Microsoft AI-500 Practice Exam Questions & Answers

Each customer request requires three independent agent runs. Dequeuing ten requests and launching a concurrent three-agent workflow for each request creates at most 30 simultaneous agent runs, exactly meeting the concurrency ceiling. Because the three agents execute in parallel and one run takes about five seconds at the 95th percentile, ten requests can complete in roughly five seconds, which corresponds to 120 requests per minute. Microsoft Agent Framework's ConcurrentBuilder performs a fan-out to independent participants and a fan-in that can produce one consolidated response, solving the downstream partial-message problem. Sequential workflows would require roughly three agent durations per request and miss the throughput target. A 90-participant workflow violates the concurrency limit. Therefore C is the only option that satisfies the numerical and orchestration constraints together. The implementation should also preserve clear inputs and outputs around this step so that later agents receive only the information they require. This improves debuggability and keeps token, permission, and state growth under control as the workflow becomes more complex.

Official Microsoft reference: Microsoft Agent Framework - Concurrent orchestration

The updated guardrail must be evaluated with known benign and adversarial cases so false positives and false negatives can be measured explicitly. Running labeled cases through every relevant workflow path also tests whether the guardrail is applied consistently at the intended intervention points. Microsoft AI-500 guidance includes guardrail testing with synthetic or curated data and emphasizes evaluation before production rollout. Playground-only spot checks are too narrow to establish coverage. Switching production to annotate-only would change live behavior and use customers as the test population, violating the requirement to avoid production changes. A compliance configuration review verifies that a policy exists but does not measure whether it correctly detects or misses real inputs. Option B is therefore the only approach that produces repeatable evidence of intervention accuracy and policy coverage without changing production behavior. A robust evaluation program separates process metrics from final-response metrics. The selected answer measures the layer where the stated failure actually occurs, which is essential for deciding whether to change retrieval, orchestration, prompt behavior, or the final generator.

Official Microsoft reference: AI-500 Study Guide - guardrail testing and evaluation

Both stated restrictions concern content that must not appear in the agent's final output. A custom blocklist should therefore be applied to agent output so the two sanctioned company names are detected before a response is returned. The customer-account requirement likewise belongs at the output boundary, where PII/sensitive-data detection can identify protected identifiers in generated content. Applying the controls only to user input would not stop the model from generating the prohibited names or identifiers itself. Task Adherence focuses on whether agent behavior stays within assigned procedures and is not a substitute for explicit output filtering. Microsoft Foundry guardrails are designed to attach controls to the intervention point where the relevant risk occurs. Because the risk is disclosure in the response, option C is the configuration that places both controls at the correct boundary. 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.

Official Microsoft reference: Microsoft Foundry guardrails - intervention points

The business requirement says that a physician must approve any triage assessment that recommends an emergency room visit. The workflow must therefore pause and collect a human response before continuing. Ask a question is the option that introduces that interactive human-in-the-loop step. Deliver a message only informs someone; it does not capture a decision. Agent invokes another AI component, which would not satisfy the explicit requirement for a physician's approval. Go to changes the workflow path but likewise does not obtain human authorization. Microsoft Agent Framework documents human-in-the-loop request/response patterns for cases where execution must wait for approval or additional human input. For a medical escalation decision, the control must be deterministic and auditable rather than encoded as a prompt suggestion. Consequently, Ask a question is the workflow node that best enforces the business requirement. 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 - Human-in-the-loop workflows

The corpus already contains reliable document hierarchy in Markdown and HTML, so the ingestion pipeline should preserve those author-defined boundaries rather than infer new ones from token counts or topic shifts. Format-specific header splitters can divide Markdown by heading levels and HTML by structural headers, producing chunks that align with meaningful sections. This directly solves the current problem of fixed-size token chunks crossing section boundaries and requires less custom logic than building a semantic topic-shift chunker. Recursive character splitting can be configured with structure-aware separators, but it remains a more generic fallback when format-specific structure is already available. Microsoft Azure AI Search guidance recommends exploiting document structure such as headings when chunking. Therefore B is the most direct and maintainable approach. 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: Azure AI Search - structure-aware chunking and Markdown indexing

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

GitHub issues are the authoritative source for current requirements, and those requirements can change while a long-running workflow is still executing. Each handoff should therefore reconstruct context from the current issue state and discard memory that has been invalidated by newer requirements. Keeping an accumulated summary allows obsolete criteria to remain present and also causes the prompt to grow until older content is truncated. Persisting the transcript in Cosmos DB preserves history but does not make that history current. Giving subagents a GitHub lookup tool while continuing to inject the stale summary can expose the model to conflicting versions of the same requirement. Microsoft Agent Framework context providers are intended to inject current external context at invocation time, making C the approach that simultaneously preserves freshness and bounds context growth. 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 - Context providers