What you'll learn

A support inbox fills faster than anyone can read it, and every ticket in it needs the same three things: a category, a read on how angry the customer is, and a reply. In this course, Building the AI Backend, you'll gain the ability to build an event-driven generative AI pipeline on Azure that handles all three without a person in the loop. First, you'll provision the data tier: a blob container where tickets land, and a Cosmos DB container shaped for the documents the model produces. Next, you'll add the compute and AI layers, deploying a gpt-5-mini model and a Python Function App, wiring the environment variables that connect every service, and tracing one ticket from upload to stored document in Application Insights. Then you'll teach the pipeline which tickets not to answer. When it sees a legal threat, an abusive escalation, or a threat toward staff, it holds back the drafted reply and flags the record for a human. Finally, as a bonus, you'll watch the same pipeline running against a live Microsoft 365 tenant, where Microsoft Graph pushes real email in, replies go back out on the original thread, and critical tickets raise an alert card in a Microsoft Teams channel. When you're finished with this course, you'll have the skills and knowledge of Azure generative AI pipelines needed to run an LLM-backed automation workflow in production.