/ Insights / View Recording: Mastering Multi-Agent Systems with Copilot Studio Insights View Recording: Mastering Multi-Agent Systems with Copilot Studio July 31, 2025Orchestrate Multi-Agent Systems in Copilot Studio—Low-Code, Fast, and SecureUnlock the full potential of Microsoft Copilot Studio by learning how to build and manage multi-agent systems that work together to streamline complex workflows and boost productivity. In this session, we’ll explore how to design task-specific agents that collaborate, communicate, and scale across your organization—safely and efficiently.You’ll gain practical insights into use cases, architecture patterns, and governance strategies that enable secure, data-driven AI orchestration. Whether you’re building your first agent or expanding into a network of intelligent copilots, this session will help you confidently move from concept to execution. WHAT YOU’LL LEARNIn this webinar, you’ll learn:How to define and build AI agents in Copilot Studio (low-code vs. pro-code)The difference between child and connected agents—and when to use eachHow to orchestrate multi-agent systems for HR, IT, sales, and website automationIntegration options with Microsoft 365, Azure AI Foundry, and FabricSecurity, governance, and deployment best practices for Copilot Studio agentsHow to quickly deploy agents to Teams, SharePoint, and web appsFREQUENTLY ASKED QUESTIONSWhat is a multi-agent system in Copilot Studio?A multi-agent system in Copilot Studio is a collection of AI agents (child and connected) that work together to automate business processes, answer queries, and perform tasks across departments—all managed through a low-code graphical interface.How do I choose between child and connected agents?Child agents are tightly integrated and deployed with the main agent, sharing authentication and configuration. Connected agents can be managed and deployed independently, ideal for larger teams or when agents need separate governance or access controls.Can Copilot Studio agents integrate with external data sources?Yes. Agents can connect to public websites, SharePoint, Dynamics 365, Salesforce, ServiceNow, and over 1,000 Power Platform connectors. You can also use HTTP calls or custom connectors for additional data sources.What governance and security features are available?Copilot Studio leverages Power Platform’s governance, including data loss prevention (DLP), compliance certifications, geographic data residency, and audit logging via Microsoft Purview. Admins can set guardrails, restrict connectors, and monitor agent activity.How do I deploy Copilot Studio agents to Teams or SharePoint?Agents can be deployed to Microsoft Teams, SharePoint, and web apps with just a few clicks. Teams admins may need to assist with organization-wide deployments. Agents respect user authentication and access controls for secure internal use.ABOUT THE SPEAKERAjay Ravi is a Solution Architect at Concurrency in Chicago, specializing in low-code automation and multi-agent orchestration with Copilot Studio and Microsoft Power Platform. With over six years of experience, Ajay helps organizations streamline business processes, integrate data sources, and deploy secure, compliant AI agents across departments.EVENT TRANSCRIPT Transcription Collapsed Transcription Expanded Ajay Ravi 0:07 Hello everyone. Good morning, good afternoon. So thank you all for joining today’s webinar on multi-agent systems with Copilot Studio. So we’ll be going over a few interesting and new features that is. Is available in Copilot Studio which Microsoft released recently in preview. So we’ll be kind of like going over some of the to do like you know what are the different options you have for MultiAgent systems and then yeah kind of like go from there. So my name is Achi Devi and I am located in Chicago, so working with concurrency for a little over six years, mainly on the low code, no code side of things along with Dynamics Power Platform. In Copilot Studio. So I have worked on a lot of like different projects where we are leveraging low code tools for building our different business solutions, everything. So so yeah, that is a little bit about me and. During my free time, I like to explore new places or like, you know, I like soccer. I was watching a soccer match yesterday in Chicago, so that was fun. So that’s a little bit about me and feel free to follow me in LinkedIn. I’ll happy to, you know, help you guys in any way. OK, so let’s talk a little bit about what are agents, right? So agents in simple form like R, AI or like programs that allow you to do various things for you. So it can be like a simple retrieval information. On defense. Knowledge sources like your public website or you can build automations or like different tasks or activities that agents can do for you. So agents could probably be like a personal Assistant or like a helper for you so that you can delegate a lot of these things and also you can. Use these to automate All various processes and things as well. So it can be like a simple one or like an advanced task automation or like you know business process automation you know agents. So you can kind of like use the Scopilot studio or low code agents. To kind of automate and execute different business process, work alongside yourself and then for your Team and organization as well. So with the whole AI intuition, there are a lot of agents and organizations are heavily using these. Copilot studio and different AI agents approach. So like it is estimated that by 2020 like about 1 billion plus agents will be there. So you need definitely like different skill set or like different business groups who will actively working on these agents like. The process owners, subject matter experts, not just the devs, right? So these are