Build an AI IT Support Agent with Azure Search, Entra ID & Jira

Description

An intelligent IT support agent that uses Azure AI Search for knowledge retrieval, Microsoft Entra ID integration for user management, and Jira for ticket creation. The agent can answer questions using internal documentation and perform administrative tasks like password resets.

How It Works

The workflow operates in three main sections:

Agent Chat Interface: A chat trigger receives user messages and routes them to an AI agent powered by Google Gemini. The agent maintains conversation context using buffer memory and has access to multiple tools for different tasks.

Knowledge Management: Users can upload documentation files (.txt, .md) through a form trigger. These documents are processed, converted to embeddings using OpenAI's API, and stored in an Azure AI Search index with vector search capabilities.

Administrative Tools: The agent can query Microsoft Entra ID to find users, reset passwords, and create Jira tickets when issues need escalation. It uses semantic search to find relevant internal documentation before responding to user queries.

The workflow includes a separate setup section that creates the Azure AI Search service and index with proper vector search configuration, semantic search capabilities, and the required field schema.

Prerequisites

To use this template, you'll need:

n8n cloud or self-hosted instance
Azure subscription with permissions to create AI Search services
Microsoft Entra ID (Azure AD) access with user management permissions
OpenAI API account for embeddings
Google Gemini API access
Jira Software Cloud instance
Basic understanding of Azure resource management

Setup Instructions

Import the template into n8n.

Configure credentials:
Add Google Gemini API credentials
Add OpenAI API credentials for embeddings
Add Microsoft Azure OAuth2 credentials with appropriate permissions
Add Microsoft Entra ID OAuth2 credentials
Add Jira Software Cloud API credentials

Update workflow parameters:
Open the "Set Common Fields" nodes
Replace <azure subscription id> with your Azure subscription ID
Replace <azure resource group> with your target resource group name
Replace <azure region> with your preferred Azure region
Replace <azure ai search service name> with your desired service name
Replace <azure ai search index name> with your desired index name
Update the Jira project ID in the "Create Jira Ticket" node

Set up Azure infrastructure:
Run the manual trigger "When clicking 'Test workflow'" to create the Azure AI Search service and index
This creates the vector search index with semantic search configuration

Configure the vector store webhook:
Update the "Invoke Query Vector Store Webhook" node URL with your actual webhook endpoint
The webhook URL should point to the "Semantic Search" webhook in the same workflow

Upload knowledge base:
Use the "On Knowledge Upload" form to upload your internal documentation
Supported formats: .txt and .md files
Documents will be automatically embedded and indexed

Test the setup:
Use the chat interface to verify the agent responds appropriately
Test knowledge retrieval with questions about uploaded documentation
Verify Entra ID integration and Jira ticket creation

Security Considerations

Use least-privilege access for all API credentials
Microsoft Entra ID credentials should have limited user management permissions
Azure credentials need Search Service Contributor and Search Index Data Contributor roles
OpenAI API key should have usage limits configured
Jira credentials should be restricted to specific projects
Consider implementing rate limiting on the chat interface
Review password reset policies and ensure force password change is enabled
Validate all user inputs before processing administrative requests

Extending the Template

You could enhance this template by:

Adding support for additional file formats (PDF, DOCX) in the knowledge upload
Implementing role-based access control for different administrative functions
Adding integration with other ITSM tools beyond Jira
Creating automated escalation rules based on query complexity
Adding analytics and reporting for support interactions
Implementing multi-language support for international organizations
Adding approval workflows for sensitive administrative actions
Integrating with Microsoft Teams or Slack for notifications