How to setup an Azure OpenAI service

In the last blog post Azure OpenAI Service in an enterprise environment, the topic of AI was mentioned in our blog for the first time. This served as a very good introduction and inspiration for thought on how AI could be used in an enterprise environment together with Microsoft Azure services.

Today I would like to show in a few steps how such an Azure OpenAI service can be built and configured within an hour with the help of Azure Portal and other UIs from Microsoft without much prior knowledge.

Let’s get to the prerequisites so that you can follow the how-to and build it up for yourself.

Steps
  1. Creation of resource group
  2. Creation of the Azure OpenAI resource
  3. Open the Azure OpenAI Studio now
  4. Select the previously created Azure OpenAI resource
  5. In order to be able to chat with the service, a so-called deployment must be created. To do this, go to Management > Deployments > Create new deployment on the left-hand side
  6. The common models already known from OpenAI ChatGPT can be selected here and made available to the API
  7. The service is now in place so that the AI model can be addressed using the Azure OpenAI API or used for the first time in Azure OpenAI Studio under Chat.

Now this is all well and good, but of course not useful for end users. As a further step, we will now deploy a first web frontend which is similar to ChatGPT and is protected for customer employees.

UI Deployment
  1. To do this, go back to the Azure OpenAI Studio under Chat
  2. At the top right you will find the blue Deploy to button
  3. The configuration and deployment of the frontend as an Azure WebApp can now be started directly from the Azure OpenAI Studio.
  4. The deployment of all Azure resources takes approx. 5-10 minutes and is displayed at the top as a notification.
  5. The frontend can now be opened directly from the notification with Launch web app

  6. When opening the website for the first time, Authentication Not Configured appears immediately. This is correct, as no identity provider has yet been configured.
  7. To do this, go to the Azure Portal > Web App > Settings > Authentication and click on Add identity provider

  8. Select Microsoft as the identity provider and the configuration can be left as it is by default
  9. After approx. 5 – 10 minutes you can reload the website and the chat will load successfully incl. integration of Microsoft as identity provider.

Summary

The following setup can be build in your Azure tenant within approx. one hour:

  • Azure OpenAI service with your choice of data location
  • Deployment of the appropriate predefined AI model
  • Chat accessible for users of your organization including the chat history and protection via Microsoft Entra as Identity provider
You might also like