How Copilot Agents Work in Microsoft 365 and Teams (Full Transcript)

[00:00:00] Speaker 1: Agents expand the capabilities of AI tools, allowing you, your partners and clients to work with customized AI assistants for specific jobs or purposes. Agents can accomplish tasks within preset specifications. They can be set up to work independently or can create workflows across multiple applications. Hi, I'm Nick. Welcome to my full-length training course on using agents in Microsoft 365 Copilot. This is a fairly long video, but it is organized into 11 sections. You can use the YouTube chapter markers to jump directly to each one. In this first section, we'll clarify what agents are and give you an idea of what you can expect in the rest of this course. Now, there are many different AI systems and chat assistants including Gemini, Clawed, Copilot and many others. And most of those systems have their own tools for working with agents. This course will focus on tools from Microsoft's Copilot brand. Naturally, this course will assume that you have some familiarity with Copilot in general. And of course, anybody can use Copilot for free at copilot.microsoft.com. However, Copilot agents are not available in the free system and they're pretty limited in subscriptions for individual users. The Copilot agents and agent building tools that we'll use in this course do require a Microsoft 365 subscription from a company or similar organization, also known as Microsoft 365 business and enterprise accounts. Now, there are three main goals in this course. First, I wanna help you understand what an agent is and how agents can help you focus or streamline your work in powerful ways. Second, we're going to see how to work with existing agents that have already been set up and are available for you to use in Microsoft Copilot and Microsoft Teams. Some agents are built by Microsoft, some are built by third-party developers to support their apps and services, but there are many agents that you can work with today. And third, we're going to give you the tools you need to build your own agents. I believe that anybody who can use an agent in Copilot can also build an agent in Copilot. You just need to be willing to do a little more work and planning. However, it's very important to recognize that there are different types of agents. Building agents has a lot in common with developing applications. No single training course can teach you all the ways to make agents. There are tools and systems offered by Microsoft and others that require knowledge of application coding and programming APIs. But in this course, when we build agents, we will be using Copilot Studio and the Copilot Agent Builder, two options offered by Microsoft, which integrate into known Copilot systems, allowing you to build powerful agents fairly quickly without the need for coding or application development skills. With that in mind, the agents we build here will not be able to work independently and will not integrate with third-party systems. Those are more advanced features that are a little outside of this course. So with the stage set, the rest of this course will be much more hands-on. If you're taking your first steps into the world of Copilot Agents, I wanna start by helping you understand why agents are so valuable, how they can save you tons of time and get you much more focused results from Copilot. As we do this, we're going to actually go hands-on and use agents that Microsoft offers inside of Copilot and in Microsoft Teams at no additional cost. So let's get to work in Copilot. Since I'm using a Microsoft 365 business account, I'll go to the Microsoft 365 website and sign in. And I'll make sure to click New Chat on the left to work with Copilot. Now, let's say I'm considering a career change and I want to work in the forestry service. I could ask Copilot a simple question about that. I'll tell Copilot to help me find and apply for jobs in the forestry service. After a moment, Copilot shows me websites and gives me some examples of jobs that are open now. But it's not really helping me navigate a career change. Copilot can be much more helpful, but I have to write a better prompt. You could apply prompt engineering techniques or you can just be more detailed and clear about what you want. With that in mind, I could tell Copilot, you are my career advisor with over 20 years of experience. Help me identify the challenges and opportunities in making a career shift. Ask me questions that I have not considered to help me identify what my goals should be. After giving Copilot a specific role, instructions and context, then I'll finish the prompt with the same request. Help me find and apply for jobs in the forestry service. And now I'm not just getting a quick answer. Copilot is engaging with me, helping me understand the challenges of a career change and asking questions to help me identify what I need. This may be the start of a much more powerful conversation with Copilot, but consider the work it took to get here. This upfront work is one of the main things that an agent can solve for. So in the Copilot chat interface, there's a category for agents on the left. So I wanna click all agents and these are pre-made agents available for you to use in Copilot. We'll start with the category of agents built by Microsoft and I'll click the button that says, see more. These are agents that are already designed around specialized parameters or specific jobs. For my career change, I might want to try this agent called Career Coach. This agent is designed to guide you through your professional journey, explore career opportunities and give actionable advice. That sounds like what we need. There's a button to add that agent, but you should be aware, not all of these agents can be used by everybody. If you do not have the full Microsoft 365 Copilot subscription, then you may see this add button on some agents, but others may have a button that prompts you to upgrade your account. So I'll add this agent and now the Career Coach agent is listed on the left under the agent section. I can select new chat up at the top when I wanna work with the standard Copilot chat, or I can select the Career Coach when I wanna work with that agent. And we can assume that the agent has all the context it needs to help with a career question. So I can just ask the main question, help me find and apply for jobs in the forestry service. The agent gets to work as my career coach and it starts by asking me some leading questions to help identify my goals. Once I answer those, it's going to do some very specialized research for me. I can expect this to provide great information, but also engage me in a conversation to help me identify what I need. Now, once you've added an agent in Copilot, there are a few other ways to work with them. If you click new chat on the left to use the main Copilot chat, in the chat field, you can type the add symbol followed by the name of an agent. You can only choose an agent that you've already added. And in fact, let's say you don't choose an agent here. Maybe you just start in the normal Copilot chat and ask a question. And if you don't get the information you want, you can invoke an agent and a follow-up question in the same chat. So without starting a new chat, I'll type the add symbol, start typing in career coach. It finds it there and I can just click on it. Then I can ask my follow-up question. And now the agent is working with me in this chat. And if you ever want to remove an agent that you've added in Copilot, you can just find it in the sidebar on the left and point at it. A button with three dots will appear. That will open a menu and you've got the option there to uninstall it. And if you work with Microsoft Teams, you can use agents there as well. In Microsoft Teams, you can