low code and the procode agents that allows you to Bill All these agents for your organization so that way you can have your dedicated agent for your different business processes. So getting into a little bit around like what best fit you need. So there are definitely like two different approaches, right? One is to go with the low code, no code approach or the other one is the traditional procode approach. So it purely depends on like what you’re trying to build and then your specific use cases. So if you are a process owner or a subject matter expert or like a business analyst, then you might not be that much hands on on the procode side of things. So you might potentially lean into some of. The low code side of things where you are using All visual UI, graphical interface and like you know less program to build All these different agents. So there are a lot of different prebuilt and default connectors and then automations and everything that is available in the low code side of things. And then if you are like a dev or a Team lead, so you might probably lean into the procode side of things where you are building your own custom controlled services because you have All those skill set and you would like to have your own DevOps process. So again, depending on the use cases and your specific department or area, you can probably choose between low code and the pro code set of things. So we’re going to be focusing mainly on the Copilot studio which is on the low code side of things and then kind of like deep diving into some of the MultiAgent functionality and what are some of the new things which Microsoft introduced, what you can do, the different types of agents and how you. And build those as well. So just a quick introduction around the Copilot Studio if you have not used already. So Copilot Studio was something which Microsoft introduced a few years back in Ignite and before that there was like the Power Virtual Agent, but there was no. Intelligence or All those orchestration functionality. But with the introduction of Copilot Studio you have the graphical low code tool where you can build All these agents and automations and close to help your business processes, task automation or even kind of like simply. You know, connecting to public websites and then answer queries and All those things. So this is Copilotudio can be like your AI companion that can handle a range of different tasks and functionalities for your business processes. And it can connect with a lot of different knowledge sources or data sources like if you have like a CRM system like Dynamics 365 or Salesforce or if you have ServiceNow if you want to like connect All these different applications or tools and bring your internal data and then. Perform different tasks and operations. So you can definitely do that using Copilot Studio low code, no code agents. So since it is low code and graphical interface, you don’t have to spend a lot of time building from scratch because you already have those, so you just need to incorporate. All those logics and make sure that everything is set up appropriately so the time to market and you can actually build some of these agents pretty quickly, maybe within hours or days depending on the complexity. So it’s a powerful local tool and we’ll be looking at some of the examples. As well. So before I go into some of the multi system agent system details. So in order to kind of get started with the Copilot Studio, you actually have a copilot within Copilot Studio, so with a simple interface like a chat type. You can actually Start building a copilot studio. So like for example I was trying to create a quick copilot agent so you can see that I just provided some basic information that provide what my aim or goal for that agent is and copilot will be able to kind of like. And. Then appropriately add certain instructions and everything in the copilot. So this is how you can kind of like Start spinning up some of these copilot agents and definitely you can Add All those different additional logics and then task automations and everything to make sure that the agent performs All those. Automation as well. So let’s talk about multi-agent orchestration in Copilot Studio. So this is something which is currently in preview mode. So Microsoft introduced the multi-agent functionality where you can actually connect one agent with other agents. So instead of building or spinning up one agent with probably like 10 or 20 or 50 different specific functionalities or tasks, you can actually build each agent design or specific to perform a specific set of instruction and you can do that. And you can actually integrate and connect All these agents with the main agent like in few clicks and we’re going to like take a quick look at that as well shortly. So few things to kind of like consider regarding the whole MultiAgent orchestration so. You know you can enable the agents to kind of like actually talk with each other. So either you can create a child agent that will pretty much follow All the security and authentication and everything from the main Copiler Studio agent, or you can build your own Copiler Studio. And then connect it with the main one. So there are a couple of different ways or options for integrating these agents. So again, you can build All these agents to perform specific tasks dedicated to one single operation or type of operations. We can build one main agent and then creates our child or connected agents like one for the Sales team, one for the Marketing team, one for the HR operations or one for employees. So each agent can have its own dedicated task and operations and then. When you kind of like interact or expose the main agent, it can actually interact with all those sub agents and then I’ll give you