Ushering in the Era of Agentic Cloud Operations with Azure Copilot

Microsoft

Nov 18, 2025

The world of cloud operations has reached a turning point. What once offered limitless scale now brings increasing complexity. Teams are stretched across siloed tools, dashboards, and manual workflows. At the same time, the next leap in cloud innovation is here. We’re entering an era of AI agents that are making cloud operations truly intelligent.

This brings us to a frontier moment, where AI and cloud are converging to redefine operations. It ushers in a new operating model for the cloud that we call agentic cloud ops, where intelligent agents empower your cross-functional ops teams to create with speed and run the cloud with confidence.

The driving force behind agentic cloud ops is Azure Copilot, a new agentic interface that orchestrates specialized agents across the entire cloud management lifecycle, with the governance built in to support complex operations at enterprise scale.

The assistive capabilities of Azure Copilot that you rely on remain at the heart of the experience. Previously, the Copilot experience in Azure was a conversational AI assistant to help you get started, explore Azure, and get questions answered, but now, we took a step further by enabling you to work with agents which can perform chain-of-thought reasoning, plan, and execute actions on your behalf.

Azure Copilot’s innovations are focused on three priorities: agentic interface, agentic capabilities and governance. Let’s dive into each of them.

Agentic interface

The new Azure Copilot interface is a smarter, more intuitive way to work. It’s seamlessly woven into your workflows, and always there when and where you need it.

Personalized and contextual: We’ve built a fully immersive, personalized, and contextual agentic interface, embedded in your workflows, and designed to simplify your work. You can switch easily between chat, console, and CLI. A new immersive experience lets you chat with Copilot in one place and view your past conversations. You can now multi-task with Copilot starting a new chat while another prompt might be processing. The “agent” toggle in the Copilot chat allows you to use our new agents.

Right agent at the right time: Azure Copilot intelligently surfaces the right agent at the right time wherever you are in your workflow. When you work within a specific service or task, Copilot automatically invokes the agent best suited for what you’re doing. We give you the ability to dig in and see all the steps the agent took to arrive at its answer, and you can browse any artifacts Copilot procured while it processed the task.

One place for what matters most: With the new operations center dashboard, you can get a unified view across observability, resiliency, configuration, optimization and security, all from one place. The operations center also serves as a seamless starting point for your Copilot and agentic cloud operations experience.

Agentic capabilities

At Ignite, we’re launching new agents that streamline every stage of the cloud management lifecycle from ‘Day 0’ migrations and deployments to ‘Day N’ operations and optimization of IT environments. With Copilot orchestrating these agents, manual workflows give way to intelligent automation where you set the intent and agents act on your behalf.

We built these agents based on user feedback and extensive research. A recent report from Harvard Business Review confirms this trend: 52% of all agentic AI implementations are happening in IT operations. We also learned that monitoring, troubleshooting, configuration, and provisioning are among the top gen-AI use cases. This is in line with how we have been prioritizing Azure Copilot’s roadmap. Furthermore, we have focused on delivering complete, connected capabilities rather than isolated tools. Here’s a glimpse of what Azure Copilot agents can do:

Migration: Accelerate migration and modernization by automating discovery, generating AI-powered IaaS and PaaS recommendations, and using GitHub Copilot integration to refactor .NET or Java applications. Copilot helps transform inventory data into actionable blueprints for faster, smarter migrations. Try asking: “Assess my on-prem apps and recommend the best Azure modernization approach.”

Deployment: Streamline infrastructure planning and deployment with natural language guidance grounded in Azure Well Architected Framework best practices. Whether you’re setting up web apps, chatbots, or microservices, Copilot helps you generate Infrastructure as Code and validate configurations with ease. Try asking: “Deploy a Python Flask app with a PostgreSQL backend, secure secrets in Key Vault, and enable monitoring with Application Insights.”

Observability: Investigate performance issues across applications and infrastructure with AI-driven correlation of Azure Monitor metrics, logs, and traces. Copilot highlights anomalies, connects related alerts, and recommends mitigation steps all within your monitoring workflow. Try asking: “Start an investigation for my alert: [insert alert ID]” or “Can you help investigate this alert?”

Optimization: Identify cost and carbon savings opportunities, validate recommendations, and execute improvements through guided, conversational workflows. Copilot explains why changes are suggested, compares alternatives, and even generates scripts to implement them. Try asking: “Show me my top cost-saving opportunities” and “Generate PowerShell commands to apply them.”

