Agents Gone Wild: The Best (and Worst) AI Agents You Can Build with Microsoft Today

AI agents are everywhere—but not all of them should exist. In this session, we’ll explore how to build powerful, practical AI agents using Microsoft’s modern AI stack, including Azure AI services, Azure AI Foundry, and Copilot-ready integrations.

Through real-world examples, we’ll look at useful agents such as a support assistant that summarizes customer tickets, a knowledge agent that answers questions from internal documentation, and an automation agent that triages alerts from cloud infrastructure. But we’ll also explore the “gone wild” side—agents that hallucinate policies, automate the wrong tasks, or create chaos when guardrails are missing.

Attendees will learn how to design, deploy, and manage AI agents responsibly using Microsoft tools, understand where agents add real value, and recognize common design mistakes before they cause problems.

By the end of the session, participants will understand how to build practical AI agents, integrate them with enterprise systems, implement safety and governance controls, and decide when an AI agent is the right solution—and when it absolutely isn’t.

Don't Ship That Agent

A pre-launch checklist talk: structured as the five things you must verify before any agent goes to production, in the order you should check them. More prescriptive and less narrative than Agents Gone Wild — built for teams who are about to ship and want a final-mile review.

AUDIENCE
Teams in the final stages of shipping their first or second production agent. Practical, checklist-driven audience.
EXPECTED OUTCOMES
• Walk away with a literal pre-launch checklist (provided as a takeaway doc)
• Understand the cost, safety, and grounding checks specific to agentic systems
• See how to wire these checks into CI/CD using Azure AI Foundry evaluation gates

The Honest AI Talk

An unusually candid look at the current state of agentic AI — separating genuine capability from hype, and giving attendees a calibrated sense of what's actually production-ready today versus what still needs guardrails, human oversight, or simply isn't there yet. Less demo-heavy, more analyst-style, with select live examples to back up each claim.

AUDIENCE
Mixed technical and business audiences, including decision-makers evaluating where to invest in agentic AI this year.
EXPECTED OUTCOMES
• Leave with a clear, opinionated framework for what to trust agents with today
• Understand the gap between demo-ware and production-ready agent patterns
• Get a vendor-neutral way to evaluate AI agent claims in pitches and proposals

My Agent Did What?

A rapid-fire tour of the strangest, funniest, and most expensive things AI agents have done in production — told through real incidents, then fixed live on stage. Built as a spiritual sequel to Agents Gone Wild, this session widens the lens beyond the original five patterns to cover scheduling agents, multi-agent handoffs, and tool-chaining disasters.

AUDIENCE
Developers and architects who have shipped (or are about to ship) an autonomous agent into production. Comfortable with code, new to agent-specific failure modes.
EXPECTED OUTCOMES
• Recognise 3 new failure patterns beyond the original five (handoff drift, tool-chaining runaway, silent permission creep)
• Walk away with a one-page incident checklist for agent code review
• See live fixes using Azure AI Foundry's evaluation and tracing tools

Create Your Own Agents with AI Foundry and Azure Logic Apps

The future of automation is agentic. With AI Foundry, developers can design, deploy, and manage custom AI agents that seamlessly integrate with enterprise workflows. When combined with the power of Azure Logic Apps, these agents can act on real-time events, trigger business processes, and orchestrate complex tasks—without writing endless lines of glue code.

In this session, we’ll walk through how you can:

1. Build and customize AI agents with AI Foundry.
2. Connect these agents to external systems using Azure Logic Apps.
3. Automate real-world business scenarios like task management, incident management, approvals, or customer support.

Whether you’re a developer looking to experiment with AI-driven automation or an architect aiming to scale intelligent workflows across your organization, this session will show you how to move beyond simple bots and into the era of true autonomous agents.

By the end, you’ll leave with the know-how to create your own agents that don’t just respond but also take action.

Supercharge Azure Management with Copilot,PowerShell and Resource Graph: Efficient Insights at Scale

Unlock the full potential of PowerShell by integrating it with Azure Resource Graph, transforming how you manage and gain insights into your cloud environment. This session dives deep into harnessing the power of Azure Resource Graph for large-scale resource management, querying, and reporting, all through seamless PowerShell automation. Learn how to leverage Copilot along with Resource Graph to visualize complex data sets, optimize cloud resources, and streamline governance, all while boosting your efficiency with scripts that provide real-time visibility across your entire Azure infrastructure. Whether you're a cloud architect or an IT pro, walk away with actionable strategies to scale your Azure management using these powerful tools.

Alexa! Run PowerShell!

Using Alexa with Azure PowerShell Functions for common tasks such as start-stop VMs, generating reports and some more.

Azure Copilot ile Serbest Ucus!

Bu oturumda populer bir yenilik olan Azure Copilot (Preview) uzerine konusacak ve canli demolarla Copilot'un birbirinden guzel yeteneklerine goz atacagiz.

Azure OpenAI ve PowerShell Kullanarak Azure uzerinde Dogal Dilde Sorgulama Yapmak

Bu oturumda Azure OpenAI, PowerShell ve Azure Functions kullanarak kullanisli bir arac olusturmaya odaklanacagiz. Bu arac sayesinde, dogal dilde sorgular olusturabilecek ve Azure kaynaklarini cok daha basit bir sekilde sorgulayabilecek ve raporlayabilecegiz.

Azure Firewall Manager with Latest Features

Preview Features of Azure Firewall Manager to centrally manage security policies, Azure Firewall instances and Virtual WAN Hubs

Building External MCP Servers with Azure Functions

The Model Context Protocol (MCP) is redefining how large language models interact with the outside world, enabling them to securely call external tools and APIs. In this session, we’ll explore how Azure Functions — enhanced with the new mcpTool trigger — can act as scalable, cloud-hosted MCP servers.

You’ll learn how to:
• Deploy a Python-based Azure Functions MCP server in minutes using Microsoft’s public sample.
• Define custom MCP tools that allow AI clients (like VS Code Agent Mode) to take real-world actions.
• Implement a practical scenario where MCP tools interact with Azure DevOps to fetch data, automate workflows, and respond to natural language prompts.
• Extend the pattern to integrate with any API or service, from cloud management to enterprise systems.

We’ll cover both the technical setup and real-world use cases, showing how this approach unlocks a new generation of AI-driven automation. By the end, you’ll see how to turn Azure Functions into a powerful bridge between LLMs and the tools your business runs on.

Who Let the Agents Out?

A session about boundaries: what happens when an agent has too much autonomy, too much tool access, or too little supervision. Framed around the idea that good agent design is really about deciding what an agent is NOT allowed to do. Strong fit for a security- or governance-adjacent track.

AUDIENCE
Platform teams, security engineers, and architects responsible for agent governance, approvals, and access control — not just the builders.
EXPECTED OUTCOMES
• Leave with a concrete checklist for scoping agent permissions before launch
• Understand the human-in-the-loop pattern and when it's mandatory vs. optional
• See a live example of an agent escalating correctly vs. incorrectly