View Recording: Real Results with Microsoft Copilot: What’s New and What Works

Tabatha Frozena 0:08 Hello, good morning. Corey Milliman 0:10 All right. Good morning, everybody. Going to go ahead and get started. I’m Corey Millerman, a technical architect with concurrency, and I’m joined by. I’ll let you introduce yourself tab. Tabatha Frozena 0:24 Yeah. Hi everyone. I’m Tabitha Frozena, director of sales at concurrency power, user of copilot. So I wanted to compliment Corey’s done a great job on our past webinars. But compliment share a little bit about if I can use it and it’s helping, anyone can use it? So excited to share some new advancements and hopefully some great value and tips and tricks for you. Corey Milliman 0:47 Awesome. So what we’re going to go over today, we’re going to go over our new capabilities with copilot. So we have a new interface that’s been rolling out. We have a new mobile app experience. There’s an application enhancements. Reasoning agents have been released, and we’re also going to go through some deep thinking capabilities and then we will also be going through everything. Live where we need to. So first of all, we do have a streamlined interface that is rolling out. Not everybody is going to see this right away. I don’t see this interface yet, so I have a screenshot, but the first time you open copilot chat going forward, you’re going to get an experience that introduces some new capabilities. The chat module, which was recently or previously rather the copilot chat, or the copilot module, you can. See, that’s been renamed, and that’s going to give you access to your history, copilot pages and agents. Now everything is moving over the left side. So you can collapse that, move that around. But right now, when I go into copilot, everything I have is on the right side. But as this new interface rolls out, you’re going to start seeing that change and everything moving over the left and some new verbiage there. On the mobile side, we are seeing that copilot pages are going to be available in the mobile app as well. So this feature started rolling out in May, depending on if you’re an iPhone or Android, you may or may not see this. But again, you’re going to see a simplified experience. OneDrive is being renamed to search in the mobile app. Copilot again is being renamed chat. And then there’s a new module called create. So these three modules are going to be pinned to the bottom of the copilot app, and that’s going to give you the user the ability to access all of your AI chats, apps, and files directly from within Nicole App. So those apps, those updates started rolling out. In May. Tabatha Frozena 2:45 So if you haven’t downloaded your copilot app on your phone, please make sure to do that ’cause. That’ll continue to enhance your experience, not just only on your laptop during work, but also on your phone. Corey Milliman 2:56 Absolutely. So in OneDrive now and I was just trying this to see if I could share this with you before our demo, but I had a little trouble with the audio, but now with your OneDrive files, if you click the OneDrive icon in compile it, you now have AUD. Overviews so you can actually click through one of your files. Couple of your files and it’s a really good way to generate kind of an audio. Overview perspective. But it does a podcast style as well, similar to what Google was doing with Google Notebook and what I use it for is sometimes. I’ll take someone on meeting recordings and I found that I can take three or four meeting recordings and then I get this objective like commentary about this is what was discussed. These are the pain points. This is what’s going on. So it’s a really interesting way. To turn a lot of information almost into a podcast that can just kind of listen to. And it’s a little interesting. So you have two different options with that, one is going to be the podcast style where you have two people. Sounds like you know, typical podcast and the other is just an audio overview that it’ll create. So again, if you’re driving or something, you can get an overview of, say, meeting notes or meeting agendas or different things like that to help you prepare for a meeting or prepare for a project. And now there are also as it relates to teams in copilot. Now we have additional multilingual capabilities. So what is happening is you are able to as a meeting organizer, enable these multilingual meetings. Now when that’s enabled, each meeting participant can choose their own spoken language and translation language. So meeting organizers are going to see this new control option called Enable multilingual speech recognition. This will start the meeting transcript. It’ll start the intelligent meeting recap and it will accurately reflect each participant’s spoken language of choice. So it rolled out to public preview in April and now that is rolling out here from Mayon. So the intelligent meeting recap now supports the multilingual meetings. You can catch up in your own language. Key discussions, even when multiple languages were spoken. And the recap is automatically generated. Rated in the translation language that you selected. So if I’m meeting with multiple people and we have different languages that are spoken, I can see the entire meeting note meeting overview and all my copilot action items in my own native language. Tabatha Frozena 5:38 This is really big, especially for our companies that interact globally. I know that’s been a common use case for AI and just the the Gen. AI of how well it is doing translation. So excited to see this enhancement to just better improve meeting collaboration. Corey Milliman 5:56 So now we’re going to hit on a couple of new agents that Microsoft put out. We have the researcher agent which helps manage complex multi step reasoning tasks, so it’s going to combine open Ai’s research model with Microsoft 365 Copilot search capabilities. And some different things I’ve seen this used for as an impact of tariff on different business lines preparing for vendor negotiations or gathering some insights before a client call. And I’m actually going to stop the PowerPoint presentation and just switch over here. So I can show you my screen and I’ve actually already gone ahead and sometimes the researcher has anybody on the call used this before. The researcher agent was looking for a show of hands. Show of hands here while I’m bringing this up tab, if you could watch that and let me know what we see. Tabatha Frozena 6:54 Yep. And if you haven’t, and when you get this, I do have a step guide of where you can find it, how you can then pin it to your taskbar and and use this afterwards so I can share that out too. Corey Milliman 7:07 So for me, I have these agents here on the upper right hand side, right? So I already have agents with all of these new Microsoft agents. If you click get agents you can see the agents. I have and then we have all the agents that have been built by Microsoft. And so I’m gonna go ahead