Copilot Agent and Agent for Copilot for Microsoft 365

Hi Everyone,

I notice in Copilot Studio, we have our Built Copilot Agent that we can feed the knowledge for a chatbot.

I notice as well in Copilot studio, we have Copilot for Microsoft 365 which then we can built an agent under Copilot for Microsoft 365
Question, What is the difference between Normal Copilot Agent vs Agent under Copilot for Microsoft 365

What i understand is Agent under Copilot for Microsoft 365 is supposed to be extension of Copilot for Microsoft 365 (Personal Assistance plus Customized knowledge), however, the Agent under Copilot for Microsoft 365 cannot read our personal information such as emails, teams chat etc. So, what's the intention of it and what is the difference? I can't seem to find the comparison anywhere.

Also can I recommend that Agent under Copilot for Microsoft 365 to be renamed to maybe Co-Agent since the name is very confusing.

Below are the difference that I've identified but can i get someone from Microsoft or some expert to validate and explain a bit more

	Copilot for Microsoft 365	Agent under Copilot	Copilot Agent
Description	Personal Assistant	Extension of Copilot	Chatbot for Others
Autonomous Capability	No	No	Yes (Public Preview)
Topic Configuration	No	No	Yes
Knowledge	General (GPT)+ Intranet and Personal	Configured (Web and Custom)	Configured (Web and Custom)
Access to Personal Data (Outlook,Teams)	Yes	No	No

Categories:

General

Hello,

the main difference between the two is that, as you've said, M365 Copilot agent is an extension to M365 Copilot, while agent created in Copilot Studio is a standalone entity.

The first one can be used only within the context of M365 Copilot: let's say you'd like to feed the M365 Copilot with information from a public website or Sharepoint site. You can create an agent which acts as a separate knowledge source - you can ask it about the datasources that it has defined. You can also create separate agents for various purposes like: writing blog posts, creating text content, helping in writing code etc. It's good to perfom it this way, because you can scope the functionality for each agent and scoping makes it better at generating more relevant responses.

The second one gives the ability to create a separate, standalone bot, which can connect to anything that's currently defined. You can add it to various channels and it can communicate with anyone inside or outside the organization.

I think this comparison is better: