What can I build with Azure AI Foundry templates?
With Azure AI Foundry templates, you can quickly stand up several practical AI solutions without starting from scratch:
1. **AI-powered chat agents**
- Build a web-based chat application that uses an Azure AI Agent integrated with Azure AI Search.
- The app can pull answers directly from your own content (for example, uploaded files or internal documents) and return responses with citations, which is helpful in document-heavy environments like knowledge bases, policy libraries, or technical documentation.
2. **Conversation knowledge mining**
- Analyze large volumes of conversation data such as customer calls, support transcripts, or internal meetings.
- The solution extracts key topics, trends, and relationships so you can uncover patterns in how customers ask questions, where they struggle, or how teams collaborate.
3. **Document generation and summarization**
- Turn unstructured documents into usable content.
- The solution automatically identifies relevant files, extracts meaningful information, and generates structured drafts or summaries.
- This helps you scale document-heavy workflows like report creation, proposal drafting, or policy updates.
All of these templates are designed to be customized, so you can start with a working pattern and then adapt it to your data, workflows, and user experience.
How do the AI agent and chat app architectures work?
The AI agent–powered chat app in Azure AI Foundry follows a modular architecture that connects the front end, back end, and AI services:
1. **Front end and back end**
- You build a web-based chat interface that handles user interactions.
- The back end manages requests from the chat UI and connects to the AI Agent Service.
2. **AI Agent Service + Azure AI Search**
- The AI Agent Service receives the user’s question and uses **Azure AI Search** to look up relevant information from your uploaded files and indexed content.
- The agent then generates a response that includes **citations** back to the original documents, so users can see where the answer came from.
3. **Azure Container Apps for deployment**
- The chat application is deployed using **Azure Container Apps**, making it easier to run, scale, and manage the service in the cloud.
4. **Developer workflow**
- You can use **Visual Studio Code** or **GitHub Codespaces** to edit the template code and deploy directly to Azure.
- This lets you iterate on the agent’s behavior, data sources, and UI without rebuilding the entire stack.
The result is an AI chat experience that connects your users to your own content, rather than generic web data, while keeping the architecture clear and maintainable.
How does Azure AI Foundry handle unstructured conversations and documents?
Azure AI Foundry provides templates that focus specifically on turning unstructured data into something you can act on:
1. **Conversation knowledge mining**
- **Input:** Call transcripts and audio files from customer support, sales calls, or internal meetings.
- **Process:** The architecture ingests these files, applies Azure AI services to enrich and structure the data, and then surfaces insights through an interactive web experience.
- **Output:** Key topics, trends, and relationships across conversations, helping you spot recurring issues, emerging themes, and opportunities for process improvements.
2. **Document generation and summarization**
- **Input:** Unstructured documents such as reports, manuals, policies, or knowledge articles.
- **Process:** The solution uses **Azure AI Services**, **Azure AI Search**, and **Azure OpenAI Service** to:
- Process and index documents.
- Retrieve the most relevant content for a given task or prompt.
- Generate summaries, drafts, or templates via a web-based interface.
- **Output:** Structured drafts and concise summaries that help you scale document workflows without manually reading every file.
These patterns are meant to be starting points. You can extend them, connect them to your own data sources, and reimagine how your teams work with conversations and documents by automating the most repetitive analysis and drafting tasks.