AI is fundamentally changing what IT teams operate in the cloud and how they operate it. As organizations put more AI agents to work, IT teams face two challenges. They need to manage workloads that behave differently from traditional applications while finding better ways to keep systems they already run reliable and efficient. We need to operate agents, and we need agents to help us operate.
That’s the thinking behind agentic operations on Azure. Teams can operate on Azure by connecting their own preferred agents through Model Context Protocol (MCP), by using Azure’s own built-in agents, or, by using and extending Azure SRE Agent to external systems of record. Each path builds on the Azure’s operational context, observability, management, and governance so agents can investigate issues and act within the permissions and guardrails teams already define.
AI is changing demands on operations
Three things are driving the need for a different approach. First, agents don’t behave like traditional automation. Their runtime decisions make behavior and dependencies harder to predict, requiring continuous evaluation across varied inputs to maintain quality, reliability, and performance while controlling costs. Second, operational signals are fragmented. Connecting telemetry, health, configuration, and recent changes across systems takes time, especially during an incident. Third, expertise is siloed across teams. Resolving issues requires application, infrastructure, and production knowledge. Adding another tool won’t solve this. We need to bring operational context and expertise together so teams and agents can act on what they know.
What agentic operations means in Azure
In Azure, we’re building on the resource and management foundation customers already use. Azure Resource Manager (ARM) and Azure Resource Graph (ARG) provide information about resources, their relationships, and changes across the estate. The ARM MCP Server makes that foundation to MCP-compatible agents so they can inspect and manage resources through the existing control plane.
Giving an agent access to tools and operational data doesn’t, by itself, make it a better operator. The critical ingredient is context. While an alert can tell an agent that a service is failing, understanding the service’s dependencies, configuration, and recent changes helps it determine why and what to do next. Outages happen most often through the unexpected interactions of multiple systems. We must connect diverse operational data to build a complete picture of what is happening.
That’s the role of the intelligence engine we’re building: turning operational data from Azure’s vast resource estate into context that people and agents can act on, right where they work. It grounds investigations and actions in how systems actually work together. The value of an agentic system comes from how well it understands intent, how quickly it acts on it, and whether it comes back with an outcome people can act on right away. That outcome should feel personalized, shaped by the user’s own context and environment. Our intelligence engine is what makes that possible. It lets an agent read a request considering the user’s actual environment, so intent is understood in context. It keeps dependencies, configuration, and recent changes already connected, so the agent can move quickly instead of gathering them from scratch. And it shapes the answer around the user’s own estate, so the outcome is specific to them rather than generic. Because the engine absorbs the complexity of how every resource fits together, people can spend their time on what they want to accomplish instead of how the estate is wired.
Specialized skills bring operational expertise to this foundation. Azure agents come with skills for different areas of cloud operations, and organizations can develop custom skills that capture their own architectures, procedures, and requirements. These skills help agents apply that expertise to the environment they’re operating in. Azure’s existing identity, permissions, policies, and approval requirements govern their actions, so teams can automate more of their operations without losing oversight.
Advancing agentic operations at Microsoft Ignite 2026
At Ignite, we’re introducing new capabilities across Azure Resource Manager MCP Server, Azure Copilot, Azure SRE Agent, and Azure Monitor to give customers more ways to put agents to work with the operational context and controls they need. Whether teams connect their own agents, use built-in Azure agents, or use and extend Azure SRE Agent, these updates help them build on the same Azure management and governance foundation.
Azure Resource Manager (ARM) MCP Server lets agents take governed action across Azure
The ARM MCP gives MCP-compatible clients and agents read access to your Azure resources through the control plane, across more than 3,500 resource types. Now, agents can also create, manage, and update resources, bringing your own operational agents into the same management workflows teams already use. Requests through the ARM MCP server are evaluated by Azure Policy. Policies that block noncompliant deployments, require tags, or restrict regions and SKUs apply whether a change comes from the portal or an agent. And as agents themselves become Azure resources, teams can also use Azure Policy to audit and enforce their configuration. Learn more about the ARM MCP Server.
Azure Copilot expands agentic capabilities across the cloud lifecycle
The deployment agent is now generally available, helping teams translate natural language requirements into infrastructure and deployment code. Azure Copilot spaces and custom skills, now in public preview, let teams provide persistent context and operational knowledge tailored to their environments. Azure Workload Center, now in limited preview, extends agentic operations to Oracle and SAP, helping teams monitor workload health and validate configuration.
