AI agents are redefining how enterprises operate by enabling intelligent automation, decision-making, and cross-system orchestration. In this full-day workshop, you will explore how to design, build, and operationalize AI agents that deliver measurable business value and align with enterprise architecture principles.
The hands-on lab bridges strategy and execution, starting with identifying high-impact use cases and building a business case for AI agents. You will then design agent architectures using modern patterns such as multi-agent systems, Retrieval Augmented Generation (RAG), and tool integration across enterprise systems.
Through guided hands-on exercises using the Microsoft AI stack, including Microsoft Foundry ( Azure AI Foundry), Copilot Studio, and Microsoft Fabric, you will implement agents with governance, security, and scalability in mind. The hands-on lab also covers evaluation, observability, and lifecycle management to ensure production readiness.
By the end of the hands-on lab, you will have a practical framework and working solutions to confidently lead AI agent initiatives in your organization.
You will learn:
- How to identify and prioritize enterprise AI agent use cases aligned to business outcomes and ROI
- How to design and build scalable, secure agent architectures using multi-agent and RAG patterns
- How to operationalize AI agents with governance, evaluation, and lifecycle management for production readiness
Hands-on lab attendees will receive access to Azure, Microsoft Foundry, Azure OpenAI services, Microsoft Copilot Studio, and Microsoft Fabric (a $500 value) for use during the hands-on labs. Attendees who already have their own Azure subscriptions, Microsoft Fabric Capacity, and Microsoft Copilot Studio licenses may use those environments instead.
Attendee Workstation Requirements:
- You must provide your own laptop for this hands-on lab
WHAT TO BRING
- A laptop with an up-to-date version of Microsoft Edge or Google Chrome
- A headset with a microphone (required for the voice agent lab)
- A plain text editor, such as Notepad or Visual Studio Code
BEFORE THE FIRST SESSION
- Make sure your browser has access to your microphone and speakers.
- Use a personal or guest browser profile to prevent your work account from automatically signing in.
PORTALS YOU'LL USE
ACCOUNT ACCESS AND SUBSCRIPTIONS
For the hands-on labs, you'll need access to the following:
- Microsoft Foundry: An account with deployed AI models from OpenAI or another supported provider.
- Microsoft Copilot Studio: Access to Copilot Studio with available Copilot credits.
- Microsoft Fabric: Access to a Microsoft Fabric environment with at least F2 capacity.
**Good news!** If you don't have access to these environments, we'll provide a pre-provisioned account with all the required access so you can fully participate in the hands-on labs.
However, **we recommend using your own accounts and subscriptions whenever possible.** This will allow you to continue exploring and building on what you've learned after the workshop.
ADDITIONAL REQUIREMENTS FOR CODE LABS
If you'd like to follow along with the code-based labs, the following additional requirements apply:
- Admin rights on your laptop to install the tools below (install them before the workshop)
- Visual Studio Code**, with the Python and/or C# Dev Kit extensions
- Python 3.10 or later** (for the Python samples), and/or the **.NET 10 SDK** (for the C# samples)
- Azure CLI**: https://aka.ms/installazurecli
- Azure Developer CLI (azd)**: https://aka.ms/azd-install, then run: `azd ext install azure.ai.agents`
- Sign in before the session:** run `az login` and `azd auth login` with your Azure account
- A Microsoft Foundry project** in your subscription, with a deployed chat model (e.g., gpt-4o or gpt-4o-mini)
- Azure permissions:** at least the **Azure AI User** role on your Foundry project. To deploy a hosted agent, you'll also need **Contributor** (or Owner) on the resource group.
- Network:** make sure your laptop can reach Azure endpoints and pip/NuGet package feeds. Corporate VPNs and proxies sometimes block them.