Skip to content

01 Get started with AI agent development on Azure

1. Problem ❓

How can you build intelligent AI agents that autonomously automate business tasks using generative AI models in Azure?

2. Solution with Azure ✅

Leverage the following to develop, manage, and deploy AI agents that combine generative AI, contextual data, and programmatic tools:

  • Azure AI Foundry Agent Service
  • OpenAI Assistants API
  • Frameworks:
  • Semantic Kernel
  • AutoGen
  • Microsoft Copilot Studio

3. Required Components 🧩

  • Model:

  • OpenAI or Azure AI Foundry catalog models

  • Knowledge Sources:

  • Azure AI Search
  • Bing

  • Custom documents

  • Tools:

  • Built-in: e.g., code interpreter, Bing

  • Custom: e.g., Azure Functions

  • Conversation Thread:

  • Maintains context and data exchange history

  • Agent Frameworks / SDKs:

  • Azure AI Foundry SDK
  • OpenAI Assistants API (OpenAI models only)
  • Semantic Kernel Agent Framework
  • AutoGen
  • Microsoft 365 Agents SDK
  • Microsoft Copilot Studio / Agent Builder

4. Architecture / Development 🏗️

Single-Agent Scenario (e.g., Expense Agent)

  1. User submits a question
  2. Agent grounds the prompt using a knowledge source
  3. Model generates a response
  4. Agent performs an action (e.g., submits an expense claim)

Multi-Agent Scenario (e.g., Travel + Expense Agents)

  1. User provides trip details to Travel Agent
  2. Travel Agent books and collects receipts
  3. Travel Agent invokes Expense Agent
  4. Expense Agent submits the claim

Agent Construction in Foundry Agent Service ⚙️

  • Use visual interface or SDK
  • Define:
  • Model (LLM)
  • Knowledge source
  • Tools (built-in/custom)
  • Manage conversations via threads

5. Best Practices / Considerations 🧭

  • Copilot Studio Agent Builder:
    Ideal for business users with no coding skills

  • Copilot Studio (low-code):
    Best for Power Platform users

  • Microsoft 365 Agents SDK:
    Suitable for agents in Teams, Slack, etc.

  • Foundry Agent Service:
    Best for enterprise-grade, scalable agent development

  • Semantic Kernel:
    Recommended for multi-agent orchestration

  • Tool Selection Criteria:

  • User skill level
  • Target channels (e.g., Teams, Web)
  • Required integrations
  • Flexibility and scalability

✅ Tip: Prefer Foundry Agent Service over OpenAI Assistants API for richer integration with Azure.

6. Exam-like Questions 📝

Q1: What are the three core components required when building an agent in Azure AI Foundry Agent Service?
A1: Model, Knowledge, Tools

Q2: Which framework is best suited for orchestrating multi-agent solutions in Azure?
A2: Semantic Kernel

Q3: What distinguishes Azure AI Foundry Agent Service from the OpenAI Assistants API?
A3: Foundry provides greater model choice, integration with Azure services, and enterprise-grade features

Q4: A business user with no coding experience wants to build a simple task automation agent. Which tool should they use?
A4: Copilot Studio Agent Builder in Microsoft 365 Copilot

Q5: What Azure component allows agents to retain user interaction context and generated files?
A5: Conversation threads

7. Module Assessment 🎓

Q1: Which of the following best describes an AI agent?
A1: A software service that uses AI to assist users with information and task automation.

Q2: Which AI agent development service offers a choice of generative AI models from multiple vendors in the Azure AI Foundry model catalog?
A2: Azure AI Foundry Agent Service

Q3: What element of a Foundry Agent Service agent enables it to ground prompts with contextual data?
A3: Knowledge