and start. If I start with my researcher agent, I can pin that up here and have that available all the time. So I’m actually going to bring up. A research agent that I recently used. Instead of going to the copilot, store here. And. OK. Open up the other window. I have too many things open here. Switch to work versus love. We see my work chats and webchats, so here’s my work chats. And. So this was a scenario I put together right? So I wanted to prepare for a copilot studio conversation. Low code, no code. Kind of scenario for a bottling company and so I gave it a prompt saying help me prepare for a meeting with new customer. We’ll be discussing the impact of copilot Studio’s low code solutions on how they can be leveraged in the modeling and packaging industries. So I. Gave a couple of follow up questions and I walked away. And it came back and it generated. It took about 12 minutes for this one. And it generated a lot of information about the benefits I was looking for. But really it generated more information because I don’t have a solid understanding of the bottling industry and food packing industry. So I was going out and finding use cases that related to that, so it actually pulled up information about. Overall equipment effectiveness and different things I don’t know about, but are going to be relevant and it actually referenced information from Coca-Cola. Recent deployment of copilot studio in a low code scenario. So this is a really good way. You can use it on the public Internet. You can also use it against all of your own documents and meeting notes. And tell it that you want to maybe prepare a project brief for a new member on the team or something like that to get them up to speed. So it is a really powerful way using reasoning models to take a step by step process to get to a solution for you. So tab, how do you use that one? Or Oh yeah, there you go. Tabatha Frozena 9:40 Sorry it’s it’s been. I was on mute. Sorry about that. Corey Milliman 9:43 We’re on mute. Tabatha Frozena 9:44 It’s been a new one for me with just the recent updates from Microsoft build and just how to do it, but it’s been really helpful for me from an A sales role to get a deeper understanding. Two of the customers that I’m going to be meeting with maybe. Some offerings or pain points that they have to just get that more comprehensive view than just some of the kind of Q&A that. Chat gives you. To to help better answer. So it does take a little while to get more information, but the information is meaty and there’s a lot to go through. Corey Milliman 10:17 So another one is the analyst agent and this one is a transition over more like copilot is thinking like a skilled data scientist. So it’s built on 03 Minis reasoning model. And it is optimized for advanced workplace data analysis. So again, it utilizes chain of thought, reasoning, breaking down complex problems, step by step, refining answers, and bringing a more human like analysis. So it does run Python code. To solve data queries. And I found a couple of different ways to use this, and one that I looked at here. If I open up more is I actually uploaded a few different log files that I had and I had some different SSH things that were going on. I had some log. I had. Someone had a Windows 2000 log file and HPC events and open SSH events. And I was trying to understand, you know from these log files that were really large. What is going on? What are the patterns, failures, security risks that that we see here? So went through and we can see the chain of thought that it used, the reasoning and the coded executed to get there. Starting the analysis and again this took some time. We went through a couple of iterations but after it completed this analysis, let’s keep scrolling here. It went through and it gave me an analysis. To these log files. So it showed me how many failed login attempts they were and where they came from, what users were targeted. I see git. I see root, you know all these different items here for failed SSH login attempts. I see some issues with the Windows log file and then it also gave me information about the network and system failures. So it gave me a list of brute force security risks, system failures and usage patterns. So I said OK. Well, I would like a visual report focused on these risks and patterns and failures. So then it broke this down for me and it showed me the IP addresses that were in this simulation attacking something. So these log files were not real customer log files. These were actually generated by AI for this and I wanted it to simulate what I would see in a log file if there was some kind of attack. So here we have an SSH brute force attack. It’s showing how many requests there were. And it is showing the targeted usernames route. Of course being the biggest target showing the error related items in my HPC logs and just give me a graphical overview of everything through this and then I asked it for graphic report of the top ten issues that were ident. Here are the top 10 issues. Essentially failed. Gives me an kind of a picture of what that looks like. I asked it to provide an analysis based on. Geographical data and it was unable to do that because it doesn’t have an ability to tie into an API for that. So that is something we can do with, say, copilot studio. But when it goes through here to give me a summary. Of all my findings recommended mitigation steps I’d ask for and then I went as far as to say. All right, give me the PowerShell scripts necessary to mitigate these issues. So it gave me my hardening script, gave me a Windows Update and repair script. It gave me some information on how to check a node, so with this kind of scenario I can upload log files a lot of log files are pointed to a directory with these log files and start to identify trends and patterns in data that would typically be really. Difficult to do so. That’s one way you can use that. And again, that’s the analyst and by default it’s going to give you some. Prompts that you can use, so create a table. What are some quick insights? What? What trends do you see? And then it’ll start servicing information. Your data that maybe you don’t see and it will help you move through that. So I’m just gonna pause there and see if there are any questions and dropping over to Chad quick. All right. Tabatha Frozena 14:29 Nothing in there. Corey Milliman 14:30 Any questions? All right. Just jumping back over and wanted to make. Tabatha Frozena 14:34 Yeah. One thing I know a lot of organizations have very complex Crms and other platforms that are doing certain things. But how? I’ve been using the analysis trying a little bit Corey on even just jumping out Excel and saying tell me which deals have been sitting in the same stage for over 30 days and then asking it to give some insights and reports or compare certain reps. Win ratios. So of course I could do all those things looking at it, seeing. But giving it to analyst agent and getting different