click the apps button in the sidebar on the left or click the button with three dots and then choose get more apps. These apps are add-ons that can give you more functionality inside of Teams. And there's a category here for agents. I'll choose one here. Let's try the idea coach. I'll click add on this agent. And once it's added, I'll click open with Copilot. And that takes me to the Copilot chat here inside of Teams. I can see that Copilot is selected on the left and the idea coach agent is selected. Now, some agents will work differently in Teams. Some will create a dedicated chat in the chat section and some will appear as an additional button in the sidebar on the left. So just keep an eye out for different ways to interact with agents in Teams. Now, one last thing, I'll go back to the Copilot website and I'll select the agent that we set up. And when you add a new agent that you've never worked with before, I think it can be very helpful to ask the agent what its parameters are. Here's a prompt that I recommend. Describe your parameters, including capabilities and functions, your rules and constraints and your technical configuration. And the agent tells me about the upfront parameters that it uses. We can also see the functions it uses. One important thing to see here is that this agent can use enterprise search. So it can access my data in my email, messages, shared documents and more from my Microsoft 365 organization. We often refer to this as the work mode in Copilot. And it's a feature that is only available for users with the full Microsoft 365 Copilot license. This is something that we covered in my other course on Microsoft 365 Copilot for organizations. Now you do not have to ask an agent about its parameters, but this can really help you understand whether an agent will do what you need. So we've seen what the career coach agent can do, but that's just one agent that works to help with a specific type of goal. And I also wanna stress that this is just the beginning of what agents can do. But you can start to see the value of a Copilot assistant that is pre-programmed with specific roles, instructions and context. And if you stick with this course, one of the things you'll learn is how to build your own custom agent like this, but with specific parameters that you define. And once that agent is built, you can share it with your teammates, which can save a lot of time and effort in your organization. I think it's a good idea to get familiar with two very powerful agents that Microsoft has built in and spotlighted in the main Copilot chat. We're talking about the researcher agent and the analyst agent. These agents are only available in the full Microsoft 365 Copilot license, at least at the time of this recording. In the main Copilot app or website, we'll go to the agent section on the left in the sidebar, and I'll start by selecting the researcher agent. Now, of course, most of the questions I ask Copilot can be considered to be research. However, most of the time, I'm also looking for a quick answer. The researcher agent does not deal in quick answers. It is designed to take extra time to compare resources, clarify requirements and context, and take precautions to avoid inaccuracies. Now, when I worked with a new agent, I like to ask what its parameters are. So I'll ask it to describe your parameters, including your capabilities and functions, your rules and constraints, and your technical configuration. And in the response, we see that this agent does deep research, comparing many sources. And it can also use information from my organization. But there are some trade-offs here. In rules and constraints, we can see that the researcher cannot generate pictures or images, all very important things to know as I ask my starting question. And to start, I want to refer to my employee manual. You could drag a file to the prompt field, or if you have the full Microsoft 365 Copilot subscription, you can type the forward slash, then find a file from your OneDrive or SharePoint. I'll find my employee manual, then I'll tell it to consider standard policies that should be included in a corporate employee manual. Tell me if there are any subjects that may be outdated, or any subjects that may be missing. The researcher will almost always ask one round of follow-up questions to clarify your needs. So I'll answer these questions. Once I answer those questions, the researcher will take a very long time to research and come back with a response. When you get the response, all of the normal rules of Copilot apply. You can ask follow-up questions. You can and should look at citations to check the sources of where it found this information. But when you want detailed information where accuracy is vitally important, the researcher agent is pre-configured for those needs. Next, let's look at the analyst agent. I'll select that from the sidebar on the left, and I'll ask my standard question about the agent's parameters. Basically, the analyst is built around getting insights from actual data. To give the analyst a try, I'll refer to my company's product inventory spreadsheet. Then I'll write my prompt. Several of our products are backordered. Based on the prices and values of the items in the inventory and information about market trends, what insights can you give me to determine whether we should restart production on the backordered items? The analyst agent will also take a long time to generate responses. And let's be clear, you certainly do not need to use the analyst agent for a question like this. However, the analyst agent is already designed with parameters that make it very effective for questions like this, saving you the time and effort of writing those parameters into a normal Copilot prompt. The results include some actionable strategies for helping me decide when we should restart production. It also identifies some missing data. If I can provide that, it may be able to provide more options. And it finishes by offering to make a chart or simulate scenarios. The researcher and the analyst are two very powerful agents, and they're already set up for you as long as you have the full Microsoft 365 Copilot license. Now, I'm not sure about this, but my guess is that Microsoft identified some of the more powerful tasks that people expect from Copilot. And instead of making the main Copilot chat tool more powerful and potentially slower, they decided to offer these pre-configured agents to help with those types of tasks. And they're also very useful to help us understand how valuable agents can be in general. Within the Copilot app or website, there are several pre-built agents that you can use to expand Copilot's capabilities. Some of those pre-built agents are made by Microsoft, but many others are provided by third-party app developers. These give you the ability to interact with other applications within a Copilot chat interface, leveraging Copilot's AI features to supercharge those apps and services. From the main Copilot website, look for the agents section on the left and click all agents. I'll scroll down to the section labeled more agents, and I'll click see more to see the full collection of third-party agents. There are many different apps and services that can be integrated into Copilot using these agents. But remember, they won't all work exactly the same, but we can choose one to get an idea of how they work in general. I use Canva a lot, so I'll choose that one. You may want to choose one that you work with regularly, but keep an eye out for details that may work differently. And if you're not familiar with it, Canva is an app for designing graphics, for social media, marketing promotions, YouTube thumbnails, all sorts of things. So choose an agent, then click the add button to install it. And now I see the Canva agent