the appropriate answers and everything. In addition to that, if you want to do a cross-platform orchestration like for example you have like a specific Copilot studio. That is in prediction and you want to integrate that with a new agent, you can do that. In addition to that, if you want to connect to Azure AI Foundry and then also like M365 Copilot SDK or like Fabric, you can do those things as well. So the Azure AI Foundry and Copilot SDK. So those are currently coming soon, so hopefully it’ll be available shortly. But the Copilot Studio integration and the Fabric agent integration is currently available and you should be able to select that as well. So the goal for the whole multi agent orchestration is that each agent have like a specialization where it is focusing on specific set of activities or things and we’ll be looking at a quick demo as well shortly. So before we get into some of those demo set of things, so this is an example of how Copilot Studio Anatomy or like the agent structure looks like today. So we have like Microsoft have introduced the whole powerful orchestration so it can. Based on users question or details, it can actually look at the different knowledge sources or like the tools that are available and then provide the appropriate response. And in addition to that if you want to build specific topics for narrow approach so you can do that as well. So this is how it is. Looked before, but with the whole introduction of the agent. So this is potentially how it would look like. So within the same Copilot Studio agent, you can Add all those additional agents that lets you do certain designated operations as well. So this way you are building 1 main agent and integrating with us different child agents that let you do certain things for you and you don’t have to like deploy any of these agents separately because everything will be like a child, so you know using these child agents so. It lets you do certain business press related things as well. And in addition to that, we also have the connected agents where it lets you connect to a different Copiler Studio agent or M365 agent as a Foundry fabric data agent as well. So this way you can integrate your single Copiler Studio with a different. The agents as well. So you guys might be thinking like which one should I use, right? So again, that depends on your business use case and scenario. So if you have like a single developer or a small team managing the entire agent solution, then might be like. Agent would be beneficial because you are not working on different agents simultaneously. So in that scenario maybe going for the agent would be beneficial and then but in scenarios where you have like different teams or developers managing different agents independently. Like for example, you have a big team with multiple developers like you know the HR team or the marketing team, sales team, everyone is trying to build their own agents. Then in that scenario it might be better to kind of like separate it out so that way you are not you know holding up others. So you can actually kind of like. Work on these agents parallely and then integrate it together. So in that scenario maybe the connected agents would be beneficial and then if you want to logically group instructions. So for example you have like a sales agent where it specifically look at the CRM system or different. Sales platform to perform certain things, so you can actually create those child agents. Then suppose if you do not want a sales agent to be deployed separately in that scenario, you can use the child, but if you want to. Expose those connected agents separately. Like for example if I want like an HR agent and I want to just deploy an HR agent but also include that logic for the main organization agent. So in that scenario it might make sense to spread it out so that we you. You can actually use that independently or from another agent as well. And if you don’t want to separate the configuration settings like if the child agent should follow your parent agent authentication deployments and everything so you can go with child agent. But if it needs to have its own dedicated authentication configurations and everything, then connected agent would be the preferred option. And when you’re using that child agent, like I mentioned, you don’t have to deploy those separately. So when you migrate or deploy to production, the main. So all the child agents associated to that will be deployed. But if you’re using connected then you’ll have to make sure that those connected agents are separated and then deployed appropriately and if you don’t want other agents to be used across different. That prompts and it should be purely used from the main. Then in that scenario you can go with the child, otherwise not connected would be like the preferred approach. So hopefully that makes sense to kind of when to use which agents and then we are going to take a quick look at an example in demo. So what I’ve done here is I’ve created one main agent, our concurrency agent and. Like associated several smaller agents so that way we can take a look at how some of those things work. So what I did here is so I’ve created three child agents because I don’t want these agents to be used individually and it needs to be part of the main agents, but in addition to that. I have also integrated four different agents which is like the connected agent. So what that means is I can use the IT help desk or HR agent from the concurrency agent or I can deploy and use these agents individually as well. So this will allow give me like two options like either use from the main one or like if you want to like separate it out and then deploy then you can use this approach as well. So the first agent is like a quick cell agent