What's changed: Added per-section figures (cert-figure-retrofit). New AB-900 Chapter 3 (Domain 3 "Copilot and agent administration": capabilities = Copilot vs agents/monthly-PAYG/enable-disable/Researcher-Analyst-custom agents; Copilot admin = license assignment/PAYG billing policies/Copilot Analytics usage-adoption/prompt management; agent admin = user access config/creation/approval process/monitoring via M365 and Power Platform admin centers)
3.1Features of Copilot and agents
Understand the difference between Copilot and agents’ built-in capabilities, Copilot license models (monthly vs pay-as-you-go including SharePoint), which features can be enabled/disabled, and use cases for Researcher, Analyst, and custom agents.
Microsoft 365 Copilot is an AI assistant embedded in apps like Word, Excel, Outlook, and Teams. Agents are customizable AIs specialized for specific tasks or knowledge that work semi-autonomously, extending Copilot. Admins must understand their capabilities, licensing, and enablement scope to deliver them.
3.1.1License models: monthly vs pay-as-you-go
Copilot has two billing models: a monthly license (a fixed per-user monthly fee for Copilot broadly) and pay-as-you-go (PAYG) (charged by usage, applied to specific uses such as SharePoint agents). Monthly suits steady, heavy users; PAYG suits variable or limited usage. Admins choose based on target users and scenarios.
3.1.2Researcher, Analyst, and custom agents
- Researcher: a built-in agent for deep research (investigate, summarize, analyze) across multiple sources.
- Analyst: a built-in agent that supports analyst-style tasks such as data analysis and visualization.
- Custom agents: agents you build specialized for org-specific knowledge or processes (e.g., internal FAQ, request workflows).
Admins can enable/disable certain Copilot features (adjusting what is offered to match org policy) and control agent availability. Governing "which features to whom" is central to safe deployment.
| Item | Characteristic | When to use |
|---|---|---|
| Copilot (built-in) | AI assistant in apps | Everyday productivity |
| Agents | Specialized, semi-autonomous | Specific tasks/knowledge |
| Monthly license | Fixed monthly | Steady, heavy use |
| Pay-as-you-go (PAYG) | Charged by usage | Variable/limited use (e.g., SharePoint) |
| Researcher | Deep research | Investigate/summarize/analyze |
| Analyst | Data analysis | Analyze/visualize |
Watch the mix-ups: (1) Copilot (general assistant embedded in apps) vs agents (specialized, semi-autonomous, customizable). (2) monthly (fixed) vs PAYG (by usage; SharePoint, etc.). (3) Researcher (deep research) vs Analyst (data analysis)—don’t confuse use cases.
Common: requirement → feature/model. E.g., "do deep cross-source research" = Researcher; "support data analysis" = Analyst; "build an AI specialized in internal knowledge" = custom agent; "variable usage, pay per use" = PAYG; "everyone uses it constantly, fixed monthly" = monthly license.
3.1.3Section summary
- Copilot = general AI embedded in apps; agents = specialized, semi-autonomous, customizable
- Licensing: monthly (fixed) vs pay-as-you-go (by usage; SharePoint, etc.)
- Researcher (deep research) / Analyst (data analysis) / custom agents (org-specific)
Sign in to track progress — Log in.
Quick check
(just a quick review)Q1. Which best distinguishes Copilot from agents?
Q2. Which model charges "by usage" for variable use or specific uses like SharePoint agents?
Q3. Which built-in agent performs deep research (investigate, summarize, analyze) across multiple sources?
Q4. Which built-in agent supports analyst-style tasks such as data analysis and visualization?
Q5. To deliver an AI specialized in org-specific knowledge/processes (e.g., internal FAQ, request workflows), what do you create?
Q6. What can admins do to adjust Copilot’s scope to match org policy?