4.1.3. Researcher, Analyst, and Custom Agents
š” First Principle: Microsoft ships several built-in "premium" agents alongside the base Copilot experience ā Researcher and Analyst ā each designed for a specific deep-work scenario that goes beyond standard Copilot summarization. Custom agents are what IT teams or advanced users build themselves using Copilot Studio for organization-specific workflows.
Researcher: Designed for deep research tasks that require synthesizing information across many sources ā both internal M365 data and the public web (via Bing). Use Researcher when you need a comprehensive brief on a topic, not just a quick summary.
- Example use case: preparing a competitive analysis, researching a prospective client, summarizing a regulatory framework
- Researcher can browse the web in addition to accessing M365 content
Analyst: Designed for data analysis tasks ā particularly working with structured data in spreadsheets and databases. Analyst can run Python code in a secure sandbox to perform calculations, generate visualizations, and identify trends.
- Example use case: analyzing sales data in Excel, identifying anomalies in financial reports, generating statistical summaries
- Analyst doesn't just describe data ā it can run actual computations
Custom agents (built with Copilot Studio):
- Built by IT admins or power users using a no-code/low-code interface
- Can be scoped to specific SharePoint sites, SharePoint lists, or external APIs
- Can be published to Teams channels, SharePoint pages, or the Copilot Chat experience
- Require an approval and publishing workflow before users can access them
| Agent | Built By | Data Sources | Primary Use |
|---|---|---|---|
| Researcher | Microsoft | M365 + Web | Deep research and synthesis |
| Analyst | Microsoft | M365 + Python execution | Data analysis and computation |
| Custom agent | IT / users (Copilot Studio) | Defined at creation | Organization-specific workflows |
ā ļø Exam Trap: Researcher and Analyst are premium agents ā they may require additional configuration or specific licensing tiers to access. They are not automatically available to all Copilot license holders without being enabled by an administrator.
Reflection Question: A finance team wants to use AI to analyze their monthly P&L spreadsheet, identify variances, and generate commentary. Which built-in agent is most appropriate, and what distinguishes it from using standard Copilot in Excel?