insights and then prompting it to go even further than in chat of saying based on these results, can you help me with the coaching plan or how I should performance manage? So just other ways to create efficiency in using? The data, but seeing it in a different way. Corey Milliman 15:23 Absolutely. So there are also some other ways to access these analysts. These different agents that are out there by hitting the ampersand. I can start having a chat with these different agents and showing me all the agents that I’ve actually have access to in my environment that I brought. In and these are my copilot agents. So again, we have researcher, we have analysts, we have a visual creator that’s going to help create graphics, and these are all out-of-the-box from Microsoft, and they’re releasing more and more. So again, to use these, once they’re enabled by your organization, click get agents and you will see all the agents that are published by Microsoft. And when you start creating agents, it’s going to give you the ability to use some of these as a framework to create your. Own agents for the organization. So there are a couple of ways to start creating agents in copilot. It’s really easy and I’m actually going to share it in a use case that I actually had in a meeting right before this. As we go through this, taking about 3 minutes to do that and share it with the team so we can create agents and copilot chat and it’s just eight steps, we go to copilot. We define the purpose, add some data, define the behavior. You can do some testing. And some interface, testing, publishing, sharing and then use so that is the same whether you build that agent and copilot chat. Or you go ahead and you can actually do this directly from a SharePoint site. As well. So it’s really easy to get started with these and anybody with a copilot license is able to create these agents. Microsoft has done a really good job of enhancing the visibility of these agents. You’re going to start seeing more information available on the copilot dashboard, so you can actually see how your users are deploying agents, how those are being used across the organization. So you can decide if those need to be something. Maybe it needs to be brought in and cop. Studio. So we’re bringing it into a more robust framework where. If I have different people in the organization that are relying on something quite a bit I want might want to formalize that transfer ownership of that and bring some enhancements that allow us to do some different things. But we do have that visibility to determine who can create agents. And one thing to point out is if users create agents and those are shared across the organization by creating that agent, it is not creating a new security. Bypass. The source content that was used. So if I create an agent on a document on my OneDrive and I share it with tab and tab decides it says hey, this is really cool and she forwards it to somebody else in the organization. If they don’t have access to. That storage document. They won’t be able to use it, so by sharing these agents and sharing these links, we are not exposing the data that was used to create those agents to people that don’t have access to that. So we’ve covered a couple of there are some additional ones as well. Resume review, contract review research assistant these are some examples of agents I’ve built out with copilot chat. Resume reviewer to compare resumes against job descriptions. We can actually look at contract and legal review, and I’ve included these scenarios at the end of the deck. That kind of gives a step by step on how you would do this so that you can go and implement some of these in your own organization. But what? We’re going to cover from a live demo perspective. Today is copilot in Excel, is something that is really important, a lot more organizations are using that. So that allows us with no programming experience, to look at copilot and Python And Excel. We’ll create an FAQ for HR. So in HR chatbot that’s grounded with different policy documents across the organization, we already jumped ahead and covered the log file analysis. I do have an agent. But I also written to do the log file analysis, but I found the deep research or the analyst option actually works a little bit better. And then there’s deep research. So before we jump into the live demos though, I just wanted to pause here because while we do have different examples of use cases across different silos in the organization, different departments and I kind of want to pause here before we go on some live demos. So we have our oh, go ahead. Tabatha Frozena 19:50 Yeah. Oh, sorry. Yeah, well, this is really important because on the call you could be like what? Maybe if some of you are already building all these agents and doing this, but for those that haven’t gotten started and Corey, maybe I think it might be the next slide, I’ll just think about your role, the function, things that your teams are answering for and think how. I use copilot to help me find that right. If you’re constantly going to a team site looking up a document every day. Hey, can I create an agent that’s going to already be pointed there with the right prompt to help me be more efficient? So using this guide of saying step back think what role, what function, what individual thing am I doing? And can I take a second to say how can I use a copilot? Which should be your assistant. It’s not a full pilot, right? It’s not going to be perfect in doing your role. But how can it help be an assistant to you to get what you’re looking for done more efficiently? I am going to drop a link. Another great asset that I use and live in every day is Microsoft’s scenario library. So it goes into a little bit more details. Then that great summary. You can go back to that slide, Corey. Thank you. Of just certain functional areas, it can give step by step guides of how to use the tool sets that you might already have licensing for. But yeah, I’m excited. Corey, for you to go to your demo, but wanted to hit on a couple of these of just customer service. I think the SharePoint agents of connecting to a knowledge base or going to a service. There might be great ways of. Like you had already showed from the log files like draft me, a response of common inquiries using language from past tickets, right, and just giving an idea for either a manager or an individual contributor to, say, use the information I have put it into copilot and tell me. Are there are there certain things I’m missing or pulling out? Maybe even to prep for certain renewal calls or how a customer service manager can walk through. Better servicing their customers. From Ahr one, I think the most common use cases is help me draft a job description, right? That’s been an easy one to use. And also I think a lot of us are recording meetings or using copilot and teams transcripts, but taking it that next level of using copilot to say help me summarize key actions help me