is listed on the left and it's already selected. And I always like to ask each new agent what they can do. So I'll ask it to describe your parameters, including your capabilities and functions, your rules and constraints, and your technical configuration. I'll send that prompt. And we can see here that this agent can help me create new designs in Canva, retrieve existing designs from my Canva account, or give me suggestions about design styles. Great, so let's give this a try. Now, one of the best things about using an agent in Microsoft 365 Copilot is that I can reference other files and if you have the full Microsoft 365 Copilot subscription, you can refer directly to files in your organization or information from your meetings, messages, and emails. So I'm going to tell this agent I want it to refer to information from my email. So I can type the forward slash key, then start typing in a word that I think is in the subject of that email. I'm looking for an email thread about our Volunteer Day event. It found it here, so I can select it. So that email thread is being referenced in this Copilot prompt. I'll tell it, based on the information from this email, design a promotional graphic for Instagram to advertise the event. Now, the key to third-party agents is that they will need to connect with a separate app or service. So I will need to allow that connection to continue. But you should also consider whether you trust that app enough to connect it to your Microsoft 365 account. You may want to get some guidance from your IT department or security team about third-party integrations. In this case, I'll just click Always Allow. And with most of these agents, you'll need to sign into your account with that service at some point. But after some time, it shows me some design concepts from Canva. Remember, in Copilot, you can always make follow-up requests. In the Canva agent, I found that it doesn't always like to make direct changes, but I could ask it to generate new designs that include the date. I'll send that follow-up request and give it a moment. And here are those new designs. Now, I say that the agent doesn't like to make changes, but once you have a design that you like, you can click the link to preview and edit it. And it's at this point where I do need to sign into my Canva account to continue working on this design. And after I'm signed in, it opens that design and I can work with it. You can make any changes you want or download the final graphic. Now, of course, Copilot itself can generate graphics, but the Canva agent in Copilot can create complex designs with multiple layers, including stock imagery, adjustable color palettes, and text that you can actually edit. Going back to Copilot, let's consider privacy and security. At some point, you may want to remove the permission for Copilot to access this agent or app. Click the button with three dots near the top right and go to settings. Then go to the subcategory for agents. And I can see here that I have connected Copilot to the Canva app. I can click on that and there's an option to reset, which will remove the permission for Copilot to connect to Canva. If I try to use the agent again, it will ask me to grant that permission again, which I can always decline. And if you want to completely remove the agent from the Copilot interface, reset to the main Copilot chat, then find the agent and the sidebar on the left, point at it and click the button with three dots to open the menu, and there's the option to uninstall it. Now, clearly I chose an easy app to integrate with. I encourage you to look through the other third-party agents available in Copilot. It's incredible how we can basically turn Copilot into an AI-powered front end for each of these apps, along with the ability to reference the emails, files, messages, and meetings in your Microsoft 365 account. The agent mode in Excel is a specialized agent that only works in Excel. It's designed to make independent decisions based on your requests, and it significantly expands what Copilot is able to do in Excel. It's different from other agents we've seen because you have to go to Excel to use it, and it mostly presents as an Excel feature. You could open Copilot on the web by going to the Office website, click the app button on the left, and choose Excel from there, or you can open the Excel desktop application on your computer. When you open Excel on your computer, I recommend you click the Account button on the left and check to make sure you are signed into Excel with the right account. The Microsoft 365 account you use for Copilot is the account you should be signed into here. Then go back to the Home section. So I'm going to open a file that I have stored on OneDrive. In Excel, you can go to the Home ribbon, click the Copilot button, and open the Copilot panel. And here you can make a general Copilot request, or you can ask a question about this spreadsheet. I'll ask Copilot, which are the top five items with the highest profit margin? It finds the answer to that question based on data it found in the sheet. So let's try something else. I'll start a new chat, and I'll tell it to sort by retail price. But we'll find that the Copilot chat does not make changes directly to a spreadsheet. It may answer the question in the chat panel, or it may offer a new Excel file with the changes that you can download, but it will not edit the original spreadsheet directly. Instead, this is where we would use the agent mode. The agent mode is able to make changes to your spreadsheet, but that's just the beginning, as we'll see. So I'll start a new chat, then click the Tools button in the chat field and choose the agent mode. And now I'll tell it to sort by retail price again. And after a moment, it does make that change. This is similar to a feature that used to be in Excel called App Skills, but it's no longer there. If you have a request like this, you should be using the agent mode. Of course, you can always undo the change or ask Copilot for something else. I'll ask it to filter to only show accessories. In the category column in my spreadsheet, there are several different types of products. Some of them are listed as computer accessories. Copilot can identify those items and filter the spreadsheet down to only show those. I'll reset by telling Copilot to clear the filter. And next, let's ask the agent mode to automatically label certain data. I'll tell it for each item with a profit value higher than $100, color the row in green. The profit value is not in the sheet, but Copilot can calculate it using the wholesale price and the retail price, which are in the sheet. Then it applied conditional formatting to label those rows. Conditional formatting can apply a color or other formatting element to specific cells based on specific conditions. But I didn't ask for conditional formatting. I asked Copilot to do a task using natural language and Copilot decided how to do it. It used conditional formatting to highlight the items with the highest profit value in green. Next, let's add that profit information so we can actually see it. I'll tell Copilot to calculate the profit margin for each item and add that to a new column. And after a moment, it adds a new column with that calculation. So the agent mode can handle requests that require changes to your workbooks. But like I said earlier, that's only the beginning. You can ask the agent mode to analyze your workbooks or add specific content. It can perform research to find what it needs. It cannot make up information or draft new original content like Copilot and Microsoft Word, but it can project trends and it can do calculations and it can generate dashboards, tables, and graphs based on data it calculates or finds through research. So for example, I'll ask it to make a dashboard