AI-103 · Practice set 7 of 13
Matching Foundry Services to Tasks: 10 practice questions
10 free AI-103 practice questions on Matching Foundry Services to Tasks, with an explanation for every answer. Untimed. The full mock exam and the timed version are in the app.
-
Question 1 of 10
A team only needs to send a prompt to a model and display the reply, with no tools. Where should it start?
- AA toolbox endpoint that exposes the model as a tool
- BA model call, adding an agent only if needed
- CA workflow that orchestrates the prompt and the reply
- DA hosted agent so the call runs in a managed runtime
Show the answer
Match the build to the need. Sending a prompt to a model is the shortest path to a working call, and you add an agent only when tools or orchestration become necessary. Starting with agents or workflows adds parts this task doesn't use.
Next → 1 / 10 -
Question 2 of 10
A model gives weak answers because it has never seen the company's current product data. Which lever does the capability map recommend first?
- ARetrieval of the missing context
- BFine-tuning on the product data
- CThe model router, for a better model
- DA workflow that splits the question
Show the answer
Fine-tuning and retrieval fix different problems. When the model simply lacks the context, retrieval supplies it at query time. Fine-tuning is the lever when prompting and retrieval still can't reach the quality you need.
Next → 2 / 10 -
Question 3 of 10
An agent must answer questions about announcements made this week on public websites. Which grounding capability fits?
- AMemory across user sessions
- BFile search over the uploaded files
- CWork IQ over Microsoft 365 data
- DWeb search and Grounding with Bing
Show the answer
Choose grounding by where the content lives. Fresh public information lives on the web, and Web search and Grounding with Bing ground answers in current web results.
Next → 3 / 10 -
Question 4 of 10
One agent must answer from three PDF files that a user provides. The team wants the quickest setup, and no other agent needs this content. Which grounding choice fits?
- AFile search over the uploaded files
- BFabric IQ over the PDF content
- CFine-tuning on the three files
- DA Foundry IQ knowledge base over the files
Show the answer
Both file search and Foundry IQ can ground answers, which is why this is a real choice. Foundry IQ is agentic retrieval over a knowledge base, but for only a few uploaded files, file search is the lighter-weight option.
Next → 4 / 10 -
Question 5 of 10
Which Foundry capability provides the managed storage and indexing behind file search?
- AMemory
- BAgent identity
- CRoutines
- DVector stores
Show the answer
File search doesn't need you to build an index. Vector stores provide managed storage and indexing for file search, so the files an agent receives are searchable by meaning without extra infrastructure.
Next → 5 / 10 -
Keep the ones you got wrong
In the app, every question you miss comes back exactly when you’re about to forget it.
-
Question 6 of 10
An organization already runs Azure AI Search indexes over its manuals and wants an agent to query those same indexes. Which capability fits?
- AMemory with a user scope
- BThe image generation tool
- CNew vector stores for file search
- DThe Azure AI Search integration
Show the answer
Reuse what exists. The Azure AI Search integration lets an agent use existing Azure AI Search indexes as knowledge. Vector stores are a separate managed store for files given to file search, so they wouldn't query those same indexes.
Next → 6 / 10 -
Question 7 of 10
A team built an agent with Microsoft Agent Framework, including custom libraries, and wants Foundry to run that code as it is. Which build fits?
- AA hosted agent
- BA workflow of prompt agents
- CA voice-based prompt agent
- DA prompt agent with tools
Show the answer
Ask whether instructions and tools are enough. Here they aren't: the team has its own code and dependencies. Hosted agents run your own agent code or framework with full control, while prompt agents are declarative and have no runtime of your own.
Next → 7 / 10 -
Question 8 of 10
A claims process needs an intake agent, a compliance agent, and a drafting agent, each finishing its stage before the next starts. One agent with more tools can't do it all. Which capability fits?
- AA routine that packages the intake steps
- BMemory so all three agents share their progress
- CA workflow that orchestrates the three agents
- DOne prompt agent with every tool attached
Show the answer
The deciding signal is that one task spans several agents and stages, and a single agent with tools can't finish it. That's exactly when workflows fit: they coordinate multiple agents and steps into one orchestrated process.
Next → 8 / 10 -
Question 9 of 10
A design review lists image generation as a capability for its agent. What kind of capability is it, and what's its status?
- AA preview agent tool, not for production use
- BA generally available Foundry Tool for speech
- CA guardrail feature that filters output images
- DA deployment type for serving image models
Show the answer
Image generation is listed among the tools you attach to an agent, and it's marked preview. Preview capabilities aren't recommended for production workloads, so check its status before you depend on it.
Next → 9 / 10 -
Question 10 of 10
A retailer wants customers to talk with an assistant in real time. Instructions and a few tools are enough, and the team doesn't want to run any speech pipeline. Which option fits?
- AA standard prompt agent with text replies
- BA workflow that chains several text agents
- CA voice-based prompt agent
- DA hosted agent with a custom speech stack
Show the answer
Two constraints decide it: a spoken, real-time experience, and no speech pipeline to host. Voice-based prompt agents speak and listen in real time with managed voice orchestration. A hosted agent could work, but it means running the speech pipeline yourself.
Next → 10 / 10 -
You’ve finished this set
That’s 10 questions on Matching Foundry Services to Tasks. In the app the ones you miss come back exactly when you’re about to forget them.
The whole course, on your phone
Lessons you can read, audio you can listen to on the way to work, and practice that remembers what you got wrong.
“This app is the ‘One Lake’ for preparing the certification exams. Really worth it.”