which is like the account use agent. So this is pretty much you know looking at the latest. Account related news. So for this demonstration I’ve used Microsoft as an example so you can ask like information or details regarding the news about Microsoft and things like that. The other one is the Employee assist agent, so I’ll be using this for the employers. Can use this agent to kind of like ask questions about like what is the travel and expense policy, what is the leave process like vacation balance, things like that, right. So that is the next agent. The next one is like a website agent where I have connected a public contingency website and then we can kind of like ask few. Basic questions regarding other website as well. The next four these are the connected agents. So this one is mainly for opening new tickets, accessing or asking questions regarding the ServiceNow knowledge base or like you know get existing ticket updates and things like that. And the next one is the HR agent which can be used for like you know onboarding. I get the onboarding status like create operators and training materials and all those different things. The task agent is for creating to do task and then finally sales agent is for you know the sales team who can. CRM or like the Dynamics 365 related information about their accounts, contacts or like you know the pipeline details. So we’ll be kind of like in a quick look at this now. So the goal is that we have used like a main primary agent and then a lot of different smaller agents. And so let’s take a quick look at that. OK, so I’m in my Copilot Studio portal right now and so since some of these features are in preview, so if by any scenario you are not able to see some of these agent functionality, you might want to add a preview after Copilot Studio so that way you can. See some of those. So that was something which I noticed few weeks back, but uh, hopefully should work for uh all of you guys. So you can see that here I have like 4-5 different agents, but in the previous slide I kind of like showed 7 and the reason for that is like some are child agents. So if it is a child agent, it will be purely associated to the main primary agent and any connected agent that’s what. You can see in your environment. So concurrency agent is the one which I created. So in here what I’ve done is I’ve provided like some basic description and instructions so. And you can kind of like give more additional information as well as needed based on your business case and scenario. So but for now I’ve just added some high level details and you can see the different knowledge sources and knowledge that we have connected to this main agent. So even if you. To add some of these knowledges to the Chai agent, it will pretty much show up in your main agent as well. So the way you can actually add some of these agents is so you should be able to see a new option called agents in here and if you go in here so you can see all the different agents that are here. Are connected to the main agent. So you can see that there are two types of relationship. One is like the child. That means all these are directly connected to the main agent and the connected points are individual agents and it can be used either independently or it can be connected and used from the main primary. And can see agent as well. So this way and in order to add an agent so you can pretty much click on add an agent and that will show you the different options. So like I mentioned M 365, SDK and Azure AI Foundry, those are coming soon. So if you go. Go to the preview mode that you can see that, but for now these are the three different options which are available. So create an agent is what you can use in order to connect or use an agent. But if you use Copilot Studio connect an existing agent, this is what. The connected agent will be displayed. This way you can actually connect different child or connected agents to your main Copiler Studio agent. Let’s take a quick look at some examples in here. What I’m going to do is I’m going to ask you questions regarding some of these different agents and let’s see how the copilot is working. So what I’m going to do is I’m going to ask an account related information about Microsoft. So what I’m going to do is I’m going to ask. I’m going to do us give me latest Microsoft news. So right now I have not mentioned you know go look at this particular agent or anything. So when I mentioned give me the latest Microsoft news, it pretty much reviewed all the different agents. Agents and connected agents and then it identified that I’m asking something regarding the knowledge source or information from the account news agent. So you can see that it looked at the accounts news agent and then provided me some of the latest information. So and if I want to ask like some additional things like is there any news? So you can see that again it pretty much looked at the account news agent and then looking at some of the documents and you know the latest news from the Power Platform area and then provided me those details so. You can see that now, but you don’t have to mention that you know I’ll go look at this agent, the powerful orchestration which is available in the main agent that will accordingly look at the different knowledge sources and data source and then provide the appropriate response. So the next thing which I’m going to do is I’m going to ask something which is regarding the employee assist agent. So I’m going to ask what is our? Travel and. OK, so now I was. So until now I was asking questions about different accounts, but when I mentioned what is a travel and expense policy, now you can see on the left hand side that it is looking at the employee assist agent and then giving me the appropriate