that will help me understand my data and focus on giving me insights to help me understand how much profit we lose by not selling items in our inventory that are backordered. And I'll even ask it to give me advice to help me decide when we should restart production on those backordered items. The agent mode will need to do some research and do some analysis, which takes a long time, so you should give it several minutes. But when it finishes, I can find that dashboard in a new tab in my workbook. There are several tables that contextualize how much money we're missing out on from our backordered items. And it also includes recommendations and advice to help me decide when to restart production, as I requested. And all the values in the dashboard are based on calculations from the original spreadsheet. So if I update that spreadsheet later, the connected values in the dashboard will update as well. Now, this is just one example based on my simple inventory spreadsheet, but the Excel agent mode's ability to do independent research, evaluate data, compare that data to information from the internet, and its ability to modify my workbook and generate dashboards makes it an incredibly sophisticated tool for data analysis. This is much more than Copilot responding to questions in a chat field. This is independent action based on my instructions. And that's what agent mode can do in Excel. Once you see what agents in Copilot can do, you might start thinking about making your own agent. There are lots of different types of agents, and some of them do require specialized skills, but some types of agents are surprisingly easy to build. The first tool you should get familiar with is Microsoft's Copilot Agent Builder, which lets you build your own agents and share those agents with your teammates. The Agent Builder is available right inside of the Copilot interface for all users with a Microsoft 365 business or enterprise account. Just be aware, some features are limited to people with the full Microsoft 365 Copilot subscription. Starting from the Microsoft 365 Copilot website, we'll go to the Agent section in the sidebar and click New Agent. There are two tabs at the top, Describe and Configure. We'll start with the Configure tab, but after you know how this works, you might prefer to start in the Describe tab. You'll understand as we go. Now, before we do anything, if you have not already identified what you want your agent to do, you should take some time and really think about your goals. I want to make an agent that can answer HR questions in my company. It can save my HR department some time by helping team members find quick answers about company policies. Now, you could start from a template here. These are mostly based on the agents that Microsoft provides in Copilot. So if you want something similar to the Career Coach agent, you might choose that as a starting point, then set your own custom parameters. But I'm going to make my own agent without a template. So I'll start by giving it a name. I'll call it HR Helper, and I'll put in a brief description. Scrolling down, the Instructions field is where you do most of the work. Think about prompt engineering techniques, or just consider all the detail you might put into a Copilot prompt to get a robust answer. You might give Copilot a role, describe a certain set of rules or context to consider before answering questions. You can put all of that into the Instructions here. So I might add lines like, you are an HR director with 30 years experience managing HR departments in large companies. I'll start a new line, then add, refer to well-known established HR practices when answering questions. Or here's one of my favorites that I like to include with most of my agents. I'll tell it, do not guess or estimate. Check multiple sources to ensure accuracy. If you do not know an answer for sure, tell the user, and encourage them to contact the HR department. And if you're not sure of all the instructions you should include, that's where the Describe tab can be very helpful. This is an AI assistant within the Agent Builder to help you generate your instructions. So I could tell it that I want an agent that can answer questions about HR policies in my company. I'll type that in, then send the prompt. And if it asks any follow-up questions, make sure you answer those. But now it says my agent is set up. So let's go back to the Configure tab and see what's changed. I can see that it has added several more instructions, and it even rewrote some of the instructions that I provided. I strongly encourage you to read through these and make sure that everything looks right. And these instructions might give you ideas for more that you want to add. But we're not finished yet. Still in the Configure tab, scrolling down, we'll get to the section for Knowledge Sources. And this may be incredibly valuable. I want this agent to answer questions about HR policy at my company. My policies are well-documented in my employee manual. So I'm going to tell the agent to use the employee manual as a knowledge source. There's an option to add a link to a website as a knowledge source. But if you also have the full Microsoft 365 Copilot subscription, then you will have the ability to upload documents or use documents that are stored or were shared with you on OneDrive or SharePoint or Teams. To use a document from my OneDrive, I can click on this field, and I can start typing in the name of the file. It searches through my OneDrive, finds a file that I want so I can select it. And after a moment, it's added at the bottom. But here's an important variable. Will you be sharing this agent with your teammates? If so, the files that you attach must be accessible to those teammates. So if I choose a file from my OneDrive, but I have not shared that file from my OneDrive with my team, then the agent will not be able to access it when my teammates use the agent. So in that case, I'm going to click the X to remove that file and confirm. And alternatively, if you upload a file from your computer, it will be bundled with the agent and any user will be able to access any information from that file. So there's an important trade-off between privacy, permissions, and usability. I have a copy of my employee manual and my company's social media manual on my computer. So I'll upload those. And now those files have been uploaded and will be included with this agent. Before moving on, you should set whether you want the agent to find information on the public internet or if you only want it to use the sources that you've provided. I'm going to give it permission to search the web and move on. Next, under capabilities, you can decide whether the agent can generate documents, code snippets, or images. And then there are suggested prompts. These will appear below the chat field when people use the agent. The suggested prompts that are already filled in here are based on the description I put in when I used the describe mode. You can edit or delete the suggested prompts here. So those are the core steps to building a basic agent. There is a test version on the right, so you can try some test questions and see how it responds. And then you can make changes in the configure tab. But in my experience, I think it's safer to finish the creation process rather than depending on this preview. So when you're ready, just click the create button at the top. And don't worry, you can still make changes later. When it's done, I'll click go to agent. And that agent is listed in the sidebar on the left, just like any pre-built agents that you might add. You can go to the new chat option at the top when you wanna use the normal co-pilot chat, and you can select your agent when you want to use that. And I cannot stress this enough. You are not finished