Free AI-500 Exam Braindumps
Designing and Implementing Multi-Agent AI Solutions Questions 36–40 (page: 9)

Updated On: 2-Oct-2026
View Related Case Study

You need to recommend a knowledge integration design for a Microsoft Foundry multi-agent solution. The agents answer questions by using shared documentation. The solution must meet the following requirements:
● Updates must be available from a single maintained knowledge layer.
● Retrieval responses must include citations and query details.
● Content must support natural-language queries.
Users will ask the agents complex conversational questions. The questions will include follow-up context and terminology that does NOT always match the wording in the documentation.
Which knowledge type should you recommend?

  1. File
  2. Fabric IQ (OneLake Catalog)
  3. Azure AI Search Index
  4. Work IQ

Answer(s): C

Explanation:

To meet your requirements for a Microsoft Foundry multi-agent system, an Azure AI Search Index using Hybrid
Search (Vector + Keyword) with Semantic Ranking is the best knowledge configuration.
This approach acts as your centralized, single-source knowledge layer while natively handling the linguistic gaps between your users and your documentation.


Reference:

https://learn.microsoft.com/en-us/azure/cloud-adoption-framework/ai-agents/build-secure-process



View Related Case Study

You have a Microsoft Foundry agent that completes benefits enrollment during a single user conversation. The agent collects the required enrollment fields during 15 turns. Users can correct earlier values before final submission. The current implementation sends the complete transcript with every model request.
You need to change the context accumulation strategy for the active enrollment. The solution must meet the following requirements:
● Preserve the latest value for each required enrollment field until submission.
● Bound the maximum number of tokens sent with each model request.
● Preserve user corrections until submission.
● Prevent durable cross-session memory.
What should you do?

  1. Use an enrollment-state snapshot with a recent-turn sliding window.
  2. Use response chaining with a fixed number of previous responses, relying on the conversation history to preserve the enrollment field values.
  3. Use conversation-scoped, server-managed history with automatic input truncation.
  4. Use as elective summarization of older low-value conversation turns.

Answer(s): A

Explanation:

Using an enrollment-state snapshot combined with a recent-turn sliding window is an excellent and industry-standard architecture for this problem. This approach perfectly balances state tracking with token efficiency.
Requirement Mapping
Preserve latest values & Preserve user corrections: The enrollment-state snapshot acts as your "source of truth." When the user provides a piece of data (or corrects an earlier one), the agent updates this explicit
JSON/object state. Because you only store the latest validated value in the snapshot, corrections automatically overwrite old data, satisfying both requirements.
Bound the maximum tokens: The recent-turn sliding window (e.g., keeping only the last 3–4 conversational turns) ensures that the context sent to the LLM does not grow linearly over the 10 turns. This keeps your token count strictly bounded and predictable.
Prevent durable cross-session memory: Because the state snapshot and sliding window are tied entirely to the active session's context, you can simply wipe or discard the session state object upon submission or session timeout. No data will bleed into long-term or cross-session storage.
Incorrect:
[Not C]
Using conversation-scoped, server-managed history with automatic input truncation is not the best action,
because automatic truncation directly risks violating the requirement to preserve user corrections and the latest enrollment values.
If a user corrects a value they provided early in the 15-turn conversation, automatic truncation (which typically cuts off the oldest turns first) could erase the original context or the correction itself. This creates a risk of the model losing track of the true state or reverting to a stale value.


Reference:

https://learn.microsoft.com/en-us/azure/foundry/agents/concepts/runtime-components



View Related Case Study

Note: This question is part of a series of questions that present the same scenario. Each question in the series contains a unique solution that might meet the stated goals. Some question sets might have more than one correct solution, while others might not have a correct solution.
After you answer a question in this section, you will NOT be able to return to it. As a result, these questions will not appear in the review screen.
You have a Microsoft Foundry Agent Service solution that includes two agents.
You need to configure memory for the agents. The solution must meet the following requirements:
● Isolate the memory between end users.
● Isolate the memory between the agent domains.
● Support the deletion of one user's memory without deleting other users' memory.
Solution: You create a dedicated memory store for each agent and configure a static agent scope value for each memory search tool.
Does this meet the goal?