details so you. And see how the knowledge is getting scanned and then you know looking at the different documents that are available or associated to the employee assist agent and giving me that response and you can see that it gave me like some citation and details from the. Documents as well. So you don’t have to mention that go check this particular agent and you know without that you can actually just kind of like actually converse with the agent and all the agents should be able to interact with the different child or connected agents and give you the appropriate response. So I’m going to ask one more question regarding this. So what is the? So in this scenario again, I’m asking another question from some of the documents and SharePoint sites that are all linked to the Scopilot Studio. So you can see that again it is pretty much looking at the employee assist agent and looking for the different information and everything and giving me all those citations and. Details as well. So if I go to policy so it based on my access so you can see that it pretty much take me to the internal document as well. So this way you can interact with the different agents. So I’m just going to switch and ask something specific to a different. Agent. So just going to mention who is? So now you can see that the agent got switched and it is looking at the website agent because I have added some public websites and everything regarding concurrency to all that particular agency. You can see that it is looking at our our concurrency website and then. Providing some of those. So again, you don’t have to mention like go to this agent, but other agent is able to kind of like you know accordingly orchestrate and then make sure that it is looking at the appropriate knowledge sources. So what I’m going to do is I’m going to so BB and I’m going to something like, you know, I want to see the latest blog post from our website. So we have like Brian who writes a lot of wonderful blogs. So I’m just going to mention show me. Let’s see how it is kind of like looking at. So again you can see that it is looking at our website and then giving me some of those. Looks like there’s not one so. I’m going to change. Let’s see if that returns something. Again, it is scanning some of our website content and then looking at the different things. So yeah, we were able to like find the different articles and you know our blog posts so you can see that it was able to look at the website agent. And then accordingly give the results. So at times we can definitely add additional instructions and everything to these child agents so that it will accordingly look at those details and give those results as well. So the next thing which I’m going to do is so I want to open a new. Our tickets. So I’m going to mention a ticket. So now we’re kind of like stretching gears and then trying to open a new ticket. So you can see that when I mentioned I want to open a new ticket, the main agent was able to orchestrate and make sure that it is calling the IT help desk agent. So what I’m going to do is I’m going to give a description. So now what it is trying to do is it is trying to create a ticket. So now you can see that a new ticket has been opened and it provided me the ticket number as well. So you know from the same agent we are trying to interact with the different smaller child or. And then without much additional details. So you can see that I was able to open it simple ticket from the copilot agent. So the next thing which I want to do is I’m going to ask something from an HR side of things, so I’m going to say something like I want to. Start onboarding because I want to onboard a new employee so you can see that the human resource agent got triggered. It’s asking for the employee name, so I’m just going to mention something like Tom. Williams. So now we have like automations in the back end that triggers and then all creates those employees. So you know we can see that our onboarding has been successfully started and then alert has been sent for manager approval and here is the employee ID. So if I come to the system and if I just refresh, so now you can see that a new record got created just now from the agent. So you can build all these powerful agents and another thing which I’m going to do is I want like the onboarding status. What I’m going to do is. I’m going to. So again, you can see that the human resource agent got triggered. So what I’m going to do is I’m just going to provide the employee ID for Dom Williams. So I’m just going to mention. So now we have like other automation to kind of like look at the current status and then give me that. So we can see that the application submitted is returned. So if I go here you can see that the status is application submitted. So you know, this way you can build all these additional agents and automations around the human source. And in addition to this, if you want to like send an offer letter, automate that or send tailing materials to new employees, you can actually do that using these Copilot Studios and everything as well. The next thing which I want to quickly show is a scenario where suppose if I want to create a task for myself. So what I’m going to do is I’m just going to mention create to do task. So we have like Microsoft to do which we can use for creating task for yourself. So I’m going to like show a quick scenario where how quickly you can actually automate some of those operations as well. So in this scenario it’s pretty much asking for like the task name. So what I’m going to do is I’m going to mention. Follow up with. So I just want to like create a task for myself and then make sure that I don’t forget that. So you can pretty much give the different like the priorities. So I’m going to go with medium