Microsoft Designing and Implementing Multi-Agent AI Solutions AI-500 Exam Questions

Page: 1 / 14
Total 73 questions

Want more questions? Get Premium Access.
()

Question 1

You need to design a solution to resolve the Scheduling agent issue.

What should you include in the design?

AImplement a retry mechanism that uses exponential backoff.

BRoute the tool execution directly to the Lead Orchestrator agent.

CRoute the tool execution directly to the Patient Intake agent.

DImplement a retry mechanism that uses immediate retries.

Answer : A

The Scheduling agent's MCP server fails transiently during peak load. Microsoft Azure reliability guidance recommends retries with exponential backoff for transient faults so that callers do not immediately repeat requests at the same rate and worsen a temporarily overloaded dependency. Immediate retries can create a retry storm, especially when many agent requests encounter the same service condition. Routing the tool call through a different agent does not remove the underlying MCP dependency and would only add coupling. A well-designed retry policy should normally include a bounded retry count, increasing delay, jitter where appropriate, and respect for server-provided retry guidance when available. The scenario asks which design element directly addresses the timeouts, and exponential backoff is the established resilience pattern for transient service saturation. Therefore A is the correct choice. At implementation time, the same rule should be expressed through the framework or service configuration rather than left only as a natural-language convention. That makes the behavior repeatable across runs, easier to test, and less sensitive to model variability.

Official Microsoft reference: Azure Well-Architected Framework - Handle transient faults

Question 2

You are designing a Microsoft Foundry multi-agent solution for claims processing. The design includes multiple specialized agents.

You need to specify the agent personas. scopes, boundaries, and autonomy levels. The solution must meet the following

requirements:

* Provide a clear owner for conflicts between specialist agents.

* Validate agent outputs before downstream agents consume the outputs.

* Prevent specialist agents from invoking tools outside the assigned domain.

* Isolate each business domain so that adding a specialist agent affects only that domain.

What should you do?

ADefine connected agents grouped by domain under a main agent, use natural-language delegation, and accept narrative summaries from the specialist agents.

BDefine a claims hub that has direct specialist delegation, gate the final settlement output against a checklist, and route conflicts through the hub.

CDefine domain-scoped sub-orchestrators under a claims supervisor, gate each output against a structured contract, and route conflicts through the supervisor.

DDefine domain-scoped workflows that have local quality gates, publish the accepted results to a shared case state, and let the consuming domains resolve conflicts.

Answer : C

Domain-scoped sub-orchestrators under a claims supervisor provide the clearest ownership and isolation model. Each domain can contain its own specialists and tools, so adding a new specialist affects only that domain. The supervisor becomes the explicit authority for cross-domain conflicts. Requiring every domain output to satisfy a structured contract before it is consumed downstream provides a deterministic validation boundary instead of relying on unconstrained narrative summaries. Microsoft AI-500 architecture objectives emphasize agent scopes, tool boundaries, structured interfaces, and explicit control loops. Option B gates only the final settlement and therefore allows invalid intermediate outputs to propagate. Option D deliberately leaves conflict resolution to consuming domains, which violates the requirement for a clear owner. Option A lacks a strong validation contract. C is therefore the most robust architecture. The architecture should still be validated with representative end-to-end tests, but the selected component establishes the correct structural boundary first. Microsoft's AI-500 blueprint consistently favors explicit scopes, interfaces, and persistence or identity boundaries over prompt-only conventions.

Official Microsoft reference: AI-500 Study Guide - agent personas, scopes, boundaries, and workflows

Question 3

You have a Microsoft Foundry multi-agent solution.

A developer publishes a new version of a specialist agent. Once the agent goes live in production, the solution starts mishandling requests.

You need to restore the previous behavior as quickly as possible

What is the fastest way to roll back the agent?

ACreate a new agent

BRedeploy the agent

CChange the endpoint.

DDelete the published version

Answer : C

The fastest safe rollback is to route the stable endpoint back to the previous known-good immutable agent version. Current Foundry lifecycle guidance supports versioned agents and endpoint/version selection so production traffic can be redirected without rebuilding the agent from scratch. Deleting the newly published version is a destructive cleanup action and is not the preferred rollback mechanism because it removes an artifact that may be needed for diagnosis. Creating a new agent changes the lifecycle identity and takes longer, while a complete redeployment is unnecessary if the earlier version already exists. Option C is therefore correct when interpreted as changing the endpoint's active-version or version-selector configuration to the previous version while keeping the endpoint URL stable. From a security and governance perspective, the control should be enforced at the narrowest platform boundary that can deterministically block or constrain the action. Relying only on prompt text is weaker because the model can still be induced to behave unexpectedly.

Official Microsoft reference: Microsoft Foundry agents - development lifecycle and versioning

Question 4

You have a LangGraph multi-agent workflow that uses Azure Cosmos DB as a checkpointer.

You plan to change the pruning rules for the running message list before the next release

You need to implement evaluations tor memory. The solution must meet the following requirements:

* Verify that claim-related state values remain correct at transition points.

* Verify that conversation context is preserved after an interruption.

* Identify the source of any loss of continuity during a run

Which three actions should you perform Each correct answer presents part of the solution NOTE: Each correct selection is worth one point.

AStore the expected values in short-term memory and remove durable external artifacts before each run.

BRun each turn with a different threaded and evaluate the final response text for the case.

CCapture the tool.call.arguments and tool.call.results attributes as the memory evaluation data for each tool invocation.

