This is what I am looking for, similar to OpenAI with custom GPTs, but the declarative agents seem more promising in terms of freedom regarding creation and innovation.
I still don't get it. Why adding this complexity of Copilot agents? For me, the goal of having Copilot for M365, is to propose to all users/employees of our company a single entry point where they can ask (almost) everything and we use connectors/API Plugins to fetch the relevant information and display them nicely to our users. If we start to create agents for each need, then for big companies, it will become a mess and users will be lost because, before asking a question, they will have to start to ask themself: which Copilot agent will help me to get my answer.. It makes no sense. Why can't we simply extend Copilot for M365 and add connectors/API plugins for all users in our company? Why can't we manage activation of these plugins for all users from the Admin Center (and not ask them to active these plugins manually)? I know that Copilot topic is quit new and Microsoft still have a lot of work to do, but simplicity should be the goal for users and employees, not complexity.
Speaking from experience more than official positioning, I will say I've noticed serious usability issues with stand-alone plugins. A user asks about her projects, for example, and there are projects in documents, plugins, Planner, etc. - so you have to word your prompt very specifically, much as you would have to you always spoke with just one person who tried to be an expert on all subjects. Users understand the idea of going to an expert depending on what they want; those are the agents. This also ensures requests won't be sent to the wrong back-end (privacy concerns) and old random documents won't be selected (since declarative agents' can specify the authoritative documents to be included.) Note that OpenAI also stopped supporting stand-alone plugins I suspect for the same reason. OpenAI moved them to inside GPTs now, and though Declarative agents are different than GPTs they both have the effect of providing a more bounded and understood context. - Bob
@@Microsoft365Developer Thanks Bob for your share, I really appreciate it. It makes sense and I also suspected the same reason for the introduction of these declarative agents. I also confirm that we struggled a lot with stand-alone plugins :)
Do custom agents work with Word, Powerpoint and Excel Copilot 365?
This is what I am looking for, similar to OpenAI with custom GPTs, but the declarative agents seem more promising in terms of freedom regarding creation and innovation.
One of the requirements for this hands on labs is to have copilot license. How can do the lab without paying £300+ for the copilot license?
I still don't get it.
Why adding this complexity of Copilot agents?
For me, the goal of having Copilot for M365, is to propose to all users/employees of our company a single entry point where they can ask (almost) everything and we use connectors/API Plugins to fetch the relevant information and display them nicely to our users.
If we start to create agents for each need, then for big companies, it will become a mess and users will be lost because, before asking a question, they will have to start to ask themself: which Copilot agent will help me to get my answer.. It makes no sense.
Why can't we simply extend Copilot for M365 and add connectors/API plugins for all users in our company? Why can't we manage activation of these plugins for all users from the Admin Center (and not ask them to active these plugins manually)?
I know that Copilot topic is quit new and Microsoft still have a lot of work to do, but simplicity should be the goal for users and employees, not complexity.
Speaking from experience more than official positioning, I will say I've noticed serious usability issues with stand-alone plugins. A user asks about her projects, for example, and there are projects in documents, plugins, Planner, etc. - so you have to word your prompt very specifically, much as you would have to you always spoke with just one person who tried to be an expert on all subjects.
Users understand the idea of going to an expert depending on what they want; those are the agents. This also ensures requests won't be sent to the wrong back-end (privacy concerns) and old random documents won't be selected (since declarative agents' can specify the authoritative documents to be included.)
Note that OpenAI also stopped supporting stand-alone plugins I suspect for the same reason. OpenAI moved them to inside GPTs now, and though Declarative agents are different than GPTs they both have the effect of providing a more bounded and understood context.
- Bob
@@Microsoft365Developer Thanks Bob for your share, I really appreciate it. It makes sense and I also suspected the same reason for the introduction of these declarative agents. I also confirm that we struggled a lot with stand-alone plugins :)
❤
0:01