build a plan. Help me take this into my next one-on-one with my. Team to really maximize the meeting recording and the notes to not be manual. And then, yeah. Corey Milliman 22:23 No, go ahead. Sorry. Tabatha Frozena 22:24 Lastly, I was just gonna say some of the sales ones we’ve been using copilot internally. Obviously prospecting sometimes in true sales roles is the last thing people want to do or spend time on from cold calling standpoint. But putting in these are the personas. These are the offerings. These are the companies I want to target. Can you build me an 8 touch plan with phone and e-mail scripts to have a consistent language? Clearly there still needs to be human review because. Those on the call don’t want sales emails. That are just you can tell their AI generated, so you still want them to be authentic, but it helps take that step of building a custom e-mail, building a plan, putting a phone script together. So it’s really accelerated my team’s ability to put prospecting plans together and. Then, as a sales leader review those to have consistent language to make sure that we’re having common stories and sharing what works. So it’s been a really great efficiency saver. By adding more to the prompt and really using it on things that. Our pain points today of what people don’t want to spend time on. Corey Milliman 23:28 Absolutely. All right, I’m going to jump in then to some more demos here. So I’m going to go ahead and minimize this one thing I mentioned right was at the beginning of right before this call. We were on an internal call and we were talking about all the differe. Acronyms that Microsoft has, right? So there are so many and somebody said, is there something out there, some kind of acronym dictionary we talked about? Tabatha Frozena 23:51 Let me see. Corey Milliman 23:57 Yes, Microsoft has some stuff internally, but what I did during that call was said wait. Let’s create an agent that does this, so we have the ability to centrally start building this out. Understand that. So I’m actually going to go here, go to create an agent when you click create an agent, you can start with a new one, or if you’ve already created agents, you can go back here and bring those up to modify their. Modify how you configured those, so I’m actually going to go ahead and modify this one. So I create this Microsoft. Acronym dictionary to give it some instructions. Provide definitions and explanations of Microsoft acronyms. Excuse me, Microsoft acronyms respond promptly. Maintain a professional tone. Avoid jargon or overly technical language. And so I pointed over to his site. I found that had a Microsoft acronyms glossary and then somebody else on the call quickly gave me a document and I said OK, I can go ahead and put this up here. So when we actually start this, I can just say. What is C4? And C POR is certified partner of record and I’m in the testing and when I do that, this is really funny. It says it’s not actually in the document that I uploaded. However it is in the website. So tell me what the this one is actually saying. Claiming partner record is a Microsoft program for partners, blah blah, blah blah blah. So this is a really good way that we now have a spot where I create a quick team folder. Now different people as we learn more or want to centralized more information. That gets uploaded now. Anybody in the organization that can use this? I’ve shared that with a group of about 15 so far. But now they can do something. There’s some things here that started out, but it does give us then that baseline of what new acronyms are out there. What is it doing? I turned on looking on the Internet as well. So people have different questions about acronyms. It’s super easy. So that was something that we’ve done before the meeting ended today. So really. Tabatha Frozena 26:05 Yeah, and think about that from an onboarding experience. Hey, I don’t want to have to sound silly and not knowing something right, but servicing. Sometimes the inks of having to understand every acronym. How awesome to create that to yeah help people self-serve, feel more educated and then continue to ramp up so. Corey Milliman 26:27 Yeah, real simple real quick. And that’s just an example, right? So if you start, if you have information in your organization where we want to turn that into something self-service, this is a really easy way to start taking that that knowledge that we have across the organization and make it so that people can actually start having convers. About that. So another example, let me see. Let me go back to my list. So I’m actually going to go over to Excel and Python And this one is a pretty. So what? We’re going to be doing is looking at how we actually. Get into data visualization with Python here. So let me just go ahead and start this demo here. Here we go. So another window. Oh my goodness. Tabatha Frozena 27:26 Live demos. Corey Milliman 27:30 It was just right here. I made sure it was all available before the call. I just have to click start now again, I’m sorry accept terms and conditions and sometimes. So sometimes I do have some sensitive data that shows up in my history or things like that. So this time I’m using something that is a little more sanitized and allows us to go through a document here without exposing any proprietary information. So Excel has charts and with copilot we can do that a lot differently. We can do that a lot better in this data set. We have 23,000 rows of vehicle registration data by state and by county. And what I’m going to do is analyze this data to determine the popularity of electric vehicles by state and county over the last few years. So that is what we are going to actually start doing here. So. To get started here, we’re actually going to go up and modify my glasses here so I can see. So I’m going to go ahead and click formulas. Now if I new Python, I would actually just click the icon. This is Python And start using Python directly here, but I’m not. I’ll just say I’m not, so I’m going to go back home and now I’m going to go over to copilot and we’re going to start using copilot to help us get information about this data. So from here I have my copilot pain plane. Excuse me? Pain. And then I’m going to actually select this divider and expand this out so we can see here when we’re down. And I’m going to say ask a question and here I’m going to say with Python, create A to create a strip plot of electric vehicle data. Go ahead. And click send. So copilot is suggesting that I use advanced analysis because I entered Python into the prompt. So advanced analysis is going to take my natural language requests and convert that into actual Python code. So I’m actually going to go here and click start advanced analysis. So this advanced analysis mode in new sheet is opened up. Copilot has gone ahead. It’s ensured Python code and it’s going to start adding adding visualizations in my data, so I’m not editing my original data. So we’re going to go ahead