priority for now. So once you do that so you can see that it new task was successfully created. So if I go to Microsoft to do so you can see that it task was created just now for under the medium priority as well so. This way you can actually automate the different tools, applications as well. The final thing which I’m going to show is regarding some sales related agents. So what I want to like to is as a sales person or someone on the sales team, I want to like look at certain. Account or pipeline related information, so I’m going to mention show the. So on the left-hand side, you can see that now it’s triggering the sales agent and looking at all the different knowledge which is available for those details. So again, you can see that everything was able to be arbitrated appropriately and you can see that these are the three different accounts. So yeah, this way you can actually look for that details. So one other thing which I’m going to do is so I want to like see my different opportunity related or like you know pipeline details. So what I’m going to do is I’m going to mention. So I’m just asking for like the different opportunities, open opportunities. So what I want to do is I want certain information, so give me topic. Launch. So for the scenario, I just want this information. So I’m explicitly mentioning some of these information. So when I mentioned that, so you can see that it is again looking at the sales agent and then looking for the different information regarding the different opportunities and. My plans that I have and hopefully we should get those details shortly. OK. So yeah, you can see that there are like 2 different opportunities and you know I I have like opportunity one and opportunity two as the name I believe. So that’s the reason why I can see that and you can see the account contact estimated closely and revenue information as well. So yeah, that was a quick demo. So you can see that you don’t have to like mention anything about which agent you want to like utilize and you can just simply converse using the copilot agent and based on the different child or connected agents that are integrated. Your main agent should be able to orchestrate and make sure that the knowledge is appropriately reviewed and give you the appropriate results. So this is definitely a really nice and great feature that Microsoft introduced recently. So just wanted to quickly hit on a few other topics where you know if even if you are using the low code, no code side of things, it doesn’t mean that you cannot go with the pro code. So you do have like the VS code extension for the Copilot studio agent. So you know if you are traveling if you want to. Work offline, you can actually pull from online and then kind of like make all the changes and push it back and everything should be all live in cloud as well. So you know you do have the ability to do that seamless to async as well. So this is again a great feature that allows the procode developers to utilize these agents and then make the necessary updates as needed. And if you want to like connect your agents to the Microsoft 365 Agent SDK, so you can do that as well. So all these copilot studio agents you can actually connect or integrate with your M365 copilot and it will you know kind of like interact and then accordingly. Provide the appropriate response as needed. So yeah, you can connect to your Microsoft 365 H and SDK as well. So a quick overview around some of the knowledge sources that are available. So here’s just some high level featured and advanced data sources that are available so you can connect to your public websites, SharePoint, Azure AI, so it’s like Dynamics, Dataverse, Salesforce service. So some of these are like some of the feature data sources or knowledge sources that you can attach with your agent. But in addition to that you do have all these different data sources like Confluence, Snowflake, Oracle databases and all these as well. But since Copilot Studio works. Pretty. Closely with our platform, there are over 1000 plus connectors available, so if you want to select or get information from some of those data sources, you do have the ability to do that as well. And if something is not available there, you can always do the HTTP call or question connector. That let you bring your data to the knowledge sources for your agent. So yeah, you do have the ability to connect with all those different knowledge sources for your agent so that the agents can review your business related information and accordingly assist your users. And in order to deploy now these copilot agents, so you can deploy that to Microsoft Teams and Copilot M365 Copilot in few clicks along with SharePoint as well. So you don’t have to go with the extensive process of like you know deploying, so you can deploy some of these for. Organization by deployment you might have to get a little bit help from your teams administrator so that way they can add those co-pilots appropriately. But within a few clicks you can actually build these agents and then add it to your personal teams and then you know kind of like validate some of those as well. And in addition to that, if you want to embed this agent in a web app or a native app or like Facebook and all these different channels, you do have the ability to do that as well. And if you want to integrate with the customer engagement hub where you want your agents to, I mean if someone. Escalates, then you can actually integrate with these different engagement hubs like D365 customer service, ServiceNow, Genesis, all these different platforms, so you can actually do that as well. Another quick thing which Microsoft recently introduced and mentioned is the model support in Power Platform. So you can actually build some of the