What is a single-agent solution?
A single-agent solution is one AI agent designed to handle a task by using:
A model
Instructions
Optional tools
Optional knowledge/data
A test chat/playground interface
Think of it as:
User request → Agent → Model + instructions/tools → Response/action
Typical Foundry portal workflow
1. Open the Foundry portal
You start in a Foundry project. Microsoft’s agent quickstarts describe creating or selecting a Foundry project before creating an agent.
2. Create an agent
In the portal, you create an agent and give it a name. Microsoft’s current hosted-agent quickstart says agents are found under:
Build → Agents
3. Configure the agent
You normally configure:
Configuration Purpose
Agent name Identifies the agent
Model The AI model the agent uses
Instructions Defines behaviour, role, tone, and limits
Tools Allow the agent to perform actions
Knowledge/data Grounds answers in provided information
4. Test the agent in the playground
The playground lets you chat with the agent before using it in an application. You enter test questions, review the answers, and refine the instructions.
5. Refine the agent
You improve the agent by adjusting:
Instructions
Prompts
Tools
Grounding data
Model settings
Exam example
Scenario:
A company wants an AI assistant that answers employee questions about HR policies in a controlled tone.
Correct approach:
Create an agent in Foundry portal
Choose a model
Add instructions
Connect relevant HR policy content
Test in playground
Refine responses
Key exam distinction
Requirement Correct Foundry feature
Manually test an agent Foundry playground
Define agent behaviour Instructions/system prompt
Let agent perform actions Tools
Make answers use company data Grounding/knowledge
Build app integration SDK/client application
Summary
For AI-901, remember:
Create agent → configure model/instructions/tools/data → test in playground → refine.