AI-900 Practice Question: Describe Artificial Intelligence workloads and considerations

Answer analysis

Option-by-option breakdown

For each option: why learners choose it and why it is or isn't the right answer here.

• ✓

  Create a QnA Maker resource, then create a knowledge base and add QnA pairs, then test the knowledge base, then publish the knowledge base.

  Why this is correct

  Create a QnA Maker resource first so a hosting service exists, including the Azure Search index and runtime endpoint that a knowledge base depends on. Adding QnA pairs then populates question and answer sets inside that resource, and the Test pane validates matching behavior before any users are affected. Publishing is the final step because it deploys the verified knowledge base to a public HTTPS endpoint, making it available to client applications.

• ✗

  Create a QnA Maker resource, then test the knowledge base, then create a knowledge base and add QnA pairs, then publish the knowledge base.

  Why it's wrong here

  This order is invalid because the Test operation has nothing to test: the knowledge base has not yet been created and no QnA pairs have been added. In QnA Maker, testing is performed from the Test pane within the knowledge base editor, which requires populated FAQ pairs or imported content. Without a content-holding knowledge base, there are no user queries to evaluate, so any test would either fail or return an empty response before the publishing step is even reached.

• ✗

  Create a knowledge base and add QnA pairs, then create a QnA Maker resource, then test the knowledge base, then publish the knowledge base.

  Why it's wrong here

  A knowledge base cannot exist independently of a QnA Maker resource; the resource supplies the required Azure region, subscription key, and backing Azure Cognitive Search instance. Attempting to create QnA pairs first would occur outside the resource hierarchy, so there is no service to attach the content to. The resource must be provisioned and selected in QnA Maker before the portal will allow you to name and create a knowledge base.

• ✗

  Create a QnA Maker resource, then create a knowledge base and add QnA pairs, then publish the knowledge base, then test the knowledge base.

  Why it's wrong here

  Publishing compiles the current knowledge base and deploys it to a live endpoint that client apps and bots call, so doing this before testing risks exposing incorrect answers to production users. The Test pane gives immediate feedback on confidence scores and answer selection, which is necessary to catch gaps before endpoint activation. A responsible flow always validates the knowledge base's responses locally and then publishes only after testing confirms the intended behavior.

Go deeper

Related to this question