Design AI-powered business solutions
Free Microsoft Certified: Agentic AI Business Solutions Architect practice — 6 questions on Design AI-powered business solutions, with explanations. No sign-up. Full 12-question mixed test →
A solution architect is designing a multi-agent system in Microsoft Foundry Agent Service for a loan-approval workflow. The process requires: (1) an intake agent extracts application data, (2) a credit-risk agent and a compliance agent must each independently review the same extracted data in parallel without influencing each other's analysis, and (3) a decision agent synthesizes both independent reviews before responding to the applicant. Which orchestration pattern should the architect use for steps 2 and 3?
Concurrent orchestration fans a task out to multiple agents that run independently and in parallel, then aggregates their outputs in a downstream step — exactly matching the requirement that risk and compliance reviews must be independent and later synthesized.
An architect building an agent in Microsoft Foundry needs it to invoke a third-party SaaS agent — built on a different vendor's agent framework — that publishes an agent card describing its skills, so the Foundry agent can discover its capabilities and delegate a shipment-tracking task to it at runtime. Which open standard should the architect use for this cross-vendor agent-to-agent interaction?
A2A is the open standard purpose-built for agent-to-agent interoperability across vendors and frameworks; it defines the Agent Card for capability discovery and a standardized task-delegation lifecycle between independent agents.
A company is building a Copilot Studio agent to answer employee questions about HR policy, sourced from thousands of documents in SharePoint, and also to return each employee's exact remaining PTO balance, which is stored and updated in real time in Dataverse. Employees need both cited narrative policy answers and precise, current numeric PTO balances. Which knowledge and tool design BEST meets these requirements?
Unstructured policy content is well suited to generative-answers grounding with citations, while exact, real-time numeric data requires a deterministic tool call against the system of record (Dataverse) rather than LLM inference over documents — combining both meets accuracy and currency requirements.
An architect has enabled generative orchestration on a Copilot Studio agent so the AI can dynamically choose which topic or action to invoke based on conversation context. However, one highly regulated topic — 'cancel account' — must always be triggered strictly by its defined trigger phrases and never dynamically selected by the orchestrator, even though generative orchestration remains enabled for the rest of the agent. What should the architect configure to achieve this?
Copilot Studio allows configuring individual topics with a trigger-phrase-only orchestration setting, which locks that specific topic to deterministic classic invocation while leaving generative orchestration active for all other topics in the same agent.
An architect must build a custom multi-agent orchestrator with fine-grained control over agent state and custom code-based reasoning, using the Semantic Kernel Agent Framework's Magentic orchestration pattern to dynamically plan tasks across five specialized agents. Low-code maintainability by business users is explicitly not a requirement. Which platform should the architect use to build this orchestrator?
Microsoft Foundry Agent Service supports building agents with the Semantic Kernel Agent Framework, which natively implements orchestration patterns including Sequential, Concurrent, Group Chat, Handoff, and Magentic — giving the required programmatic, fine-grained control.
A retail company's Copilot Studio customer service agent is integrated with Dynamics 365 Customer Service. When the agent detects strong customer frustration or an explicit request for a human, it must transfer the full conversation context, transcript, and any related case record to a live agent through Omnichannel, without requiring the customer to repeat information already provided. Which capability should the architect design into the solution?
Conversational escalation/handoff to Omnichannel for Customer Service is designed specifically to transfer live conversation context and case data to a human agent in real time, avoiding repeated information and preserving continuity.
Ready for the real thing?
The full course has two full-length practice tests, video lessons for every exam domain, hands-on labs and detailed answer explanations.
Start my full course on Udemy →