task specific models within Copilot Studio. So for example if you have a way of summarizing or creating a state network and things like that. Actually build a task specific model and then utilize that in. So some of these features are still in preview and not 100% available today, but these are really neat features where you can actually build some of these things and then utilize that. In your different copilot studios and everything so that way some of these models can be reviewed while the agent provides response and create documents and different things. So again, a really great feature that is coming soon. So just wanted to quickly touch on some of the security and the governance side of things. So Copilot Studio follows all the different security and governance control, most of the governance control from Microsoft. So you do have the geographic data residency like data loss. So your administrators can actually set up which connectors or data source your employees or users can kind of like utilize while building these agents. So you have the ability to do that. You can standardize like multiple standard certification can be added if there are. All the regulatory compliance will be followed and if you want your makers, if you have a lot of different devs or users trying to build these compiler studios or local agents or apps, then you do have the ability to set the environment routing so that users are. Automatically routed to the the development environment rather than trying to build everything in like org byte environment. So you do have the ability to do that and then also like you know our regional customization and all those functionalities as well. So regarding the data loss prevention policies, so if you want like the makers to not use specific connectors, so like for example you do have the ability to classify your different data sources as. Business, nonbusiness or blocked. So what that means is if you have a business critical data and you don’t want that to be integrated or connected with an HTTP call or things like that, right? So you do have the ability to set up all those data loss prevention policies and. And make sure that some of the data sources, business critical data sources are not exposed. So you do have the ability to do all those things in the data loss prevention policies. So you can set up most of these things in the Power Platform admin center. So as a global. Admin or Dynamics 365 administrator. So you do have the ability to set up some of these things to make sure that before organizations are utilizing this, you can set the proper guardrails and governance for this as well. And this is again something that was released sometime back. So in the agents page of the Copilot studio. So now you do have the ability to see the prediction status. So what it does is it will look at the different things like the authentication. Policies and content motivation and all those things and make sure that it gives you like an update. You know if your agent with your bill is like in predicted status. So basically there are like 3 different options and like you know so it shows as like the predicted or. Or it can be like if there are any reviews or things that needed so it will pretty much show up as needs review and then you can see some of those details like you know these many messages were blocked in the last seven days out of all these many sessions so. This also something which Microsoft introduced sometime back as well. So you know at the individual copilot level you can see if you need to make sure that you review certain you know if you need to like review any premise security side of things as well. And then again, you do have the ability to audit Copilot Studio activities in Microsoft Purview. So you can kind of like review and filter based on the log and the event type and category so that you can see the different, you know, activities that are happening in Copilot Studio as well. So some of these things will definitely allow you to make sure that you are making sure to set up appropriate guardrails and governance before you actually roll out some of these agents organization by and for departments. And it’s just a quick overview around the use cases. So Copilot Studio, I mean since it is a low code graphical UI, the time to market is relatively quicker and based on which domain or business area. You are like be it governmental, retail, financial, manufacturing professional or any area. You do have like a lot of different use cases and opportunities to automate. So some of these agents can be built in few weeks and it’s. Pretty easy and really quick to kind of like deploy and everything. So yeah, I mean you have like a lot of these different use case options using which depending on which domain or area you are, you can actually build some of these agents relatively quickly. And make sure that it is live for, you know, your organization and your department. So great next step. So we can definitely help you guys get started with some of these AI agents and how to utilize some of these AI’s. So we do have like Copilot Studio and Multiagent Discovery session and all these are. Where our solution architects and team will kind of like help you assess and get started with some of these different agent and AI activities and if you want help with AI investment session for your executive team to understand what are the different opportunities or areas where. I could start so we can definitely help you with that. And then Microsoft have also introduced a new funding review so that will it’s available for the next year. So it will help cover the cost of your AI activities and actually start identifying what are the different. Areas or opportunities for your organization and where you could