Free AI-103 Plan and Manage an Azure AI Solution Practice Questions

The Plan and manage an Azure AI solution domain makes up 25–30% of the Microsoft AI-103 exam. Practice free questions covering selecting Azure AI services, Microsoft Foundry project setup, deployment planning, monitoring, cost management, and responsible AI — each with the correct answer and a detailed explanation.

AI-103 Plan and Manage an Azure AI Solution Sample Questions with Answers

Sample Question 1 — Plan and manage an Azure AI solution

A legal operations team is building an internal assistant in Microsoft Foundry. The assistant must answer only from policy documents that change every week, and users need answers grounded in the latest content. What is the best design choice?

  1. A. Deploy a larger model and rely on its pretraining
  2. B. Fine-tune the model whenever the policies change
  3. C. Use RAG with Azure AI Search over the policy corpus (Correct answer)
  4. D. Increase content filtering to block unsupported answers

Correct answer: C

Explanation: Correct answer (C): RAG is the best fit when answers must come from changing enterprise content. By retrieving current policy documents at runtime and grounding the response on those sources, the assistant can answer from the latest information instead of relying on static model knowledge. Why the other options are wrong: - Option A: A larger model does not solve the core problem of needing current private knowledge. Pretraining alone is not a reliable source for frequently changing enterprise policies. - Option B: Fine-tuning can shape behavior or style, but it is not the primary solution for rapidly changing private documents. Retrieval quality and freshness are more relevant here. - Option D: Content filters can help with harmful content, but they do not provide factual grounding or ensure that answers reflect the latest documents.

Sample Question 2 — Plan and manage an Azure AI solution

A support assistant indexes repair manuals in Azure AI Search. Users often search by exact part number, but they also ask natural-language questions that use different wording from the manuals. Which retrieval approach is the best fit?

  1. A. Use keyword-only retrieval across the index
  2. B. Use vector-only retrieval across the index
  3. C. Use hybrid retrieval in Azure AI Search (Correct answer)
  4. D. Fine-tune the model on the manuals instead

Correct answer: C

Explanation: Correct answer (C): Hybrid retrieval is the best choice when both exact terminology and semantic relevance matter. It combines lexical matching for exact identifiers such as part numbers with vector similarity for natural-language phrasing and conceptually similar content. Why the other options are wrong: - Option A: Keyword-only retrieval can work for exact part numbers, but it is weaker when users ask semantically similar questions using different wording. - Option B: Vector-only retrieval helps with semantic similarity, but it can miss the importance of exact lexical matches such as precise codes or identifiers. - Option D: Fine-tuning changes model behavior, but it does not replace a retrieval strategy when the requirement is to search indexed content effectively.

Sample Question 3 — Plan and manage an Azure AI solution

A Foundry-based internal copilot must connect to Azure AI Search and storage that contain confidential project data. Company policy prohibits public data paths, and the security review requires that the application not store secrets in code or configuration. Which design is best?

  1. A. Use public endpoints for both services and keep access keys in app settings
  2. B. Use private networking for connected resources and managed identity with RBAC (Correct answer)
  3. C. Use a larger model so the copilot needs fewer data connections
  4. D. Keep public endpoints and rely on content filters for protection

Correct answer: B

Explanation: Correct answer (B): This design aligns with both security requirements. Private networking reduces unnecessary exposure when public data paths are restricted, and managed identity with Azure RBAC avoids embedded secrets while supporting least-privilege access to connected resources. Why the other options are wrong: - Option A: This conflicts with both requirements. Public endpoints violate the network policy, and storing keys in app settings increases secret management risk. - Option C: Model size does not remove the need to securely access enterprise data sources. The main issue is connectivity and authentication design. - Option D: Content filters do not secure network paths or replace authentication and authorization controls.

Sample Question 4 — Plan and manage an Azure AI solution

An HR assistant must open cases in an internal API. The request must include required fields such as employee ID, category, and priority. Prompt instructions alone often produce missing or misnamed fields. What should you do?

  1. A. Add more examples to the prompt and hope formatting improves
  2. B. Fine-tune the model for better JSON style
  3. C. Use tool or function calling with a defined schema (Correct answer)
  4. D. Switch to a larger general-purpose model

Correct answer: C

Explanation: Correct answer (C): Tool or function calling is the best approach when the application must reliably invoke an external API with structured inputs. A defined schema helps constrain the request format and makes API invocation more controlled than prompt-only instructions. Why the other options are wrong: - Option A: More examples can help, but prompt-only behavior is still less reliable for structured API invocation with required fields. - Option B: Fine-tuning can influence output style, but it is not the preferred control mechanism for reliable external API execution with structured parameters. - Option D: A larger model may still omit or rename fields. The issue is control of API invocation, not just raw model capability.

