# Microsoft Azure AI automation and integrations > Azure is the natural AI platform for companies already on Microsoft 365 and Entra ID. Microsoft Foundry, formerly Azure AI Foundry, provides managed access to OpenAI and other models under Azure agreements, Azure AI Search handles retrieval, and Document Intelligence reads forms. n8n has built-in Azure OpenAI, Azure AI Search vector store, Cosmos DB, Storage and Entra ID nodes. Common builds are policy assistants, document extraction and IT automation. - Category: Data & Cloud - Vendor: https://azure.microsoft.com/ - How to connect: Built-in n8n Azure OpenAI chat model node; n8n Azure AI Search vector store node; n8n Azure Cosmos DB and Azure Storage nodes; n8n Microsoft Entra ID node; Azure REST APIs and SDKs; Logic Apps and Power Automate - Authentication: Microsoft Entra ID app registration with client credentials or managed identity, or resource API keys where a service supports them - Notes: Register a dedicated Entra ID app for each automation and grant it roles on specific resources only. Model deployments in Foundry are per region with their own quotas, so deploy the models you need in your chosen region before building and request quota increases early. - Last updated: 2026-10-14 ## What you can automate with Microsoft Azure - **Policy and SOP assistant in Teams.** Index SharePoint policies and procedures in Azure AI Search and answer staff questions in Teams with links to the source. Guide: https://www.lumeor.studio/guides/build-internal-sop-assistant/ - **HR policy questions.** Answer handbook and benefits questions with citations, filtered by employee location or group. Guide: https://www.lumeor.studio/guides/ai-hr-policy-assistant/ - **Document and form extraction.** Read invoices, applications and forms with Document Intelligence and a model, validate the fields and route exceptions to a person. Guide: https://www.lumeor.studio/guides/ai-document-data-extraction/ - **User provisioning and offboarding.** Create and disable Entra ID accounts, assign groups and licenses, and log every step when HR records change. Guide: https://www.lumeor.studio/guides/automate-user-provisioning/ - **IT helpdesk triage.** Classify incoming IT requests, answer common ones from the knowledge base and route the rest with a summary. Guide: https://www.lumeor.studio/guides/automate-it-helpdesk-requests/ - **Alert digests.** Summarize Azure Monitor alerts and logs into a daily digest that groups repeats and names the likely cause. Guide: https://www.lumeor.studio/guides/ai-log-and-alert-summaries/ ## What makes Azure easy or hard to automate For a Microsoft 365 company, Azure is already halfway set up. Users, groups and permissions live in Entra ID, documents live in SharePoint and OneDrive, and the security team already reviews Microsoft services. n8n has nodes for the Azure services AI work needs most, alongside nodes for Outlook, Teams and SharePoint, and Logic Apps and Power Automate can trigger or be triggered by n8n. What makes it harder is the setup around each resource. Model deployments are created per region with token-per-minute quotas. App registrations need the right Graph and resource permissions, sometimes with admin consent. Networking options such as private endpoints add security but also add configuration that has to be right before anything connects. ## The AI layer Azure groups its AI services under Microsoft Foundry: - **Models.** OpenAI's GPT models through Azure OpenAI, plus models from other providers including Anthropic, deployed under your Azure subscription. - **Retrieval.** Azure AI Search provides vector and hybrid search with filters, which pairs with SharePoint content for internal assistants. - **Documents.** Document Intelligence extracts text, tables and fields from forms, invoices and IDs, and a model interprets the result. - **Agents.** Foundry includes agent tooling, and Copilot Studio offers low-code agents inside Microsoft 365. The usual pattern is SharePoint as the source, AI Search as the index, a Foundry model for answers, and Teams as the place people ask. ## Limits to plan around - **Quotas per deployment.** Model throughput is limited per region and deployment. Plan for peak use and request increases early. - **Permissions and consent.** Graph permissions such as reading all SharePoint sites need admin consent. Use site-scoped permissions where possible. - **Security trimming.** An internal assistant must respect SharePoint permissions. Store access groups as metadata and filter by the user's groups at query time. - **Region choice.** Not every model or feature is in every region. Choose a region that has what you need and satisfies residency requirements. - **Cost tracking.** Tag resources by project and set budgets in Cost Management before rollout. ## How Lumeor typically connects it For Microsoft-centric clients we build [SOP and policy assistants](/guides/build-internal-sop-assistant/) on SharePoint, Azure AI Search and a Foundry model, delivered in [Microsoft Teams](/tools/microsoft-teams/), and use n8n with an Entra ID app for multi-system workflows such as [user provisioning](/guides/automate-user-provisioning/). Simple approvals often stay in Power Automate, and our [n8n vs Power Automate comparison](/compare/n8n-vs-power-automate/) explains where we draw the line. ## Frequently asked questions ### What is the difference between Azure OpenAI and OpenAI's own API? Azure OpenAI serves OpenAI's models through Azure, under Microsoft's agreements, billing, regions and Entra ID access controls. OpenAI's own API is a separate service with its own account. Teams choose Azure when procurement, data residency or security reviews are already built around Microsoft. ### What is Microsoft Foundry? Microsoft Foundry is the new name for Azure AI Foundry, Microsoft's platform for deploying and managing AI models and agents on Azure. It includes OpenAI models and models from other providers, plus tools for evaluation and agents. Older documentation still uses the Azure AI Foundry name. ### Should we use Copilot Studio, Power Automate or n8n? Copilot Studio and Power Automate suit simple flows that stay inside Microsoft 365 and are built by business users. n8n is better for multi-system workflows, complex AI agent logic and self-hosting. Many clients use Power Automate for simple approvals and n8n for anything with real AI logic. ### Does n8n connect to Azure? Yes. n8n has built-in nodes for Azure OpenAI chat models, Azure AI Search as a vector store, Azure Cosmos DB, Azure Storage and Microsoft Entra ID, along with nodes for Outlook, Teams, SharePoint and OneDrive. Other Azure services can be called with the HTTP Request node. --- Source: https://www.lumeor.studio/tools/microsoft-azure/ Published by Lumeor Studio, Newtown, PA. hello@lumeor.studio. Get a written scope: https://www.lumeor.studio/start/