Feature: Described Agents in Work IQ Dev Tools (wiqd) 🔹 What It Does: Build Microsoft 365 Copilot agents by describing them. What Is It Giving You: ✅ Agents Without Writing Code: Skip building an orchestrator, hosting a runtime, or wiring an LLM. Define what the agent should do in JSON plus a natural-language instructions file — Copilot runs it. ✅ Small, Well-Defined App Package: appPackage/manifest.json (M365 app shell — id, version, icons, permissions, where the agent appears), appPackage/declarativeAgent.json (the agent itself), appPackage/instructions.txt (natural-language system prompt), plus optional API plugin manifests for actions that call REST APIs you own. Version-control the whole thing. ✅ One CLI for the Full Lifecycle: wiqd covers scaffold → edit → validate → package → provision → publish → monitor → share → evaluate. No more learning four separate tools (Teams Toolkit, M365 catalog, tenant admin, monitoring) to ship one agent. ✅ Validation Against the Current Schema: Manifest validation runs against the shipping schema — not a stale copy. Catches drift and breaking changes before you deploy. ✅ Doesn’t Replace the Underlying Platform — Just Orchestrates It: Under the hood, wiqd shells out to the right upstream tool for each step — Agents Toolkit for … Described Agents in Work IQ Dev Tools (wiqd)Read more

🔹 Feature: Generate a Lean, Custom Microsoft Graph SDK With Kiota 🔹 What wIt Does: Generates a minimal Microsoft Graph client library containing only the paths your app actually uses — dramatically smaller install footprint than the full Microsoft-published SDK. What Is It Giving You: ✅ Massive Size Reduction vs the Full SDK: The official Microsoft Graph SDK covers the entire Graph surface, but most apps only use a small slice (mail, calendar, Teams, To Do, Planner, SharePoint files). Kiota builds a client that includes just the slices you need. Perfect for mobile apps, edge functions, containers, or any environment where install size matters. ✅ Same Conventions, Same Fluent Syntax: Kiota-generated clients follow the same patterns as the Microsoft-published SDK. Calls like graphClient.Me.Todo.Lists.GetAsync() feel completely familiar. Your team won’t need to learn a new API style. ✅ Scope With –include-path / –exclude-path: Use URL prefixes at generation time to define your slice. Example: kiota generate –openapi https://aka.ms/graph/v1.0/openapi.yaml –include-path /me/todo/** –language CSharp creates a To Do–only client. Same pattern works for /me/messages/**, /teams/**, /sites/**, or any subset. ✅ Interops With Microsoft Graph Core Library: Add Microsoft.Graph.Core (or language equivalent) for cross-cutting features like PageIterator, batch requests, and large file uploads. You … Generate a Light, Custom Microsoft Graph SDK With KiotaRead more

🔹 Feature: Content App on Microsoft Fabric Apps — Sample Deployment Accelerator (New Hosting Model for Content Applications) 🔹 What It Does: Ready-to-run sample that hosts a content-driven web application entirely on Microsoft Fabric — content lives in a Lakehouse, the API is auto-generated GraphQL, and the frontend is static — with no PHP, no MySQL, no App Service, no backend server anywhere. Why This Matters — A Different Way to Host Applications: 🔸 Fabric Apps is a NEW hosting model, not just an analytics surface. Applications become first-class Fabric items alongside Lakehouses, Warehouses, and Notebooks — with the same governance, identity, and access control. 🔸 App data lives in OneLake from day one — the SAME data is instantly queryable by Power BI, Notebooks, Fabric data agents, and Warehouses. The classic “operational vs. analytical” data silo disappears. 🔸 Consumes CU from your existing Fabric capacity — no separate cloud bill for the app tier. Same monthly line item that powers the rest of the Fabric estate. 🔸 Enterprise governance inherited by default — Purview, sensitivity labels, DLP, and Entra ID RBAC apply automatically. No bolted-on compliance layer. What This Sample Is Giving You: ✅ End-to-End Pipeline: CSV data → … Content App on Microsoft Fabric Apps — Sample Deployment Accelerator (New Hosting Model for Content Applications)Read more

🔹 Feature: Legal Agent for Microsoft Word 🔹 What It Does: A first-party Microsoft agent in Word built to support time-consuming legal work — contract reviews, redlining, and clause analysis — directly in the document. What Is It Giving You: ✅ Analyzes Contracts: Understand agreements, compare versions across drafts, and cite risks directly back to the source language in the document. No copy-paste into a separate review tool. ✅ Drafts Clean Redlines: Generates precise, negotiation-ready edits with Track Changes — the format legal teams actually send back and forth. Not just “here’s a suggested rewrite” — real Word revisions the counterparty can accept or reject. ✅ Maintains History: Separates prior revisions from new proposals so the negotiation history is preserved intact. What the previous round said stays visible; what the agent proposes now is clearly delineated. ✅ Understands Nuances: Flags non-compliant clauses and recommends approved language from your internal playbooks. Playbook-driven review — not generic legal boilerplate. ✅ Secure by Design: First-party Microsoft agent that operates entirely within the Microsoft 365 trust boundary. Contracts don’t leave your tenant; existing Purview policies, sensitivity labels, and RBAC are respected. 🌐 https://techcommunity.microsoft.com/blog/microsoft365copilotblog/word-legal-agent-in-frontier/4516218