DRun a scripted multi-turn case through all the agents by using a single thread id before and after a process restart.

ECompare the persisted checkpoints documents for the evaluation thread_id with the expected State values after each handoff.

FCorrelate the evaluation run ID with trace spans for agent-to-agent interactions and state management.

Answer : D, E, F

Memory evaluation must verify persistence across interruption, correctness of checkpointed state, and enough telemetry to locate where continuity was lost. Running a scripted multi-turn case with one stable `thread_id` before and after a restart directly tests whether the Cosmos-backed LangGraph checkpointer restores the same conversation/workflow. Comparing persisted checkpoint documents with expected state values at each handoff verifies that claim-related state has not been corrupted by the new pruning rules. Correlating the evaluation run with trace spans then provides the diagnostic path to identify which agent transition or state-management operation caused a failure. Changing the thread ID each turn would intentionally create separate state scopes, and removing durable artifacts would defeat the persistence test. Tool-call arguments alone do not prove workflow-memory continuity. Therefore D, E, and F are the complementary actions required. The evaluation should also preserve correlation identifiers and version information where possible so a failed score can be traced back to the exact agent, model, tool call, or retrieval step that produced it. This turns the metric into an actionable diagnostic rather than only a dashboard number.

Official Microsoft reference: Azure Cosmos DB integrations for LangGraph and agent state

Question 5

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?

AUse an enrollment-state snapshot with a recent-turn sliding window

BUse response chaining with a fixed number of previous responses, relying on the conversation history to preserve the enrollment field values.

CUse a selective summarization of older low-value conversation turns.

DUse conversation-scoped, server-managed history with automatic input truncation.

Answer : A

The application must preserve the latest authoritative value of each enrollment field even when the user corrects an earlier value, while also placing a firm bound on prompt size. A session-scoped enrollment-state snapshot provides deterministic structured state: each correction overwrites the previous field value. A recent-turn sliding window then preserves enough conversational context for natural interaction without resending the full 15-turn transcript. Response chaining with a fixed number of prior turns can lose an important field once it falls outside the retained history. Summarization can omit or distort corrected values, and automatic truncation likewise offers no guarantee that the latest value of every required field survives. Because the state is scoped only to the active enrollment and is not written to long-term memory, the design also avoids durable cross-session memory. Therefore A is correct. At implementation time, the same rule should be expressed through the framework or service configuration rather than left only as a natural-language convention. That makes the behavior repeatable across runs, easier to test, and less sensitive to model variability.

Official Microsoft reference: Microsoft Agent Framework - workflow state and context management

Question 6

You have a Microsoft Foundry multi-agent solution for loan applications. Each agent scores a full application independently and does NOT require output from other agents.

You need to recommend an orchestration pattern that meets the following requirements:

Produces one aggregated recommendation

Preserves independent scoring -

Minimizes end-to-end latency -

Minimize development effort -

What should you recommend?

Agroup chat

Bmagnetic

Csequential

Dconcurrent

Answer : D

The scoring agents do not depend on each other's output, so their work should be fanned out in parallel and aggregated afterward. Microsoft Agent Framework concurrent orchestration is intended for independent participants that can process the same input simultaneously. That minimizes end-to-end latency because completion time approaches the slowest individual scorer instead of the sum of all scorers. The concurrent workflow also provides a fan-in stage that can aggregate the separate scores into one recommendation without requiring a complex custom conversation protocol. Sequential orchestration wastes time by serializing independent work. Group chat and Magentic-style collaboration introduce unnecessary coordination and planning overhead when the agents simply need independent scoring. Therefore D, concurrent, is the simplest and fastest orchestration pattern. In production, add telemetry and regression tests around this behavior so changes to prompts, models, tools, or orchestration do not silently alter the intended contract. The selected approach is the one that best matches the platform's native execution semantics.

Official Microsoft reference: Microsoft Agent Framework - Concurrent orchestration

Question 7

You have a Microsoft Foundry project that contains a multi-agent customer support solution. The solution includes a final response agent.

You need to provide reviewers with the ability to assess agent responses in the preview web app and record a completion score for each agent response. The solution must follow the principle of least privilege.

Which roles should you assign to the reviewers?

Aonly Application Insights Reader on the Foundry account

Bonly Foundry User on the project

CFoundry User on the project and Reader on the Foundry account

Donly Foundry Project Manager on the project

Answer : C

Microsoft's human-evaluation guidance for the Foundry preview web app requires reviewers to have Foundry User permissions on the project plus Reader access on the Foundry account/resource. That combination gives the reviewer enough project-level access to open the agent experience and record evaluation feedback while avoiding project-management or account-management privileges. Application Insights Reader alone only grants telemetry visibility and does not provide the Foundry interaction permissions needed to score responses. Foundry User alone does not satisfy the documented account-level visibility requirement for this preview experience. Foundry Project Manager would be broader than necessary and violate the least-privilege objective. Therefore C is the documented reviewer role combination. The same configuration should be paired with auditable identity, trace, and evaluation data so reviewers can prove which principal acted, which policy was applied, and why a request was allowed or blocked. That is particularly important for production multi-agent systems with external tools. Least privilege remains the governing principle: grant only the identity, data, tool, or deployment access required for the specific operation. The selected answer preserves that boundary while still allowing the workflow to satisfy its functional requirement.

Official Microsoft reference: Microsoft Foundry - Human evaluation