Microsoft Certified: Agentic AI Business Solutions Architect Expert
Independent: not affiliated with Microsoft. Free to try; the study bundle adds a study guide and two practice exams.
5 original questions, none of them in the practice exams.
Pick an answer. Show answer explains every option and links its Microsoft source.
5 questions / Explanations link Microsoft sources / Last updated Oct 26, 2026
Question 1
Multiple choice
Domain 1 / Task 1.2 Design overall AI strategy
A manufacturer's inside-sales team receives about 2,000 inbound leads a month in Dynamics 365 Sales and follows up on fewer than a third of them. Leadership wants every lead researched against the company's ideal customer profile and product list, an initial outreach email sent, and only leads that show purchase interest handed to a seller, using a prebuilt agent rather than a custom build. Which agent should the architect propose?
Show answerHide answer
Answer: A
Why
The Sales Qualification Agent researches inbound leads, sends outreach emails and, in research and engage mode, hands leads that show purchase interest and profile fit to sellers under configured handover criteria. It is a prebuilt agent, so no custom build is needed.
Why not the others
BThe Sales Opportunity Agent researches opportunities, surfacing risks and promising deals, rather than qualifying inbound leads.
CThe Sales Close Agent runs the sales cycle to closure with product recommendations and objection handling; the requirement is qualification and handover to a seller.
DThe Recommended Actions Agent prioritizes tasks for sellers on their opportunities; it does not research leads or email them.
A regional airline is launching a customer-facing Copilot Studio agent on its public website before the holiday season. Forecasts range from 2,000 to 40,000 conversations a day, finance will not pay in advance for capacity that may go unused, and the business requires that the agent is never switched off for billing reasons during the season. Which billing design should the architect recommend?
Show answerHide answer
Answer: B
Why
Copilot Credits are billed through prepaid capacity or a pay-as-you-go meter. Under prepaid capacity, enforcement starts when the tenant reaches 125 percent of its capacity and custom agents are disabled. With pay-as-you-go, Microsoft states that enforcement does not apply because overage is billed to the Azure subscription, and nothing is paid in advance, which fits unpredictable public traffic.
Why not the others
AA per-agent monthly consumption limit caps credit usage before enforcement; it controls spending rather than funding unpredictable usage.
CPrepaid capacity is enforced at 125 percent, where custom agents are disabled, so the overage allowance is a grace period rather than a billing model, and capacity sized to the peak is paid for whether or not it is used.
DTenant graph grounding is a billed feature at 10 Copilot Credits per message for users without a Microsoft 365 Copilot license, so it adds cost rather than removing it.
A retail chain's Copilot Studio agent helps the expansion team evaluate candidate store sites. Most requests are simple lookups, such as lease status and nearby stores, that must stay fast. One step must weigh demographics, lease terms and logistics data from three sources and recommend a single site with a justification; reviewers say the default model's recommendations are shallow. The agent uses generative orchestration. What should the architect recommend?
Show answerHide answer
Answer: C
Why
Copilot Studio can add a deep reasoning model (preview) to an agent with generative orchestration: turn on Deep reasoning in the agent's settings, and the agent applies it to steps whose instructions contain the keyword reason. The model is slower, so scoping it to one step keeps the lookups fast.
Why not the others
AEach step marked reason uses the slower model, and Microsoft states that using it for multiple steps can substantially slow the agent's responses, which the lookups cannot afford.
BRebuilding on Foundry replaces the whole agent to fix one step; Copilot Studio can apply a deep reasoning model to a single step.
DDeep reasoning requires generative orchestration, and a multi-source judgment is not a deterministic topic with condition nodes.
A distributor configured the Sales Opportunity Agent in a Dynamics 365 Sales sandbox environment, tuned its selection criteria and knowledge sources, and validated it with a pilot team. The administrator plans to export the agent as a managed solution and expects production to need only the import. Which two points should the architect raise about the ALM process? Choose two.
Show answerHide answer
Answers: A and E
Why
Microsoft's import procedure for the Sales Qualification, Sales Opportunity and Sales Close agents states that the Dataverse application user and the required application registrations are environment-specific and are not included in the exported solution, and that the agent is always imported in draft state: after import you select the application user, review assignment rules or selection criteria, review the knowledge sources, then save and start the agent. Agents are deployed to sandbox or test environments before production, and Microsoft recommends organization-agnostic filters in selection criteria so profiles keep working after migration.
Why not the others
BMicrosoft's export steps choose a managed solution for production; assignment rules are reviewed in the imported agent's settings, not by unmanaged edits.
CKnowledge sources appear after import when they were added to the corresponding agents in the source environment; they are reviewed and repointed where the content differs, not rebuilt.
DThe best practice is one solution per agent profile, because bundling profiles can export unintended or untested changes.
A hospital group runs an agent on Microsoft Foundry that answers clinicians' questions from an Azure AI Search index of internal protocols. Security policy now requires that neither client calls to the Foundry endpoint nor the agent's calls to its grounding data cross the public internet. The existing Foundry resource was created with public network access enabled and no virtual network. What should the architect recommend?
Show answerHide answer
Answer: A
Why
Inbound isolation is public network access set to Disabled with a private endpoint; outbound isolation injects the agent into a subnet of the customer's virtual network, so its calls to Azure AI Search stay private. Virtual network injection cannot be added to an existing Foundry deployment, so the resource must be redeployed.
Why not the others
BOutbound virtual network injection cannot be added to an existing Foundry deployment, so the agent's outbound calls cannot be isolated without redeploying the resource.
CThe selected-networks setting filters inbound access by IP address, so client traffic still uses the public endpoint, and the agent's outbound calls are not isolated without virtual network injection.
DSharePoint grounding is a public-endpoint tool that communicates over the public internet, which the policy forbids.
The study bundle has two practice exams with 120 original questions, an answer key and an explanation for every answer, plus the 58-page study guide. Practice exam explanations are shorter than the ones on this page: a short paragraph per question, with no source links.
Study bundle
$19
120 practice questions / Study guide / Answer keys
Every bundle includes 6 months of access: download the PDFs, and any new editions, until 6 months after your purchase date. Your receipt has a license key. Enter it at overqual.com/download for the three PDFs, stamped with your email and order number. They are yours to keep.