  1. Yes
  2. No

Answer(s): B

Explanation:

Correct:
* You create a dedicated memory store for each agent and set scope for each memory search tool to
{{$userId}}.
1. Provision a dedicated memory store for each of the two agents.
This follows Microsoft’s design best practices. Keeping independent stores establishes hard boundaries between your agents' specialized long-term context.
2. Set a unique user identifier for the scope parameter on every request.
Scope is a mandatory parameter that forms a strict security boundary. A search executing in User A's scope can never return results from User B's scope.
Incorrect:
* You create a dedicated memory store for each agent and configure a static agent scope value for each memory search tool.
* You create one memory store per end user and configure both agents to use each user's memory store with a static scope value.


Reference:

https://learn.microsoft.com/en-us/azure/foundry/agents/how-to/memory-usage



View Related Case Study

HOTSPOT (Drag and Drop is not supported)
You have the following tool integration configuration for a Microsoft Foundry claims-processing assistant.

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:

  1. See Explanation section for answer.

Answer(s): A

Explanation:



Box 1: Yes
Yes, ClaimsPlanner can only see final_text returned from the FraudReview integration.
According to the configuration code under the FraudReview integration block, the data_exchange rules explicitly define what data flows back to the planner:
id: FraudReview
...
data_exchange:
sent_from_claimsplanner: callabe_input_ony returned_to_claimsplanner: final_text
Because returned_to_claimsplanner is strictly set to final_text, any other intermediate data, metadata, or raw payloads generated during the fraud screening process are hidden from the planner.
Box 2: Yes
Yes, PremiumRates can be exposed to ClaimsPlanner as an OpenAI-defined tool without adding developer-managed conversation history.
The configuration code explicitly defines this behavior under the PremiumRates integration block:
OpenAI-defined tool format: The contract: format: openai setting ensures that the tool's signature and interface follow the standard OpenAI contract format.
No developer-managed conversation history: The invocation parameters explicitly disable state tracking with stored_conversation_state: none and prevent previous planner history from being forwarded to the tool via planner_history_sent: false.
This configuration ensures that the tool operates as a stateless, direct invocation within the model-driven selection loop.
Box 3: No
No, the provided configuration dictates that unsupported schemas are rejected or skipped rather than adapted dynamically before invocation.
While PartnerIntegrations behaves like an MCP-style dynamic tool pattern by loading external tools via a listing_endpoint at session_start, the ClaimsPlanner handles unsupported schemas through a strict filtering and rejection mechanism defined directly in the runtime configuration.



View Related Case Study

You have a Microsoft Foundry agent named Agent1.
You open Agent1 in the playground and update the instructions.
You need to run a full evaluation against the updated instructions. The solution must meet the following requirements:
● Test the changes by using a synthetic dataset.
● Ensure that the changes are available only for the development team that has access to Agent1.
What should you do first?

  1. Publish Agent1 as a new version.
  2. Create a new evaluation for Agent1.
  3. Preview Agent1.
  4. Save Agent1 as a new version.

Answer(s): A

Explanation:

The first step is to save the updated instructions as a new agent version within the Microsoft Foundry project,
without publishing it to an external or global endpoint.
Meets the Development Team Requirement: In Microsoft Foundry, keeping the agent inside the project workspace as a development version ensures that the modifications are completely isolated. It prevents the updated configuration from going live to production or external platforms, restricting access exclusively to teammates who hold internal role-based access control (RBAC) permissions (like Foundry User or Foundry
Project Manager) within that specific project workspace.
Prepares for Evaluation: You cannot configure an automated or synthetic evaluation framework against temporary, unsaved changes sitting in the playground UI. Saving a specific snapshot version provides a concrete evaluation target for Microsoft Foundry’s evaluation wizard or Python SDK.


Reference:

https://learn.microsoft.com/en-us/azure/foundry/how-to/evaluate-generative-ai-app



Viewing page 9 of 23
Viewing questions 41 - 45 out of 107 questions


Post your Comments and Discuss Microsoft AI-500 exam prep with other Community members:

AI Tutor AI Tutor 👋 I’m here to help!