Sample Question 5 — Plan and manage an Azure AI solution

A finance agent can request vendor payments and update payment dates through enterprise tools. The business wants faster processing, but it requires auditability and does not want the agent to act independently on high-impact transactions. What is the best design?

  1. A. Allow autonomous execution and depend on content filters
  2. B. Require human approval and use limited, auditable tool calls (Correct answer)
  3. C. Fine-tune the agent and grant broad finance permissions
  4. D. Use a larger model and remove the approval step

Correct answer: B

Explanation: Correct answer (B): High-risk financial actions should not be fully autonomous. Human approval provides an appropriate control point, and limiting tool access supports least privilege. Auditable tool calls also help the organization track who initiated actions and how the agent executed them. Why the other options are wrong: - Option A: Content filters do not make sensitive business actions safe to automate without review, and they do not replace approval controls. - Option C: Fine-tuning does not solve the governance problem, and broad permissions conflict with least-privilege design. - Option D: A larger model does not remove the need for approval and audit controls on sensitive financial operations.

Sample Question 6 — Plan and manage an Azure AI solution

An employee policy assistant uses RAG. After a major policy update, answers still cite old sections and sometimes miss the newest documents. The current model deployment is already strong. What is the best first action?

  1. A. Move to a larger model deployment
  2. B. Review chunking, refresh the index, and validate citations (Correct answer)
  3. C. Fine-tune the model on the old and new policies
  4. D. Tighten content filters for the responses

Correct answer: B

Explanation: Correct answer (B): When a RAG system returns outdated or incomplete answers after source changes, retrieval quality is the best place to investigate first. Index freshness, chunking strategy, and citation validation directly affect whether the latest documents are retrieved and grounded correctly. Why the other options are wrong: - Option A: A larger model is not the best first fix when the problem points to stale or poor retrieval. Grounding quality depends heavily on the retrieval pipeline. - Option C: Fine-tuning is not the primary remedy for frequently changing private knowledge. Retrieval freshness and source quality are more relevant. - Option D: Content filters do not fix stale indexes, poor chunks, or missing citations.

Sample Question 7 — Plan and manage an Azure AI solution

A production chatbot has rising token spend, slower responses, and occasional unsafe prompt incidents. The operations team currently tracks only uptime and HTTP success rate. What monitoring addition is most useful?

  1. A. Add traces, token usage, component latency, and safety events (Correct answer)
  2. B. Increase quota and continue using only service health metrics
  3. C. Collect only end-user star ratings after each conversation
  4. D. Replace retrieval with fine-tuning to simplify the logs

Correct answer: A

Explanation: Correct answer (A): AI application monitoring must go beyond basic service health. Traces, token usage, latency by component, and safety events help diagnose cost spikes, slow responses, and harmful interactions in ways that uptime metrics alone cannot. Why the other options are wrong: - Option B: More quota may help throughput in some cases, but it does not explain token cost, latency distribution, or unsafe prompt events. - Option C: User ratings can be useful, but they are not enough to diagnose technical behavior or safety trends. - Option D: Changing the solution architecture is not a substitute for observability, and fine-tuning does not inherently simplify production monitoring.

Sample Question 8 — Plan and manage an Azure AI solution

A company runs prototype prompts and production workloads in the same Foundry project and uses the same identities and connections for all stages. A recent change caused a production outage and made rollback difficult. What planning change is best?

  1. A. Keep one environment and document changes more carefully
  2. B. Separate development, test, and production environments (Correct answer)
  3. C. Use a larger model so prompt changes are less risky
  4. D. Rely on content filters to catch rollout problems

Correct answer: B

Explanation: Correct answer (B): Separate environments are the best planning choice because they improve version control, rollback, policy isolation, and identity boundaries. They reduce the risk that experimental changes directly affect production workloads. Why the other options are wrong: - Option A: Better documentation helps, but it does not provide the isolation needed for safe rollout and rollback. - Option C: Model size does not address environment isolation or deployment discipline. - Option D: Content filters do not prevent configuration drift or protect production from untested changes.

Keep Practicing

Take the 10-question AI-103 quick-start test across all 5 domains, or return to the AI-103 practice test hub for 200+ questions, exam details, and a study plan.