Azure SRE Agent brings secure, governed agentic operations to enterprise environments
Azure SRE Agent helps teams investigate incidents, identify root causes, and automate operational tasks. New capabilities in public preview strengthen its security, governance, and extensibility for production environments. Customer-managed keys, Entra Agent ID, and integration with Agent 365 give organizations greater control and governance across the agent lifecycle. Beyond individual agents, cross-agent communication lets agents share context and coordinate tasks, while Microsoft Teams integration brings investigations across multiple clouds, helping teams work through issues that span their production environments. Learn more about Azure SRE Agent
Azure Monitor health models bring richer context to cloud operations
Now generally available, Azure Monitor health models turn telemetry into a real-time view of application health, connecting architecture, dependencies, and business impact. Teams can quickly see what’s affected, distinguish application failures from platform issues, and prioritize their response.
That shared view also gives agents a stronger basis for investigation: understanding how a failing resource affects an application, identifying root causes, and recommending or automating remediation. It’s a concrete example of the operational context that makes agents more useful. Learn more about Azure Monitor health models.
Agentic operations in action at Ecolab
We’re already seeing what this looks like in practice.
At Ecolab, engineers are using Azure’s agentic operations capabilities – bringing automated investigation, code-level insights, and application health context into the work of understanding and responding to issues. Engineers are using Azure SRE Agent, Azure Copilot Observability Agent, and Azure Monitor health models together to help teams spend less time piecing together evidence and more time resolving issues.
As a Software Engineer Manager at Ecolab, explains:
“By applying agentic operations with Azure, we’re transforming operations from reactive troubleshooting to proactive resilience, helping our teams save time and effort from issue detection through resolution. Azure Monitor health models helped us identify issues 30% earlier, reduce customer-impacting incidents, and cut through low-value alert noise. With Azure SRE Agent, we automated root cause analysis, helping us identify root causes 60% faster while reducing escalations and reliance on specialized teams. Complementing Azure SRE Agent, Azure Copilot observability agent turned alerts into autonomous investigations across 50+ services from day one, saving 30 minutes per alert. Together, these capabilities helped us detect issues earlier, resolve incidents faster, and reclaim 3 to 4 hours per incident, giving our teams more time to focus on innovation, improve reliability, and deliver greater value to customers.”
The value isn’t just a faster investigation. It’s making operational knowledge available across the team, reducing escalations, and giving engineers time back to improve the systems they run.
Microsoft leads the way in adopting agentic operations
We're also applying this approach across Microsoft's own engineering teams. More than 11,000 agents powered by Azure SRE Agent have been deployed across Microsoft, 7 times more than when the product became generally available in March 2026. The fast growing adoption is helping Microsoft teams move from reactive incident response toward more proactive operations. Over the last 90 days, 61% of total tasks are focused on proactive work and an average 10% of millions of incidents are handled end-to-end without human intervention. The impact is tangible. Remediation that once took 100+ minutes can now take just 4, giving engineers more time to focus on improving the services they run.
Get started with agentic operations on Azure
The experience at Ecolab and across Microsoft shows the shift to agentic operations is already underway. Teams are resolving issues faster and spending more time on proactive work. The next step is to make improving reliability, performance, and cost efficiency a continuous part of how systems operate – not just work that happens in response to an incident.
That’s where we believe cloud operations are headed. AI agents won’t replace the expertise of IT teams, but they can put that expertise to work in new ways. With agentic operations on Azure we’re helping teams take that step today.
Learn more
If you’re joining us at Moscone Center this week or catching up on demand, be sure to check out these Microsoft Ignite sessions to learn more about agentic operations on Azure.
- BRK415: Put AI agents to work in Azure cloud operations — Discover how AI agents can help teams manage and improve cloud operations.
- BRK416: Drive modern operations with Azure Monitor — Learn how to turn operational signals into actionable insights.
- BRK417: Scale operations, not toil, with your AI SRE teammate — Explore how AI-powered SRE capabilities can accelerate investigations and incident response.
- THR895: AI for FinOps, continuous optimization for every workload — See how AI can help teams continuously optimize cloud resources and costs.
- LAB724: Govern your estate with Azure Arc and Azure Resource Manager MCP — see how to use the Azure Resource Manager MCP server with real resource management scenarios (in-person only)