GregLab | Exam Prep

Drive business outcomes by using agents

Prebuilt Copilot Agents in Business Workflows

Core

Recognize where prebuilt agents such as Researcher and Analyst improve a workflow, decide between a prebuilt and a custom agent, use agents from different Copilot experiences, and find agent chats again.

Aligned to the AB-730 skills measured as of October 20, 2026; product behavior verified November 13, 2026.

Why this matters

The October 20, 2026 AB-730 blueprint moved agents from building and configuring to selecting and applying them. A business user is expected to know which prebuilt agent fits a task, when a custom agent is really needed, and how to reach agents where the work happens.

Must Know

  • Agents are specialized AI assistants for specific tasks and workflows. They can follow defined instructions, use selected knowledge sources, and respond in a consistent format.
  • Use an agent when you need consistent, specialized help for a repeated task. For one-time questions or general research, use Copilot Chat without an agent.
  • Researcher runs multi-step research across internal work content and public web sources, can ask clarifying questions, and returns a structured report with citations.
  • Analyst analyzes attached files such as Excel, CSV, JSON, PDF, and Word, reasons across them to find trends and outliers, and returns a report with charts, tables, and cited findings.
  • The Agent Store lists agents from Microsoft, partners, and your organization. Microsoft-built options include agents such as Writing Coach, Prompt Coach, and Visual Creator. Review an agent's description, knowledge sources, and availability before adding it.
  • Create a custom agent only when the task needs instructions, knowledge sources, or response formats tailored to a specific process that no available agent provides.
  • Agents can be reached from several experiences: Copilot Chat and the Microsoft Copilot app, the + menu in the Word Copilot pane, and @mentions in Copilot prompts in Teams meetings.
  • You can filter Copilot Chat history by agent to find and resume earlier conversations with a specific agent.

Compare and Distinguish

  • Prebuilt agent vs custom agent: reuse an agent that already fits the task vs create one for an unmet, repeatable need with its own knowledge and instructions.
  • Researcher vs Analyst: discovery and synthesis across work and web sources vs data analysis across attached files.
  • Agent vs Copilot Chat: scoped, consistent behavior for a repeated task vs open conversation for one-off work.
  • Agent vs Cowork: task-specific help within a defined scope vs a handed-off goal that spans several tasks across apps.
  • Filter chats by agent vs rename a chat: find all conversations with one agent vs change one conversation's title.

Scenario examples

  • Scenario: A manager needs a market analysis that combines competitor moves from the web with internal plans. Think: Researcher.
  • Scenario: A finance team must compare regions across several Excel and CSV files and present cited findings. Think: Analyst.
  • Scenario: An Agent Store agent covers the right topic but cannot use the team's approved sources or escalation rules. Think: confirm the gap, then create a custom agent.
  • Scenario: During a Teams meeting a manager wants a deeper look at sales figures. Think: mention @Analyst in a Copilot prompt in the meeting.

Exam traps

  • Matching a subject name is not enough; an agent fits only if it meets the task's knowledge and behavior requirements.
  • Creating a custom agent because one response was weak is not justified; first refine the prompt or choose a better-fitting agent.
  • Researcher and Analyst require a Microsoft 365 Copilot license, and organizations control which agents users can discover, use, create, and share.
  • Renaming an agent or a chat does not change what the agent can do.

Key takeaways

  • Check for a prebuilt agent that fits before creating one.
  • Researcher researches; Analyst analyzes data.
  • Use agents where the work is, and filter chats by agent to pick up where you left off.
How it works
  • You select an agent, describe the task, and the agent applies its instructions and knowledge to produce a scoped result.
  • Researcher may ask clarifying questions before researching; Analyst reads the attached files and reasons through the data before reporting.
Objects and administrative surfaces
  • Agent Store in Microsoft Copilot: discover agents from Microsoft, partners, and your organization.
  • Researcher and Analyst: Microsoft agents in the Microsoft Copilot app, available from the left navigation or the agent list.
  • Agent Builder: creates a custom agent when no available agent fits; covered here only as the alternative to a prebuilt agent.
  • Chats list in Copilot Chat: can be filtered by agent.
When to use it
  • Use Researcher for questions that need research across sources and a cited report.
  • Use Analyst for analysis across attached data files that should end in a shareable report.
  • Use an organization's agent for a repeatable task that matches its purpose.
Security and governance implications
  • Agents work within the user's permissions and the organization's agent settings.
  • Review an agent's description and knowledge sources before relying on it for a business process.
How to reason about this
  • If an agent's answers miss required sources or rules, confirm the gap before deciding a custom agent is needed.
  • If you cannot find an earlier agent conversation, filter the chat history by that agent.
More detail
  • Researcher is optimized for discovery, synthesis, and insight generation, and it can deliver content ready to use in Word, PowerPoint, or Outlook.
  • Analyst applies chain-of-thought reasoning across multiple files at once and supports tasks such as comparing regional performance, forecasting trends, or summarizing feedback patterns.
  • Agent availability depends on the agent's knowledge sources, licensing or billing, and the organization's settings.
  • Follow-up prompts can reshape an agent's report for different audiences, such as three bullets for a leadership slide or an executive summary, without re-running the research.

Ready for the quiz?

  • Which clues in a scenario point to Researcher rather than Analyst?
  • What requirement justifies a custom agent instead of a prebuilt one?
  • Name two places where you can start or mention an agent without leaving your work.
  • How do you find last week's conversation with a specific agent?

Related objectives

  • D4.1.a — Identify scenarios where a prebuilt agent can improve a business workflow
  • D4.1.b — Choose between using a prebuilt agent or creating a custom agent for a business task
  • D4.1.c — Apply an agent to a business process from within various Microsoft 365 Copilot experiences
  • D4.1.d — Filter chats by agents

Learn more

Free Microsoft Certified: AI Business Professional prep

Build focused AB-730 quizzes from skill areas, topics, and product references.

Practice with exam-style multiple-choice and multiple-response questions, score breakdowns, explanations, and a compact reference for this lane's official exam domains.

Read Topics Build a quiz

Exam Weights

Quiz builder

Choose your practice set

Mode

Exam fidelity: Microsoft does not identify the specific item types that will appear on an individual exam before delivery. This lane counts multiple-choice and multiple-response items as exam-style practice. Ordering and matching are supplemental learning exercises and do not count toward exam-style accuracy. Difficulty labels are calibrated to AB-730 Fundamentals-level business-user scenarios, not a Microsoft-published question rating.

Reference

AB-730 topics and reference map

Study links

AB-730 resources