and we’re going to actually make this bigger by selecting page. And then. And advanced analysis is pulling all this information directly from the range of cells that I’ve selected in that workbook. So it’s giving me a quick visual. Right now the total number of electric vehicles registered across the United States. But I want to iterate down and get a little bit more detailed and get into my state. So I’m going to go in and I’m going to give it a prompt in my advanced analysis. So I’m going to tell it to change the X axis to show my state. And have the Y axis show the electrical vehicle total. So I just described that and now it’s going through and showing here by state. So it looks like Washington state, of course has the most registered electric vehicles. So now I want to get down into that data and understand which counties in Washington state have the most electric vehicles registered. So again, I just tell copilot exactly what I want to do, change the X axis to group by county in the state of Washington. Go ahead and click send. And now I have a list of all my different counties here. So my chart has been adjusted and I can see the King County has the most electric vehicles registered. So now let’s drill down a little bit more and let’s calculate the average number of electric vehicles per county and rank them highest, lowest. So we’re going to complete that average. And here we can see King County. And now that the data is ranked, we’re going to make it a little bit easier to read because I, I’m circling to read that. So now we’re actually adding some filtering. We’ve added some commas. We have a blue line that’s going to show the top ten counties. We have kind of this divider here. So now if we want to add some more color to this, we go ahead. Color each item by year. So this is going to make it easier to see how the average number of registered vehicles continues to grow in time. So each dot represents a month. Each color represents the year, and we’re showing data from 2017 to 2024. So now we might have known something, but now we have the we saw that Kings County was the top on this, but now we have the data to back this up. So we can actually add more information and we can add text box showing the number of items that are above the 20,000 line. And. It’s added that I really don’t like where that is. So I’m going to, so I did this so you don’t have to watch me type as well. So now let’s move that box. And now we have a better understanding of what’s going on there. So you whether or not you know Python, you still have the code that you can tweak. So if I click this, it’s actually showing me the code that was used and if I want to change if I know Python I can. I want to change the color of that line. Maybe go down right here and I’m going to change that and I’m going to. Bring this over to red. Get my green check mark. I’m editing that Python code here. It’s gone ahead and changed that line, so it’s just one example of this is a little bit different because we do have in that analyzer agent the ability to look at data. One thing that’s interesting about this is I can actually point it over to multiple files. I can start asking questions about multiple files and I’m actually working with the data that’s already you know something I’m already using so that is just one way that you can use Python in Excel. So I’m gonna pause there. And just bring my screen back up and see if we have any questions so far. Tabatha Frozena 33:22 No questions. Quiet group. Corey Milliman 33:25 Everybody’s quiet today, all right. Tabatha Frozena 33:27 But Corey, would you? Would you say I mean that? That’s something new in advancements, right? I think early on there was a lot of, like, oh, copilot and PowerPoint and excel, like, didn’t really meet the needs. And that continues to get better and better. With with just additional features that are rolling out in more investment to it. Corey Milliman 33:44 Absolutely right. When copilot for Excel first came out, it was pretty limited. Copilot for PowerPoint was too. And the advancements with kill pilot in Excel and even PowerPoint have been pretty significant. We also have deep thinking available. So anytime you’re asking certain questions, you can actually when you’re interacting with copilot, click the deep thinking button and it’s going to go deeper into your data, deeper into your organizational data. Take extra time extra processing. Power to give you more information about what what you asked. Tabatha Frozena 34:18 No. Which is great news because especially if you already have it part of your license had a better enable it, but also from the data security and information protection, right, there are definitely other tools that I’ve heard people use and how much better it is for some of these. Selections, but then you have to worry about data and information leaving your organization. So definitely the enhancements and better use of copilot within your environment. Is where we want to help you. Corey Milliman 34:48 Absolutely. So another one, we see a lot is we see a lot with HR, chatbots, HR agents, right. So we’re actually going to go ahead and run create one. And this is going to spin for just a moment here. And I have an hour policy document that I’ve already created, but you can see here when I go create an agent, I can try one of these as a template. So we have all of these other templates that are out there prompting coaches, meeting coach creating quizzes. Scrum assess and writing coach. You know you can use these as a baseline customer Insights assistant as a baseline. Or you can just describe what you want your agent to do and. Use natural language to do that. So I’m going to copy and paste. So you don’t have to watch me type today. And so I just have this in other document here. And I just gave it a prompt telling it what I want this to do. So one thing about these copilot agents is it also allows us to validate some of the use cases. Maybe we’re thinking, oh, I want to do this with copilot studio. This would be a great idea. We can start with some of these smaller agents that are grounded on internal documents and start understanding where we need to maybe get advanced features that are outside of copilot, where we move things over. To copilot Studio’s low code scenario. So maybe I have an HR chatbot that I put out there that’s really good at using documents. And now I want to create an integration to work day or ServiceNow or start doing other things and that’s where we would an example of when it’s time to take this from a copilot agent and bring it over into copilot studio for a low code scenario. So here. I’ve gone ahead and give it some information, so let’s determine a name. We’ll just we’ll just keep that and you can keep talking to it and telling it what you want to do. So I’m just going to give it. I have a. Wrote up on what it does. And I’m gonna drop that in here. So it’s gonna answer common HR questions. Policy enforcement