yet. Testing and refining are essential to making a good agent. You should try some questions in your agent and see how it responds. So in the chat field, I'll type, I have a client who wants to fly me to New York in exchange for publishing some social media posts about our partnership. Is that okay? I'll send that prompt. And it gives me a response based on the agent's instructions, as well as information it found from the employee manual and the social media manual. You should test several prompts. And if it does not respond exactly as you want, you should edit the agent. To do that, you can click all agents in the sidebar on the left. Find your agent and point at it. The button with three dots will appear. You can click that to open the menu and choose edit. Then make changes to the instructions or the knowledge sources, update the agent and test it again. Take the time needed to test and revise until you're confident that the agent is ready. And if you want to share the agent with your teammates, there's a share button on the edit page, or you can go back to all agents, find the agent and open the menu and choose share there. You can only share the agent with other users in your Microsoft 365 organization. If you only want to share the agent with specific people, you can choose that option. Then go to the field and start typing in somebody's name. It searches your company's directory. You can choose that person and you can add multiple people. Or you could just make it available to anyone in the organization. So choose the option you want, then click apply. And then you can copy this link and send it to your teammates. When the teammate clicks the link, they will see an option to add the agent to co-pilot in their account. However, you should be aware, if one person builds an agent and they use features that are only available to people with the full Microsoft 365 co-pilot license, then the agent will only be usable to other people who have the full license. People with the baseline Microsoft 365 account would see a note telling them that they would need to upgrade to use the agent. So anybody in a Microsoft 365 organization can make an agent, but there are some features and sharing options that are limited to people with the full Microsoft 365 co-pilot subscription. Making an agent to share with your teammates allows one person to do a little work that can save several other people lots of time. But making an agent just for yourself can also be incredibly valuable. If you write long, robust prompts to define specific parameters in your co-pilot requests, you might think about whether you can save those parameters in an agent. So then whenever you have a related request, you can go right to that agent and ask a quick question without establishing the context every time you write a prompt. If you get to the point where the agent builder is not able to do what you need, you may want to consider CoPilot Studio, which offers sophisticated tools for building more complex agents. However, working with CoPilot Studio does require the participation of your organization's Microsoft administrator who will need to set up licenses, pay for usage fees, and help with the deployment of the final agents. So there are some details that you will need to work out with your administrator, and we cannot cover all of those details here in this course. But in this video, I will help you understand how the licensing works in general. This will help you decide if you want to jump into CoPilot Studio, which we will do a little later in the video. AI tasks can be expensive. So in many cases, your organization has to pay for your agents in the form of CoPilot usage credits. And you will have to consider two important things, who will build the agent, and who will be using the agent. So first, if you're the one building the agent, the question is what type of Microsoft account or subscription do you have? The baseline Microsoft 365 business or enterprise account does not let you build with CoPilot Studio. There is a free 30-day trial available, which can be extended by an additional 30 days. But after that, these users cannot use CoPilot Studio. If you have the full Microsoft 365 CoPilot subscription, then you can use CoPilot Studio to build agents. But we also have to think about who will use the agents you build. If you have the full Microsoft 365 CoPilot subscription, and you are using the agents you build yourself, then you do not have to pay for CoPilot credits. If you share your agents with teammates in your same organization, who also have the full CoPilot subscription, they also do not need CoPilot credits. So when do you need credits? Well, if you build agents and share them with teammates who have the baseline Microsoft 365 account only, then you will have to pay for CoPilot credits to allow those people to use your agents. Or if you build an agent that takes independent action based on program triggers, then those actions must be paid for with CoPilot credits. And finally, if you build an agent that can be used by people outside of your organization, those are also subject to the CoPilot credits. So how do you buy these CoPilot credits? Well, your organization can purchase credits with a prepaid plan or with a pay-as-you-go plan. Of course, you're wondering how many credits do I need and how much do they cost? Well, for that, I will refer you to this pricing page with a link to estimate your CoPilot Studio usage. I know there are a lot of variables to consider here, which require a bit more research based on your needs. But if you're just getting started with agents in CoPilot Studio, then these steps are something you'll come to naturally further down the road. At this point, if you are ready to start building an agent with CoPilot Studio, let's jump in. I'm going to show you how to make an agent in the form of an AI chat that can respond to user requests within specific defined parameters. Just remember, there are some features like independent action and connections with other apps that are offered by CoPilot Studio, but they require additional skills that do not fit into the scope of this course. I'll show where those tools are, but we will not be able to cover them completely. And of course, if you plan to publish an agent for other users, you will have to work with your company's Microsoft administrator to do that. But we'll see more about publishing as we go. Before you do anything, I think that planning and at least identifying your goals is very important. I'm going to stick with the scenario we used earlier in this course. I want to make an agent that can answer questions and filter requests on behalf of my HR department. This will help my HR team save time because they won't have to answer all of the first level questions that come in. To get started, you can work with CoPilot Studio by going to copilotstudio.microsoft.com. I'm signed in with my Microsoft 365 account from my organization. There are a few ways to start building an agent. On the homepage, there is an AI assistant. You could choose the agent tab, then describe the agent you want. The assistant will build out the features to start your agent. Alternatively, you can switch to the agents section on the sidebar on the left, and you can click the button to make a new blank agent and then fill in every detail manually. Or you can choose a pre-made template to start with. But we also have the AI assistant here. I'm going to use that to get started. So I'll just describe the agent that I want. I'll tell it to make an agent that can answer HR questions in my company or elevate questions to the correct HR representative. Then I'll click next. This takes some time, but it starts building the agent. It gives the agent a name and a description. I could click the edit button and change