actually use these agents or you know the AI. So yeah, and I think Paige have added a link to this. There’ll be links to if you could, you know, please add that and give us comment, you know, we’re more than happy to reach out to you and help you guys spread all your AI tune. So yeah. That was a quick overview around the multi-agent system in Copilot Studio. So thank you so much for joining today’s webinar. If you have any questions, yeah, let me know. More than happy to stick around for a bit and answer those questions. Paige Wamser 49:03 Ajay, there were a few questions posted in the Q&A section. If you wanted to take a look at those, we might have covered one of them, the security and governance on that one of your last slides. Ajay Ravi 49:16 Mhm. Yep, so so since Copilot Studio is mainly built on top of like Power Platform, so you can if you’re using Power Platform if you have. So we do have all those governance and the security principles of Power Platform embedded in our Copilot. Studio. As well, so you can set up proper guardrails to make sure that the DLP policies and all those things are set up appropriately so that before actually the user start utilizing, you’re making sure that your business data or critical information is set up. Appropriately and in addition to that, our Copilot, the Power Platform are most of the compliance like HIPAA and all the different policies and the regional data and everything. So all those things are respected in Power Platform and Copilot. So you know most of those are governance and everything can be set up. And then in addition to that, you can actually install the Center of Excellence, which is like a governance platform or an option within Power Platform. So you can install and make sure that you are continuously monitoring and tracking all the different apps, automations, agents that you’re building all within Copilot Studio, so. All those governance and security principles of Power Platform, so you can actually use in Copilot Studio and make sure that everything is compliant and you know you are making sure that you’re predicting everything as well. Hope that answers. So for some of those internal applications, so if you are using M365 so it will pretty much look at those authentication. So like for example if you’re using SharePoint teams and some of those different. Approaches or like you know knowledge sources and if you want to like review details from there so you can actually use the M 365 license and make sure that your single authentication and everything is kind of like accordingly reviewed and made for your. Accessing all those Internet or internal application. So if you do not have access to certain applications then the copilot agent will make sure that you cannot hit that. So there are different ways to do that, but if you are deploying your agent to Microsoft Teams or like in a SharePoint. In our certain places regarding the internal applications, so it will respect your access and security before giving you the. Hope that answers. Um. I do think so. Uh, Pais will be sharing, uh, the presentation in our website, right? Paige Wamser 52:31 Yes, we will send a link to the recording after this workshop and then also if you would like a copy of the slides, we can provide those as well. Just let us know. Ajay Ravi 52:33 OK. Perfect. And then one other question was when a question is relevant to multiple and how does it choose which agent to draw from? So some of your instructions and everything would come into play there. So you want to make sure that you are providing the appropriate. Along with the instruction to all those child and connected agents. So that way when you’re asking a question specific to maybe like multiple agents, so it will kind of like give you all the instruction that you’re provided and then accordingly give you the best answer. So it is really important that you give really informative instructions and instruct information so that the agents can identify from where to look and give you the appropriate response. OK. So I think that was pretty much it. So any other question? I’m more than happy to seek for a better. Paige Wamser 54:10 OK. I think that’s it. Thank you so much. Ajay Ravi 54:12 Sounds good. Thanks, Tim. Have a good rest of your day.
Events View Recording: Pro-Code Meets Low-Code: Navigating the Full Stack Seamlessly Build Faster with Fusion Development—Pro‑Code + Low‑Code, One Seamless Stack The future of application development is fusion—where low-code agility meets pro-code power. In this webinar, we’ll explore how your teams can use Microsoft Power Platform and Copilot Studio to accelerate full-stack solution development. Learn strategies for blending professional development with low-code agility to enable faster… July 9, 2025
Events View Recording: Inside the MCP Protocol: Supercharging AI Agent Collaboration MCP Protocol Explained: Smarter AI Agent Integration Enterprise AI systems face a critical challenge: connecting large language models to live data and external tools without building fragile, custom integrations for every service. The Model Context Protocol (MCP) solves this by providing a standardized framework for AI agents to access databases, APIs, and business tools securely.… July 17, 2025
Events View Recording: Govern with Confidence: AI Safety, Red Teaming & Compliance with Microsoft Purview Govern AI with Confidence—Microsoft Purview, Red Teaming & Safety in Action As generative AI moves from experimentation to enterprise-wide adoption, ensuring safety, trust, and compliance is no longer optional—it’s essential. Join us for a practical look at how organizations are implementing AI governance through adversarial testing (AI red teaming) and the powerful capabilities of Microsoft… July 30, 2025