guidelines and helps employees navigate benefits, insurance, payroll and career development programs. So I’m gonna tell copilot, this is what it’s going to do. And now it is taking that prompt. It’s building out my tiles. It’s building out before I even grounded it with any information. It’s taking that and going OK. This is what I think we can do, so it’s asking me another question, more details on how the agent should respond or things like that. And I’m going to skip that. But you can give it more guidance on how you want to respond. Inquiries. I’m going to click configure. And down here then I can go and choose what knowledge source I want to use so. I can go ahead and use. File sites chats. These may be SharePoint sites, team sites, different things like that. And I have to now remember the name of my policies document. Here we go. And so I consolidated everything into one document here. This is about 100 different HR policies. I could point this up to a SharePoint site. One thing I’ve noticed is sometimes is a little difficult. If you go to a site here, it’s going to pull everything from the entire site. So if you want a specific document library, you have to go ahead and click the cloud icon. Browse over to your team site and go down to that document library where you want that content to be drawn from. We have the ability to turn on different capabilities code interpreter. This is a really good way to use other use cases like a log file analysis. It does use some of that advanced code for that, but it has the ability here I can change my suggested prompts. Change what these look like. Maybe delete this one about, I don’t know how much information I have in here about benefits and they delete that one. And here I can start doing some testing and I can actually. We start looking at my guide and go. How many PTO days do I accrue every month? It’s actually a little I put a one in front of it and here we go. It’s telling me I accurate 1.5 days of PTO per month. 18 days PTO per year according to the document that I did that I put out there. So go ahead and click create. Now when you create an agent by default. It’s going to come up and be only available for you and then if you need to, you can come back in and share this with other people and socialize the link in teams or however you want to share that. So you can see here I have a link that I can copy to my clipboard and then I have the ability to change my sharing settings. So do I right now this is only for you, so you can also create agents that only you are using and these might be helpful in your day-to-day workflow if you’re analyzing a lot of data you want to you do something all the time and you. Want to create an AI agent to help you with that task. You can create your own AI agents in copilot and start delegating some of your day-to-day. And use those to help take on some of that busy work for you. So go ahead and just. I’m not gonna change the sharing. I’m gonna leave this to align me, and I’m just gonna go to my agent. And now? We have our HR policy assistant, so we can see if this information is available. I have my default policies. It’s actually showing me it’s looking for this next piece of information and again this is a really easy way showing me my salary. It doesn’t actually have a payroll schedule in my document. So but it is giving me other information it understood. I’m asking a question about pay salary so it’s giving me all the information it has in that document that it thinks is relevant. My question, so it says you may want to check with your HR department if you have another question there. If I can’t answer your question. So you also have the ability to like or dislike start uploading and downloading the responses and bringing your responses over into pages and pages allows me to start taking all the content I’m generating and copilot and I can really easily start bringing this into other office applications bring. This over if I’m building out a Word document. For doing other things, if I’m pulling information from all of my agents, I can start bringing that over and it’s gonna open up word online and it’s gonna start helping me to build out my documents so that I can use my agents to start building out content. So. That research agent that we are looking at, I can export to pages, bring it over into word and start building that content out and then using copilot. In Word to start making changes to that information. So. Tabatha Frozena 41:49 Corey, we did have a we did have a question in the chat that I think is it’s a really good one and a lot of people have questions or concerns when building these agents on risk. So do you want to just take a look at it and just address it really quick? Corey Milliman 42:03 Absolutely. So if you share any time you have an agent and if they do not have access to the source documents, they can’t. They’re not going to be able to use it. So if you have an agency created against entire SharePoint site and different people in your organization have access to different parts of that SharePoint site, different people are going to be able to get different responses based on their access that they have to that SharePoint site. So if you’re creating an agent for a repository by department or things like that, we don’t have to have it saved somewhere where all users have access to. We don’t want to create that scenario if we don’t need to. We can actually are still going to respect all the privacy controls that were on your SharePoint site or wherever whatever source you’re using it. So and also. Tabatha Frozena 42:50 And just to confirm, Corey, like if you did share it and the a person that doesn’t have access to it typed, what would it tell them? Would it tell them? Sorry, we don’t have access to it? Would it give it? Nothing. Does it depend on how you build the agent. Corey Milliman 43:03 It puts. It would say. Now let’s just say I’m. What is the payroll? Let’s just say I was. I put out this agent and I don’t have access to the information that’s behind this question. It would say it can’t find the answer to that question. Tabatha Frozena 43:23 Right. Like that’s the default, right? Corey Milliman 43:27 So if I ask a question, it’s going to look like to the end user. I’m sorry I can’t find that information, so I don’t have access to the document library where that information is stored and it’s telling me it’s trying to come up with an answer. I didn’t give it any. Tabatha Frozena 43:32 Yep. Corey Milliman 43:42 It’s really trying to come up with an answer. We’re looking at tuition reimbursement, job transfers, mentorship. It’s it’s really trying. But I didn’t give it a guidance on how to handle that. So if we tell it if it doesn’t know the answer, not to try to make it up. Escalated or something there that is one way to do that. So it the privacy me as the user the permissions that are used in the back end those flow across. So if I don’t have access to it, I can’t ask a question