those, but we'll stick with what it said. Scrolling down, we have several configurations that you should set. And many of these configurations will look familiar if you've used the Co-Pilot Agent Builder, which we saw earlier in the course. But here's our first option that is not offered by the Co-Pilot Agent Builder. You can choose which AI model powers your agent. And there's also some information about the strengths of each model to help you choose. Next, we see triggers, which can enable an agent to act independently when certain events occur. As mentioned before, we're only going to focus on the most essential features. But if you are interested in setting up autonomous actions, this is where you can find those controls. Scrolling down, we see our agent's instructions. In plain language, you can tell the agent exactly what you expect it to do. You can use well-known prompt engineering techniques to give the agent a specific role, expectations, and context. You can tell the agent what its limits are and how you want it to respond. You can tell it if there are any special rules. Now, in this case, the AI Setup Assistant already filled in several instructions for me. And you should read these because they will help you learn the types of instructions that you can give. But also, if there are any that you want to remove or change, you can click the Edit button to make that change. Then we get to knowledge sources. Just like in the Co-Pilot Agent Builder, you can tell your agent where to look to find information it needs. Since I want this agent to answer HR-related questions in my company, I want it to refer to my company's employee manual, where it will find the answers to the most basic questions. Now, this is great. The AI Assistant has already searched through my shared documents and identified suggested sources that fit my goals, including my company's employee manual. I could click Add on one of these suggestions, and it tells me where this file is located. But be careful here. If you use a file stored on OneDrive or SharePoint, and you later share the agent with teammates, but you have not shared the source file with those teammates, then the agent will not be able to use the knowledge from that source file. These privacy protections might be exactly what you want, but they can also affect the usability of the agent, depending on who you share it with. Now, in my case, I'm going to cancel this, and I'm going to dismiss these suggestions. Then I'll click the Add Knowledge button to add sources manually. And from here, I think a specific website can make a great knowledge source. You could also use several other data storage systems listed here. But for now, I'm going to upload the documents I want from my computer, and that will bundle the knowledge sources in with my agent, and they will not be subject to privacy or sharing policies. I'm doing this because it's easier, but you absolutely must consider the sensitivity and privacy of the files that you might include this way. Anybody who uses this agent will essentially have access to the potentially sensitive information in these documents. So I'll choose Upload, and select the files from my document folder. So I'm uploading a copy of my employee manual and my social media manual. Then I'll click Add to Agent. Before we finish with the knowledge sources, you can set whether the agent can find information from the general public internet. If you disable this, it will limit the agent to only using information from the knowledge sources you provided. Scrolling down, we see the options for tools, agents, and topics. We will cover topics in a separate video, and we're not going to use the others. And finally, we have suggested prompts. Most AI chat assistants offer a few suggested prompts to the users, which helps the users understand the types of questions the AI chat is designed to handle. And they can offer users a shortcut to asking useful questions. I'll click the Add Suggested Prompts button, and I'll just add one here. So in the title field, I'll label this one Remote Work Guidelines. Then I'll write the prompt. Describe the company's policies for working remotely. You can add up to six, and then click Save. We won't really see those suggested prompts until the agent is published, but you can always come back here and edit them. Now, before we move on, I want to scroll back to my knowledge sources. My documents still show the status of In Progress. My agent will not be ready until these all say Ready. So I'll skip forward in time. Now that my knowledge sources all say Ready, I'm ready to continue. Now, testing your agent is vital, and that's why there is a preview version of the agent on the right. When you think you've done your setup that you want, you should try some test questions. Ask the agent questions or ask for help with tasks that it should be able to handle. So in the prompt field, I'll say, I have a client who wants me to fly to New York in exchange for publishing some social media posts about our partnership. Is that okay? While the agent is processing, we see some of the background options on the left. This might help you identify any problems that may come up. And here's the answer to the question, including citations confirming that it is using my knowledge sources to find this information. You should try several test prompts, and if there's anything that does not work as expected, you should go back and edit the instructions or the knowledge sources or whatever you need. You should take the time for several rounds of testing and revisions before the agent will be finished. Now, clearly this agent has not been deployed for my teammates or partners to use. We're going to see how to publish an agent in a separate video, but we have seen the core tasks required to build an agent. You can go back to the Overview tab at the top to reset to your editing tools. And you will also see that you can expand many of the building tools using those tabs at the top. So I could go to the Knowledge tab to focus only on the knowledge sources and go back to Overview to reset to the main set of tools. There's also a button at the top for Settings. You can see there is a lot of depth here that you can explore as you continue to learn. In the Generative AI section, you can set rules for how responses should be formatted. You can set up content moderation or decide whether your agent can process uploaded files. In the Voice section, you can set whether the agent is optimized for voice inputs. You can go to the Languages section to establish which languages the agent can work with. So feel free to explore the settings and click the X in the top right to close Settings when you're finished. And that will take you back to the Agent Building Console. Now on the left, I'll reset to the homepage. And from the homepage, you can go to the Agents section to see the list of all the agents that you've built. And you can click on one to continue working on it. Generally speaking, an agent you build in Copilot Studio uses AI to understand what a user is saying or asking. Then it formulates a response. However, if there are scenarios when you want your agent to respond in a specific way to establish questions or prompts, you can set that up using a system called Topics. We've already built an agent with all the required features. Now we're just going to use Topics to make sure it responds exactly how we want in certain scenarios. From my list of agents in Copilot Studio, I'll select the agent I built earlier. In the Overview tab, you can see the main tools that you use to build or modify the agent. There's a section here for Topics, but this isn't showing everything. You can click See All, or you can click the tab up at the top for Topics. And here you will find some topics that are already listed