HOTSPOT (Drag and Drop is not supported)Select the answer that correctly completes the sentence.Hot Area:
Answer(s): A
Box: Information extraction solutions that detect and read text in scanned documents and images rely on ______________.Information extraction relies primarily on Optical Character Recognition (OCR), which is a specialized branch of computer vision. Instead of simply looking at pixels, these tools use advanced AI and machine learning models to "see" and interpret visual patterns.
https://www.ultralytics.com/blog/popular-open-source-ocr-models-and-how-they-work
You have a Microsoft Foundry project that has a generative AI model deployment.You need to ensure that responses generated by the model minimize costs and remain within a defined length.Which parameter should you configure?
Answer(s): C
The Max Completion Tokens (or max_tokens / max_output_tokens in the API) parameter is exactly the tool you use to enforce a fixed length on AI-generated responses and keep costs predictable.How the Parameter Works What it does: It sets a strict upper bound on the number of output tokens the model can generate for your completion. Once this number is reached, the model will stop generating (cut off) regardless of whether it finished its thought.What it doesn't do: It does not set a minimum length or force the model to write exactly that amount. It only acts as a ceiling.
https://learn.microsoft.com/en-us/answers/questions/5828812/reasoning-models-like-5-1-fails-with-500-error-cod
HOTSPOT (Drag and Drop is not supported)For each of the following statements, select Yes if the statement is true. Otherwise, select No.Note: Each correct selection is worth one point.Hot Area:
Box 1: Yes Yes - Human-in-the-loop practices provide accountability for AI-generated decisions.Human-in-the-loop (HITL) practices are foundational for ensuring AI accountability in Azure. Rather than granting models total autonomy, HITL establishes control layers, pause-and-approval mechanisms, and audit trails so that human operators remain the final authority on critical decisions.Box 2: No No - Deploying an AI system to production environment eliminites the need for ongoing monitoring.Deploying an AI system to a production environment does not eliminate the need for ongoing monitoring. AI applications face unique challenges like real-world data shifts, unpredictable user interactions, and changing environments, making continuous post-deployment observation essential to maintaining reliability, trust, and safety.Box 3: Yes Yes - disclosing the team that designed and deployed an AI system provides accountability for the system’s output.Under Microsoft's Responsible AI framework for Azure, disclosing the people and teams who design and deploy an AI system is a core mechanism for ensuring accountability.Microsoft explicitly states that accountability means the people who design and deploy AI systems must be responsible for how those systems operate. Clearly identifying these teams establishes human oversight and prevents the AI from being treated as the final authority on decisions.
https://learn.microsoft.com/en-us/azure/machine-learning/concept-responsible-ai https://verifywise.ai/lexicon/post-deployment-monitoring
Your company processes customer support emails.You need to implement an AI solution that automatically identifies mentions of people, organizations, and locations in the emails.Which text analysis technique should you use?
The adequate text analysis feature is Named Entity Recognition (NER), which is a core prebuilt capability of the Azure AI Language service.Prebuilt Categorization: The prebuilt NER feature automatically parses unstructured text (like emails) to identify and group elements into standard classes, including Person, Organization, and Location.No Training Required: Because it relies on state-of-the-art pretrained transformer models, it can be deployed immediately out-of-the-box without requiring data labeling or custom model training. High Extensibility: If your emails contain highly specialized terms (such as proprietary product IDs or industry-specific roles), you can easily scale to Custom NER within the same architecture.
https://learn.microsoft.com/en-us/azure/ai-services/language-service/named-entity-recognition/overview
Box: accountability Ensuring that human reviewers oversee AI-generated decisions and remain responsible for the final outout is an example if the Microsoft responsible AI princioke of ___________.The principle of accountability states that the people who design, deploy, and use AI systems must remain responsible and answerable for how those systems operate and the outcomes they produce. Under Microsoft's Responsible AI Principles, accountability is defined by the following key ideas: Human Oversight: AI should not be the sole decision-maker for critical matters; human reviewers must have the ability to supervise, control, and override AI-generated outputs. Ultimate Responsibility: Even when AI automates tasks or processes, humans and organizations remain legally and ethically responsible for the final decision.
https://thedatacommunity.org/2026/05/18/describe-considerations-for-accountability-in-an-ai-solution-ai-901-exam-prep/
You need to build an AI solution that generates marketing email drafts based on a short description of a product and its target audience.Which AI workload should you use?
A Generative AI workload is exactly what should be used for this scenario. Generative AI models (such as LLMs) are specifically designed to interpret prompts and create original, context-aware content, making them the perfect fit for drafting tailored marketing emails based on a product description and target audience.
https://azure.microsoft.com/en-us/products/ai-foundry/tools/content-understanding
Box 1: Yes Yes - The temperature parameter can be set before deploying a model.The temperature parameter can be set before deploying a model. Its actual enforcement depends on where and how the model is being deployed.Box 2: No No - During inference, the model name us used to route requests to a specific deployment.During inference, the deployment name is used to route requests to a specific deployment, not the model name.When you deploy a foundational model in Azure AI (such as Azure OpenAI or Azure AI Foundry models), you assign it a custom unique deployment name. During inference, you must pass this deployment name into the model parameter of your SDK call or request body.Box 3: Yes Yes - After a model is deployed, the model deployed, both code and testing tools can be used to interact with the model?After you deploy a model in Azure AI (such as in Azure AI Foundry or Azure Machine Learning), you can interact with it using both custom code and native testing tools.
https://github.com/shivamag125/EM_PT https://learn.microsoft.com/en-us/azure/foundry/foundry-models/concepts/endpoints
Box: fairness Evaluating model outcomes across demographic groups to reduce bias is an example of the Microsoft responsible AI principle of ________.Evaluating model outcomes across demographic groups to reduce bias falls squarely under Microsoft's responsible AI principle of Fairness.According to Microsoft's AI framework, AI systems should treat all people fairly, which requires teams to analyze how a system's predictions and recommendations impact different groups of people and actively mitigate issues like stereotyping, bias, and unfairness.
Share your comments for Microsoft AI-901 exam with other users:
1z0-1078-23 need this dumps
i gave the microsoft azure az-500 tests and prepared from this site as it has latest mock tests available which helped me evaluate my performance and score 919/1000
i cannot see the button to go to the questions
good questions
q-6 ans-b correct. https://docs.paloaltonetworks.com/pan-os/9-1/pan-os-cli-quick-start/use-the-cli/commit-configuration-changes
very nice very nice
please help us with 1z0-1107-2 dumps
please upload the practice questions
need this dumps
preparing for this exam is overwhelming. you cannot pass without the help of these exam dumps.
new to this site but i feel it is good
the correct answer to q8 is b. explanation since the mule app has a dependency, it is necessary to include project modules and dependencies to make sure the app will run successfully on the runtime on any other machine. source code of the component that the mule app is dependent of does not need to be included in the exported jar file, because the source code is not being used while executing an app. compiled code is being used instead.
Delayed the exam until December 29th.
A and D are True
good one with explanation
This is one of the most useful study guides I have ever used.