Resiliency: Strengthen business continuity through zonal redundancy checks, data recovery validation, and ransomware protection. Copilot surfaces configuration gaps, recommends remediations, and generates ready-to-deploy scripts to meet RPO and RTO goals. Try asking: “Check which of my workloads aren’t zone-resilient and suggest fixes.”

Troubleshooting: Resolve issues faster with targeted root cause identification and guided mitigation steps across VMs, Kubernetes, and databases. Auto-mitigation or “one-click fixes” are also available for some issues and resource types. When escalation is needed, Copilot can automatically create support tickets to keep you moving. Try asking: “Diagnose connectivity issues on my VM and recommend next steps.”

Let's see the agents in action:

Governance

As agents accelerate modernization and streamline operations, one thing is critical- they must do so securely, consistently and in compliance. That’s where Copilot’s governance comes in.

Azure Copilot gives your organization the control and confidence to scale agentic operations safely. It honors any existing controls and does not invent actions or reach into data a user is not allowed to see. Instead, it follows established RBAC, Azure Policy, and compliance frameworks, and asks for explicit confirmation before making any changes.

To give enterprises complete control over chat and artifact data, we’re introducing Bring Your Own Storage (BYOS). This option strengthens trust and transparency by letting customers decide where their data lives. By default, conversation history is stored in Microsoft-managed storage with a 30-day retention period. For organizations in highly regulated environments, BYOS gives customers flexibility to provision their own storage, configure Copilot to use it, and ensure that Microsoft has no access to the data. Customers also have control over retention policies and can set retention periods to be longer or shorter than the default 30 days.

How Azure Copilot Works

Azure Copilot combines a deeply integrated orchestration engine with intelligent agents that interpret your intent, understand your context, and coordinate the right tools to drive outcomes. By blending flexible and multi-step workflows with transparent, agent-driven collaboration aligned to your permissions, Copilot makes cloud operations more intuitive, secure, and human-controlled.

1. User initiates a conversation

Everything begins with a user request. This triggers Copilot’s orchestration pipeline, which historically followed a deliberate, linear progression. The first stage is content filtering, where Copilot screens the input to ensure relevance and safety before deeper reasoning begins.

2. Copilot interprets intent and context

Copilot extracts intent and evaluates context, including which Azure portal resources or views the user is in and what RBAC permissions they have. During this stage, locale detection adjusts the interaction to the user’s language and regional preferences. This interpretation lays out the foundation for determining which tools or agents could ultimately fulfill the request.

3. Orchestrator parses intent and selects tools or agents

The orchestrator identifies the right capability by parsing Copilot’s catalog of tools and agents, which reason over a rich data lake of Azure knowledge sources including documentation, Azure Resource Manager (ARM), and Azure Resource Graph (ARG).

With the launch of agentic capabilities, Copilot can now sub-orchestrate with intelligent agents rather than simply invoking tools. These agents sustain multi-step dialogue, interpret user intent, and collaborate with one another to deliver deeper insights.

This represents an evolution from the rigid sequence (content filtering → locale detection → tool discovery → reasoning loop → responding) to a more flexible and intelligent multi-step process.

4. Agent or tool executes with RBAC and secure state

Once Copilot chooses a tool or an agent, execution happens on behalf of the user’s identity, fully honoring RBAC and existing environmental configurations.

Agents operate as extensions of the user, and every thought and action is surfaced transparently, so users can see exactly what Copilot is doing.

All conversational states, actions, and artifacts are stored securely. Copilot uses its own managed identity to interact with storage, enabling capability usage without granting users direct access. This flow also supports the BYOS option for stricter access control.

5. Human-in-the-loop review with safe execution

Before executing an update action, Copilot requests explicit user approval. This ensures human approval remains central, keeping execution safe and deliberate. The response or proposed action reflects the outcome of Copilot’s reasoning loop, which iterates through decisions and tool or agent calls before surfacing an answer or a clarification.

Throughout, Azure Copilot adheres to Microsoft’s responsible AI policies, grounding responses in the latest Azure knowledge sources, which may include ARG, documentation, support, and ensuring transparency and trust.

How to Get Started

To start using agents, your global administrator must request access to the agents preview at the tenant level in the Azure Copilot admin center.

This confirms your interest in the preview and allows us to enable access. Once approved, users will see the Agent mode toggle in Azure Copilot chat and can then start using Copilot agents. Capacity is limited, so sign up early for the best chance to participate.

Additionally, if you are interested in helping shape the future of agentic cloud ops and the role Copilot will play in it, please join our customer feedback program by filling up this form.