and sorted into categories for Custom Topics and System Topics. The biggest difference being that you cannot delete System Topics. You can modify them, but you cannot delete them. I'll go back over to Custom Topics, and there's a menu next to each one so you can delete them if you need to. But let's take a closer look at one. I'll click on the topic labeled Thank You. Now, every topic is invoked by a trigger. The Agent Chooses trigger is very common. This means when the agent recognizes certain input from the user, it triggers the topic. We see the instructions to the AI explaining what it should look for. So if a user includes thanks, thank you, or any of these other phrases in their prompt, this topic will be triggered. Or if the AI recognizes something that it understands to be similar to one of those phrases, that will trigger the topic as well. Then below that, we see what happens next. In this case, the agent will respond with a message replying exactly with, you're welcome. Each of these blocks in this tree is a node containing a step in the conversation process. This is a very simple topic, but a topic can get much more complex. I'll click the back button near the top left to go back to the list of topics. You may want to go over to the system category and select a more complex topic. Let's select this fallback topic. Topics can have different types of triggers. There can be conditional nodes that split the tree into different steps, depending on certain conditions. And looking at how existing topics are structured like this may help you learn how to work with the topics feature. But it will also help you if you have some experience with programming language or writing system configuration files with variables and conditions. You can also click the button with three dots at the top and switch to the code editor. And if you have experience with YML markup, that would definitely help you here as well. But for now, I'll close the code editor and click the button near the top left to go back to my list of topics. So if there is a specific scenario and you want your agent to respond in a specific way, you might want to make your own topic. You could open the add menu up at the top and choose to make a new topic from blank. But considering how complex the node system is, it's unlikely that a beginner would be able to jump in and figure it out. So instead, I suggest you use the Copilot Assistant to help you build your topics. So you can choose add from description with Copilot, then fill in the name you want for the topic. In my case, my agent handles questions related to HR issues. But if there's anything related to reporting a conflict in the workplace, that should go directly to our director of HR. So I'm going to name this topic conflict report. Then in the description field, I'll tell Copilot if the user asks about or even suggests a scenario where they may need to report a conflict, refer them directly to the HR director. This will be our starting point and we'll make changes as we go. For now, I'll click create. And it generated the conversation tree for that topic. In the trigger, the agent will choose to trigger the topic when the conditions below are met. Basically when the user mentioned something about reporting a conflict. You can modify this description here to dial it in. If that event is triggered, the agent will reply with a specific pre-written response. I would prefer to rewrite this response to assure the user that we will protect their privacy, then giving them a more direct email address or phone number of the HR director. But for now, I'm just going to leave it as is. In most situations, especially if you are new to this, starting with a topic generated by the Copilot assistant is going to be far more practical than making a new topic manually. You can and certainly should go through every step of the topic and make changes or rewrite the responses. Or you can click the Copilot button up at the top and you can ask for changes there. But for now I'll close this. And when you're finished, make sure you click the save button at the top, then go back to the list of topics. Now that I have that new topic set, I'll go to the agent test panel on the right and I'll tell it, one of my coworkers is stealing my supplies who can I talk to about this? That triggers the topic and it responds with that pre-written response. Now, obviously this was a very basic topic tree. They can get much more complex, but even the Copilot assistant can help you build the more complex topics. And these are first steps. There's lots of room to grow and more that you can learn about topics. For now, I'll go back to the overview tab. Topics are just one tool that can help your agent understand user input and respond. Topics, knowledge sources and general instructions are the tools that you can combine to make your agent respond accurately and efficiently for your users. If you've built an agent in Copilot Studio, chances are you're going to want to publish it so your teammates or partners can use it. In this video, we're going to see the steps that you can take to publish your agent from Copilot Studio. However, there will be additional steps that your organization's Microsoft administrator will need to handle. So we will not finish this video with a fully published agent, but I believe you will have a clear picture of the steps that you can take and what you'll need to discuss with your administrator to finish the process. From copilotstudio.microsoft.com, I'll select the agent I've been working on in this course. And first, you need to consider authentication or who will be able to access your agent. Will they need to sign in with an account? Click the settings button at the top, then go to the security section, then to authentication. The easiest option is to set your agent to work with Microsoft authentication, letting people use it if they are signed into their Microsoft account. If you plan to release your agent on the web or in a third-party app, and you want it to be accessible by the general public, you can set it so it does not require authentication. But it is vitally important that you ensure that all of the knowledge sources and information that may be accessible through the agent is safe and clear to be used by the general public. Do not disable authentication if your agent deals in private information from your company. There's also an option to configure an authentication service provider manually. If you have the setup for that, the option is there, but I'll leave this set to Microsoft authentication. And if you make any changes, make sure you click save. Then click the X in the top right to close settings. You can always make more changes later, but if you've built and tested your agent, set the authentication rules, and it's ready to go, you can click the publish button at the top, then click publish here to confirm. Now, publishing an agent does not actually make it available to others. It just makes it ready to push out in the distribution channels that you choose. Take a look at the tabs at the top. You may see a tab labeled channels, or you may need to open the menu to show any additional hidden tabs, and then open channels from there. Channels are the different places where you can make your agent available. In this video, we're going to make the agent available in Teams and Microsoft 365 Copilot. In another video, we'll explore the options for making an agent available in other services. So I'll select that channel, and every channel will have completely different configuration settings. This is just the first step. So first I need to confirm that I want it to be available in Microsoft 365 Copilot, then I'll click add channel. When that's done, then this agent is ready for me to use myself. I could click see agent