about it. Now. Tabatha Frozena 44:12 Hopefully Mike that helps answer it. It’s not going to say, oh, you don’t have access to this specific document, it’s just going to say you don’t have it, but you can prompt it to make sure that it’s a good customer service response and or helps so perfect. Corey Milliman 44:24 Yeah. And again, if there are things that you want to expose to everybody in the organization, then you know creating a new spot, a new document library, one thing to be cognizant of is when you are doing things like that, especially when it comes to HR documents is. What is your governance plan around making sure you’re putting the right information out there? I’ve seen instances where redline documents have ended up becoming something somebody has wanted to use for a chatbot. Meeting notes or things like that. And redline documents it doesn’t understand the redlines so it’s not going to work. Those red lines have to be accepted and approved and brought in the final content. But you do want to be careful that you are not exposing draft documents or something that you don’t want out there. So when it comes to the HR documents, I know that policies and procedures might change each year. And maybe we’re making a change. I’ve seen one instance where a customer was changing payroll frequency. And they had an hour chatbot. They had a document that was talking about pay dates and they had talked to me about the organization about it. Somebody copied that into the folder that the chatbot was using. Somebody asked a question about payroll and said you’re paid monthly now and it caused it caused chaos. So you want to be careful that the documents you are using the ground news when you that’s why those permissions stay in place. If you’re creating documents. About policies and procedures that are exposing you to risk. Then you want to make sure that you have some kind of governance plan in place to make sure we’re only putting published known good information out for this chatbot to consume. Just another thought there. Tabatha Frozena 46:05 Yep, I did want to just circle on like why this HR chatbot I think has been so popular. It definitely isn’t from AAI standpoint of how am I going to increase revenue or have a competitive advantage, but it it has been a use case that still touches every employee in the organization. Sometimes it’s to reduce help desk tickets and HR calls on payroll. But it is a way to expose. The entire organization to agents and to how to use AI. Into how to create that discipline within your organization. So it’s been and and typically it’s tied to then a handbook. So you don’t necessarily have to worry about the site or it goes to a specific document which helps with permissions and things like that, but it’s been a really great step in foot in the door to expose the whole organization, but definitely isn’t, you know, Moon shot idea. On how to use AI to to grow more revenue within your business. Corey Milliman 47:06 So I do have some other agents that were created previously. We talked about the log analysis. This is another one that was actually a log file analyzer assistant that was created to do pattern recognition. But again, this analyst tool, this is really, really powerful. I showed you that Python And excel. Another thing that is really unique that they’ve enabled us to do is directly. From copilot I can start chatting with the different agents that are out there. So I hit that ampersand and I don’t have to go find the agent and that experience is going to start moving across the entire Microsoft stack. So I could be in Word and call the. Analyst agent or call the researcher agent or call. Bring up my HR policy agent or an internal agent that we’re using to give us different types of information. So. That functionality keeps rolling out more and more where you don’t have to actually open the agent. I can actually say to my HR policy assistant, maybe if that one is published already. And so it’s keeping this so I can keep this conversation context, I’m going to say, OK. Now I’m going to bring this over and I want to go into my researcher. And I want to say, how does this company’s PTO policy compare to the industry standards? Should we adjust? And now it’s going to go out based on the information that knows from this document. Now it’s going to go into a deep research mode. And it’s going. OK, wants me to start looking at your PTO policy specific industries, particular aspects. I’m going to say just give me a report. I’m going to say I am. AI am a university. And I spelled that wrong. So now it’ll go out and start generating that. So I started with an agent that was giving me an insight to a document or a pile of information inside my organization. And now that I got those insights, they were presented to me and I can transfer those insights over to another agent and bring those across. So that’s a good way of I have this one conversation and then go agent by agent and. Go through my workflow and bring that all the way across. So I just wanna bring that up. That’s another new feature as well. Tabatha Frozena 49:41 OK, cool. Corey Milliman 49:42 So I’m going to go back to the slide deck now. Only to outline some other pieces of information that are here. So Tabatha brought up the use case. You dropped a link in chat with these different scenarios. So there are use cases here that discuss one thing we didn’t discuss today was you can do this on a SharePoint site as well. The ability to do that on a SharePoint site is. The same as in copilot. You see the copilot icon, you click it on the SharePoint site and you get the same experience. But I do have that. Here we go. So I do have these scenarios though that are built out here that talk about different ways you can use kill pilot and how SharePoint agents can help or copilot agents can help. Tabatha Frozena 50:24 Hmm. Corey Milliman 50:34 So we have finance, I put out here, we have marketing. We have HR, which is one we’ve gone through. And then I’ve also gone through and created some that I’ve actually worked with clients on contract reviews. So how do you actually use copilot to create a lightweight contractor new agent? So what I’ve done since copilot is going out looking at document libraries or different different areas of the organization, I’ve pointed it out a here are the terms and conditions I want to see in my contract. This is what I’m looking for and then brought up a second. SharePoint site that had my contracts as they came in and so I was able to ask questions about, you know, how does this compare to our standard TCS or the payment terms? What does this look like? So you can do that really easily and get a basic. Tabatha Frozena 51:23 What you don’t want to do control F all the time and find consistent language like how great to build an agent to do that so. Corey Milliman 51:32 Absolutely. Tabatha Frozena 51:33 Again, not necessarily. The moon shot ideas, but thinking about the things you’re