Azure Migrate: Moving Your Workloads with Ease

Hello Everyone!

If you’ve ever attempted to transition a few hundred virtual machines (VMs), numerous databases, and several web apps from an on-premises setup to Azure, you know that the real challenge isn’t the tools themselves. It’s all about maintaining clarity, purpose, and forward momentum amidst weeks of planning, handovers, and decisions. At the Microsoft Azure Infra Summit 2026 session MAIS15, Ankur Gupta, a Senior Product Manager with the Azure Migrate team, introduced us to a groundbreaking solution: the Azure Migrate agent. This AI layer enhances Azure Migrate, helping to carry your intentions from that initial idea to the fully deployed landing zone.

The Challenge of Infrastructure Complexity

Ankur kicked things off with a poignant observation: the complexity of infrastructure is soaring beyond what humans can manage. With MySQL instances here, PostgreSQL databases there, countless web applications, storage devices, networking tools, and multiple dashboards and alerts, we’re all tasked with moving faster and making fewer errors than ever. To put it simply, migrations typically falter not because the wrong tool was chosen but due to breakdowns between stages.

Why the Agentic Approach is Crucial

Here’s how this new agentic method can help those working on the ground:

• It maintains context throughout the entire process, ensuring that the intent established on day one is preserved through execution.
• It offers guidance when you’re uncertain, helping you choose the right discovery method without the confusion.
• It streamlines tasks that might traditionally take days to analyse into just a few hours.
• It unites IT operations, architects, and developers through a consistent flow of information, allowing for a smooth transition to GitHub Copilot for coding tasks.
• It builds upon the Azure Migrate portal you’re familiar with, ensuring you can still utilise your existing knowledge.

A Continuous Evolution

The Azure Migrate portal remains intact, but now this agent acts as an additional layer. Azure Migrate was always the hub for discovery, assessment, and migration, and now it’s evolving into a comprehensive migration control platform that spans the entire lifecycle: Decide, Plan, Execute. The Azure Migrate agent is a conversational, guidance-based tool that ties everything together.

Your New AI Partner in Migration

As Ankur stated, the agent is designed to be both educational and supportive. By asking simple, natural-language questions like, “How should I plan moving my VMware workloads to Azure?” you receive practical next steps tailored to your needs. The agent excels in three main areas:

• It preserves state throughout the migration process, so your initial preferences remain intact.
• It keeps context during discovery, planning, and execution, allowing you to navigate freely while staying aligned with your goals.
• It recommends appropriate next steps based on your chosen intent.

Experience the Full Migration Flow

During the session, Ankur demonstrated a complete VMware-to-Azure migration scenario, showcasing how these elements seamlessly fit together. The agent currently supports three discovery methods: appliance-based discovery, RV Tools, and the new Azure Migrate collector. The collector is a lightweight alternative: it consists of PowerShell scripts that you run on any machine that can access your vCenter. Then, you simply upload the resulting zip file to your Azure Migrate project. There’s no appliance to deploy or complex network configurations needed.

Upon receiving your inventory, the agent interprets the data. Ankur asked for a summary and received a card listing 207 VMs, 177 SQL databases, one PostgreSQL instance, and various web apps. He was able to filter this data for servers with unsupported operating systems and tagged the relevant ones directly in the conversation for easy reference later.

Creating Business Cases Made Simple

Next came the business case creation. Ankur prompted the agent to generate cases based on modernising preferences. Minutes later, he had Azure costs alongside on-prem installation costs and projected savings. He then asked for another case related to lift-and-shift, and the agent provided a comprehensive side-by-side comparison. Remarkably, the agent showed that the lift-and-shift option had a higher total cost of ownership (TCO) in that scenario, providing crucial data for decision-making.

Detailed Application Assessment

Continuing with application assessments, Ankur returned to the portal to create an evaluation for two apps: Airsonic and Parts Unlimited. He initiated a high-confidence plan and received a recommendation to modernise, including a cost estimate of around £580 per month and estimated CO2 emissions of 32 kg. The web tier was flagged for App Service, while the data tier recommended Azure Database for PostgreSQL, marked as “ready with conditions,” complete with links explaining any requirements.

Connecting Development and Operations

One particularly exciting moment was when Ankur integrated GitHub for a Copilot Assessment, enhancing the infrastructure readiness with code-level insights. As a result, the estimate for migration effort became increasingly precise.

Creating a Wave Plan and Deploying Infrastructure

Ultimately, the agent generated a wave plan from the assessment and crafted a platform landing zone that aligns with Azure best practices. Ankur could query the agent about specific defaults, request modifications to the deployment process, substitute in a third-party firewall, or apply naming conventions. The agent then produced a downloadable Infrastructure-as-Code template for a cloud architect to polish in their IDE using GitHub Copilot.

Understanding the Benefits

So why does all this matter? Here are some scenarios where this new approach pays off:

• Pitching migration proposals to decision-makers is now streamlined. With the agent, generating modernise, lift-and-shift, and Azure VMware Solution business cases is reduced from days to mere hours.
• It helps in tackling legacy issues. By tagging outdated servers in a single conversation, you can easily plan for necessary upgrades without cumbersome CSV exports.
• When managing mixed environments, the agent suggests applicable App Service and Azure Database for PostgreSQL targets, along with SKU recommendations and relevant warnings to check.
• Closing the gap between IT and development: The GitHub Copilot Assessment and the IaC handoff ensure that developers and architects have a shared understanding.
• Minimising intent drift during lengthy migrations, as the agent keeps the original intent fresh in everyone’s minds.

In summary, the return on investment here is measured in saved time, not just money. It could mean fewer late-night calls resolving issues due to lost context on decisions.

Important Considerations: The agentic features are currently in preview, and outputs remain advisory. Human oversight, testing, and governance on all recommendations are still essential.

Getting Started: A Practical Guide

If you’re looking to dive in, here’s a straightforward approach:

1. Set up an Azure Migrate project in the Azure portal if you haven’t already.
2. Choose a suitable discovery method for your environment. If deploying an appliance isn’t an option, consider the new collector. Download the PowerShell scripts, run them on a machine that can access vCenter, and upload the resulting zip file.
3. Integrate the Azure Migrate agent within the portal and ask it to summarise your discovered inventory.
4. Create at least two business cases (modernise and lift-and-shift) and compare them.
5. Run an application assessment on a small selection of apps. Connect GitHub and include a Copilot Assessment for deeper code-level insights.
6. Request the agent to create a wave plan and a platform landing zone template, then send the IaC to your repository for the architects.

Start small, build your skills, and expand as you grow.

Keep Learning at the Summit

Be sure to check out the complete session playlist from the Microsoft Azure Infra Summit 2026 here.

Cheers!

Pierre Roman