in Microsoft 365 to use it on the Copilot web app, or I can click see agent in Teams to use it inside of Teams. That opens Teams where I can add the agent. Once it's added, I can open it and work with it here. And by the way, I see a suggested prompt here. There's only one suggested prompt because I only set one when I created the agent. Now this is all fine if you just want to use the agent yourself, but you may also want to fully deploy the agent so that your teammates can use it in Copilot. So let's go back to Copilot Studio. And if you have closed this channel information panel, you can always get back to it by going to the channels tab and click on the channel for Teams and Microsoft 365 Copilot. Since we've already added the channel, the next step would be to click availability options. And here we see the option to show everyone in my org. This is where you go to deploy the agent to the rest of your organization. However, you can only start this process. You can click the button and submit the agent to your administrator for approval. I'll confirm that that's what I want. And after a moment, I see that it has been submitted and is waiting for the administrator. After that, your administrator has tools in the Microsoft 365 admin center to approve or reject your agent. If they approve it, they can make it available for other people in your organization. Your teammates will be able to add it from the agent store in Microsoft 365 or Teams. Just remember, as we saw in another video, there are limitations on who can use these agents in your organization based on your Copilot Studio license and any usage credits your organization has purchased. Aside from publishing your agent for teammates in your organization, you may want to make an agent available to clients or partners who are not part of your Microsoft 365 organization. As we saw before, you do have to consider the question of authentication before publishing an agent. And the authentication settings you need will almost certainly be different if you plan to publish the agent externally. Click the settings button near the top right, go to the security section, then choose authentication. Now, the reason it's not letting me change the authentication settings here is because I've already published this agent for my teammates using this authentication strategy. So making users sign in with their Microsoft account was perfect in that scenario, but that will not work if I'm sharing the agent outside of my organization. If I want to change the authentication, I'll need to reset my previous publishing channels. So I'll click the X near the top right to close settings. And at the top, you may see the channels tab, or you may need to open this menu to show additional hidden tabs, then open channels from there. We'll be exploring these channels more in a moment, but because I set up the Teams and Microsoft 365 Copilot channel in the past, I'm going to click on that channel. Then I'm going to remove the channel to reset. So I'll click remove channel and confirm. It says that that channel was deleted, so I can close this panel. Now I'll go back to settings, back to security and to authentication. And since I have reset my channel settings, I am now able to change this. To deploy an agent on the web or in a third party service, you'll need to set it to no authentication or set up an authentication service manually. If you want to set up an external authentication service, you should communicate with your Microsoft administrator or your IT department. This can take significantly more work and research, but depending on your security needs, it may be necessary. Once you have those details in place, you can connect that service to your agent here. But for now, we're going to see the much easier option. We're going to set this to require no authentication. However, if you do this, then all of the knowledge sources and information that may be accessible through your agent will be accessible to the general public. You should never disable authentication if your agent deals in private information from your company. But we are going to do that for this agent. If you make any changes, make sure you click the save button. This gives a quick warning about publishing without authentication, including list of agent features that you cannot use. Once again, I'll click save. And with this authentication set, I'll click the X to close settings. And back in my agent, the next step is to click the publish button. This gives me another warning about authentication, but I'll click the publish button here to confirm. Publishing an agent does not actually make it available to others. It just makes it ready to push out in the distribution channels you choose. Once again, at the top, you may see a tab labeled channels, or you may need to open the menu to show additional tabs, or in my case, the channels tab is already selected. We see some information about my agent status, but I'm going to scroll down so we can look through these channels. The channels are the different places where you can make your agent available. In the other channel section, we see several third-party services. So you can make your agent available inside of Slack or WhatsApp or one of the others. But there is a lot of setup required for each of these third-party services. Most of these will require you to have an established account with that service, and each one will have their own requirements for setup and deployment. Also, almost all of these channels require you to have a Copilot Studio license where you have purchased a plan for Copilot credits. When external users work with your agent, that incurs additional costs that is paid for by the Copilot access credits. Of course, different channels require different configurations, but this is where you can get started with each of them. Since we can't cover all of them, let's see how the system works in general by setting up a demo website. This does not require me to have an account with a third-party service. So I'll choose Demo Website, and here you have the option to set a welcome message or some conversation starters. I'll change the welcome message to Welcome to the HR Policy Agent. If you make any changes here, make sure to click Save. And in this panel, there is a link to the website, which you could copy and share with somebody, or you can just click the button to open the website. And this is a great way to get a finished agent available to the public quickly, and it's an excellent system to test your agent. I'll ask a question here. I'll type, I have a client who wants to fly me to New York in exchange for publishing some social media posts about our partnership. Is that okay? I'll send that question and see how my HR Policy Agent responds. And of course, it gives me some great information using the knowledge sources that I provided when I built the agent. And for now, I'll close this website and go back to Copilot Studio, and I'll close this panel. And let's acknowledge the demo website is by far the easiest way to deploy an agent to the public, but it's also the most limited, and it's not really practical for long-term use. I showed it so we can see how to set authentication and how the core publishing channels work. Of course, any of the third-party publishing channels will have their own special setup, but for now, you know where to get started with those publishing channels. And that brings us to the end of this full training course on working with Copilot Agents. Thank you for coming here and learning with me. I hope you've got everything you need to use or even build agents in Copilot. You can always use the chapter markers to jump back and review any of the sections from this video. For more videos like this and to continue learning, be sure to subscribe to this channel.