Free Azure AI Apps and Agents Developer Associate AI-103 practice questions

Sample question 1. A deployed HR agent passed its pre-release evaluation on a fixed test set. Three months later, employees are asking about a new benefits program the test set never covered, and complaints about vague answers are rising. The team wants quality scores computed on real conversations as they happen. What should they configure?

1. A scheduled evaluation that reruns the original pre-release test dataset every week
2. An Azure Monitor alert on token usage and latency metrics for the agent
3. A continuous evaluation rule that samples live agent traffic and scores it
4. A scheduled AI red teaming scan against the agent every month

Sample question 2. A developer created a personal voice with the ID "grandpa-joe" from a consent recording and a 40-second audio prompt; the creation succeeded. Synthesis that references grandpa-joe in SSML fails. The voice talent must not be re-recorded. What should the app use for synthesis?

1. The same personal voice ID, passed as a deployment ID on a custom voice endpoint created for this voice.
2. A new personal voice created from a longer 90-second audio prompt so the service accepts its ID.
3. The consent ID from the voice talent's consent record, since synthesis checks consent by that ID.
4. The speakerProfileId that the service generated and returned when the personal voice was created.

Sample question 3. An online store needs a short alt text and a longer extended description for each of 50,000 product photos, written to its accessibility guidelines. Which approach BEST fits?

1. Send each image to a vision-enabled chat model with instructions defining both description formats, and spot-check a sample.
2. Run OCR on each photo and use the recognized text as the alt text and extended description.
3. Use an image generation model to recreate each photo from a prompt, then reuse that prompt as the description.
4. Use object detection to list the objects found in each photo, then join the detected object names into a single descriptive sentence.