spending, manual effort, manual time, and how you can get that time back to to do the the more creative and or just other parts of your job that you may like a lot more so. Corey Milliman 51:49 So these you actually can start with this example description. What you would like to make. You can copy and paste these prompts into copilot, the agent builder, how to communicate. You can drop in most of this and it’s going to give you starter prompts. And then here I. Say leave as is. It’s going to ask you for your knowledge and go browse to that document library. So these are real quick ways in these scenarios that I put together, you can copy and paste these and get started right away. Here’s young. Here’s an onboarding buddy. This is one where we are looking at advising new hires. You know what are things we’re bringing up to speed on? And that again could be where that research assistant could be helpful as well. And you can also use your research assistant to start combing through your company’s documents and packaging up new insights into a chat agent so that you had that distilled information. The research aid assistant that we’ve discussed, and then there’s some information about the out-of-the-box ones from Microsoft as well that you can actually use as. Templates for some of your other ones. Prompt Coach is one that is pretty valuable actually. It actually helps you improve your prompting skills, and this is another one where I can actually hit ampersand, start typing prompt Coach while I’m in Microsoft. And say, hey, help me write a better prompt to do this, or to do that, or gets the end result on the Q4. Career coach this is about looking at self-assessment, self assessments and goal setting. And then a learning coach as well. Visual creator. We didn’t actually go into, but this is going to help you create documents, help you create images for your documents and improve creativity. There are some new visual creativity tools coming out in the next month. They are starting to roll out. I didn’t bring those up today because those are not publicly available to everybody yet. So I want to stop here and turn it back over to Tabatha. That is all I had for today. Tabatha Frozena 53:56 Yes, well, these definitely aren’t necessarily true selling events, right? We want to bring you education, bring you knowledge, but if there are areas you want to take a level deeper, understand how we’re helping customers, how we can accelerate it. We do have a copilot momentum, an agent discovery. So this is a quick phone call of how how can concurrency help you accelerate your copilot adoption as well as look at what agents that might be really beneficial and some that hey, we’ve already built them. We’ve already gone through the effort on it. It let’s accelerate that within your org. We can also do envisioning session, so really taking a step back and why and what AI means to your business and how it attached to your strategic initiatives. This has been really helpful, especially as we’re getting into Q3 and budget. And planning as it comes on. So there’s an art of the possible education piece, but also a way to facilitate use cases within your organization. And we’d be happy to bring your expertise and facilitate that. And then also if you saw some fun tips and tricks here. Corey and I obviously have the content already built if you wanna bring this to your teams, have a smaller group, do a lunch and learn. We’d be happy to bring this event to you, so thanks for joining. Please don’t stop the conversation if you have questions and stuff, but we’ve really appreciated sharing some tips and tricks and hopefully you got some good Nuggets out of it. Amy Cousland 55:16 Did you see? There was one question on here, tab. Corey Milliman 55:16 All right. Tabatha Frozena 55:18 Oh, I didn’t. Corey Milliman 55:19 Oh. Amy Cousland 55:21 It was here from about AAI usage policy, a security stance. Corey Milliman 55:28 Here, let me go back out as a company. In the process of implementing DLP with, I can’t see it. I don’t. Meeting participants able to see this chat as well. Amy Cousland 55:48 Yes. Corey Milliman 55:49 I just wanted to make sure. So AI usage policies and securities dance, right? Amy Cousland 55:52 Yeah. Corey Milliman 55:53 So that is so you’re looking at. You’ve already rolled out copilot and what kind of legal or compliance cautioning do you might have here when it comes to sharing agents? Because one thing to remember, and maybe to bring back to the legal and compliance team. Is creating an agent doesn’t change the permissions or governance or controls around the content that is grounded in or that source. So just like if I open up a web browser to go to a SharePoint site me using my phone doesn’t give me a different ability to view. That same content on that SharePoint site. So when you’re thinking about agents, it’s through remote. Think of it as maybe we’re creating a different interface. And as long as our permissions and our governance controls. Are in place then yes. But if you have not gone down the path of making sure that your your permissions are correct and you don’t have broken permission inheritance and you don’t have things where somebody in accounting decided to share something with everybody in the organization, you know. Those are the things though. You do need to look at and. You do need to look at the DLP policies, purview labeling and different things like that. So you might want to maybe start out with. Only allowing agents. To be used individually instead of being shared out to the organization. So you start going down that path of understanding where your risk is or being very specific and only doing departmentally or on a SharePoint site. So we can actually see where the agent is and where it lives. Tabatha Frozena 57:28 Yeah. Corey Milliman 57:29 So that may be one way to handle that, but when you’re sharing those agents, everybody in the organization, that’s when you want to make sure that the agent I’m sharing is grounded in the information. I really want everybody in the organization to see, even if we’re relying on permissions. To trim that, we really do want to be aware of how we’re sharing, because it’s just like emailing everybody a OneDrive link, right? So everybody do. I want everybody in the organization to see this document and that’s really where it can get difficult. So good that you’re implementing purview good that you’re doing the labeling of the DLP absolutely. Tabatha Frozena 58:01 Yes. Yeah. And I think I think you hit on it the other just way maybe for certain types of agents, you might wanna build them differently for those governance and controls to this copilot, individual use might not be the right fit so would be happy to talk more. About it, but hopefully that was that was enough to give you some perspective. Great. Corey Milliman 58:20 Awesome. Perfect, I’ll reply. Thanks all. Tabatha Frozena 58:22 All right. Well, thanks everybody. Great job Corey. Thanks guys.