🔹 Feature: Questions and Answers Tool for Microsoft Teams Phone Agent 🔹 What It Does: Grounds the Teams Phone Agent (AI voice agent that answers phone calls on your auto-attendants and call queues) on your own documents and public web pages — so callers get natural-language answers from real organizational knowledge instead of “press 1 for sales.” What Is It Giving You: ✅ Real AI Voice Answering Real Questions: Callers ask questions in natural language on your published phone number; the Teams Phone Agent answers with knowledge sourced from your own documents. Opening hours, product FAQs, policy questions, shipping status, HR benefits — no IVR tree, no waiting for a human, no “your call is important to us.” ✅ Two Knowledge Source Types: Knowledge – Document (upload files directly) OR Knowledge – URL (point at a public web page). Mix multiple tools per agent to cover different topics. ✅ Broad Document Support: JSON, XML, PDF, TXT, CSV, DOC / DOCX, XLS / XLSX, PPTX, plus ZIP archives containing any of the supported types. Push the docs your team already writes for humans — no rebuilding content in a special “chatbot” format. ✅ Reasonable Content Limits: Up to 10 files per … Questions and Answers Tool for Microsoft Teams Phone AgentRead more

🔹 Feature: Appointment Tools for Microsoft Teams Phone Agent 🔹 What It Does: Lets the Teams Phone Agent (AI voice agent) book, update, and cancel appointments during a phone call — via Microsoft Bookings or any custom booking system with an API. What Is It Giving You: ✅ Caller Can Actually Book — Not Just Ask: The phone agent confirms available time slots, takes bookings, updates them, or cancels — all via the voice conversation. No callbacks or web redirects. ✅ Microsoft Bookings Integration Out of the Box: Point the agent at a shared Bookings page. One PowerShell call (New-CsMainlineAttendantAppointmentBookingFlow) wires the booking link into the phone agent call flow. ✅ Custom Booking System Support: Connect to any scheduling backend with an API (ServiceNow, healthcare EHR, hotel PMS). Configure in Teams admin center or via PowerShell. ✅ Caller Authentication Built In: Verify callers before touching appointments via SMS code, Email code, or User Details read-back. ✅ Four API Authentication Types: Basic, API Key, Bearer Token (Static), or Bearer Token (Dynamic) — covering everything from simple internal APIs to modern OAuth-protected endpoints. ✅ Full Appointment Lifecycle Call Flows: Generate auth token → Authenticate caller → Get available time slots → Book … Appointment Tools for Microsoft Teams Phone AgentRead more

🔹 Feature: Application Gateway for Containers (AGC) — Sample Deployment Accelerator (NGINX Ingress Controller Replacement) 🔹 What It Does: End-to-end, copy-paste-ready sample that deploys Application Gateway for Containers on AKS as the RECOMMENDED replacement for the NGINX ingress controller — with a full working app, WAF, external backend routing, and every real-world gotcha documented. Why AGC Is the Right Replacement for NGINX Ingress on AKS: 🔸 Fully managed by Azure — no ingress-controller pod to size, monitor, patch, or fail-over yourself. 🔸 Native Kubernetes Gateway API — the upstream successor to Ingress. Portable, role-oriented, and where the K8s community is going. 🔸 Multiple frontends / FQDNs are FREE in the managed model — one Gateway = one frontend, provisioned automatically. No more running one ingress controller per hostname. 🔸 Built-in Azure WAF via WAF policies — no separate ModSecurity/OWASP-CRS deployment to maintain (rules don’t port 1:1 from NGINX, but you also don’t run WAF pods). 🔸 Managed identity for backend mTLS — declarative `BackendTLSPolicy` instead of annotations and secrets. 🔸 Weight-based canary / blue-green via `HTTPRoute` — native, not custom annotations. 🔸 Higher throughput than a single NGINX ingress controller pod, no self-managed HPA on the ingress plane. What This … Application Gateway for Containers (AGC) — Sample Deployment AcceleratorRead more

🔹 Feature: AKS Spot-First with Node Auto Provisioning (NAP) — Cheap by Default, On-Demand Fallback 🔹 What It Does: Ready-to-deploy AKS pattern that schedules workloads on Spot VMs first (up to ~90% cheaper) and automatically falls back to on-demand nodes whenever Spot capacity can’t be provisioned or gets reclaimed. What Is It Giving You: ✅ Cheap AKS by Default: Every pod tries a Spot node first at Spot pricing. For dev/test, batch jobs, background processors, and anything interruption-tolerant, this cuts the AKS compute bill dramatically. ✅ Automatic On-Demand Fallback: When Azure can’t provision a Spot node (no capacity in the region, quota hit, spot price above your ceiling, or reclaim event) — Karpenter automatically provisions an on-demand node in the SAME cluster from the fallback NodePool. Workload keeps running, no manual intervention. ✅ Two Karpenter NodePools, Weight-Based Priority: `spot` pool (weight 99, tries first) and `ondemand` pool (weight 10, fallback). Weight tells NAP which pool to try first — you don’t chase resource requests through custom scripts. ✅ You Define Which VM Types Get Booted: Both NodePools let you specify exactly which VM SKUs (or families) NAP is allowed to provision — pin to a single size, widen to a … AKS — Cheap by DefaultRead more