Al About Microsoft Copilot 365

So good afternoon, everybody, and welcome to today's learning objectives webinar. And we are going to be talking all about Microsoft Copilot for 365. As mentioned, I will record this session and then send you the recording and the slide deck. It is unlikely that I will cover every single thing in the slide deck, so you'll have that for reference later. Also, if you have any questions, just send them to me in the chat and I will attempt to answer them as best as I can. I will also open it up for questions at the end as I usually do. So let's go ahead and get started. So this is to remind you, because this is a screenshot I took when I was interacting with Microsoft Copilot. I was asking it where the pen feature went in the new Outlook, and it gave me the instructions. So in addition to getting to summarize, to getting it to generate text, generate images, help you plan for your day, all those kinds of things that we're going to talk about, you can also ask it questions about using the technology that you're using. So I said, where the heck does the pen feature go? Because I couldn't find it. And it gave me the instructions and told me what I needed to know, which was it was contextual. But it also said, if you still can't find the pin feature, you might wanna check out this detailed instructions from this presentation that I created. So it is looking at your data because it's looking at my computer and my OneDrive and saying there's an article here. So one of the things that I thought immediately when I saw Copilot come out was that this would be a great tool for knowledge management. In terms of looking through your OneDrive, SharePoint, Teams, email, and helping you kind of amass the knowledge from groups. And the more you're all in on using Microsoft 365, the more chance of success you're going to have in terms of using it as a knowledge management tool, kind of leveraging what people have done, what people know, that kind of stuff. I will tell you right now, I am the only person in my organization using Copilot. And that is because I am the beta tester and trying to figure out what works, what doesn't, all that. I'm still 1 month in on this product, so I'm a newbie. I'm still learning. And so at any rate, those are all my disclaimers. Also, when I do the demonstrations, they'll be doing— be doing them live in my personal 365 tenant, so subject to eccentricities, as you may see. Also, I owe Ben Shore from Microsoft a huge thanks because he did two presentations at ABA Tech Show on using Copilot. So I have definitely learned from what he knows and also Subsidized with my reality. So what is Copilot? Copilot is technically— the name of it is Microsoft 365 Copilot, and there's a variety of different flavors. The one that we're going to be talking about today is the Copilot add-on for a Microsoft 365 plan. So that Microsoft 365 Business subscription either Business Standard, Business Premium, E3, or E5. We are not talking about the Home version, although you can get Copilot in the Home version. That's not what I'm here to talk about today. So when you go to look at the plans, it says there, there's 365 Copilot Chat, which is free. It is free. It is available. And you can download it, and it basically is a secure chat powered by GPT 4.0. That's ChatGPT. It does maintain IT control and enterprise data protection. It is not, however, embedded in all of the Microsoft 365 apps. And that's the really— to me, that's the big difference, is that with Microsoft 365 Copilot, you get Chat plus you get 365 Copilot apps in Teams, Outlook, Word, PowerPoint, Excel, and all the different, um, applications within the 365 suite, which I wrote an article about yesterday. There's so many different things just, you know, beyond Office that you get with a 365 subscription and kind of knowing which one to use for what. It's a little confusing and there's a lot of redundancy, but depending on your situation, what kind of practice you're in, what kind of firm size you're in. Those are the determinations you'll make in terms of what you actually use out of it. But it is $30 per user per month. It is, um, that's an additional to the $12.50 per user per month for Microsoft 365, or if that's Business Standard. If you use Business Premium, I think it's $22 or $23 per user per month. So So I think the challenge here for us is to figure out how do we make a $30 additional add-on product worth the cost, because it's not insignificant. And what I've been challenging myself to do is figure out how can I get a dollar a day worth of use out of this product? And then see if, um, I can actually use it every day to save myself some time. And I'll definitely give you some of the examples that I personally use. So remember that this is an assistant. I know that y'all have been told this so many times at this point that I hardly need to say it, but just as a reminder, it's not going to do the work for you., and it's not going to necessarily be all that good. And the more you use it, the better it gets. But the reality is the better you are at prompting, the better it is at delivering helpful information, but you're always going to have to review the work. It's not something that you want to show anybody. It's just a personal tool for you to use. So it's, it's not going to be doing your thinking for you. The other thing is, again, the more you're kind of all in with 365, the better it's going to be at unearthing information for you. And then the other thing is, say you're, you're excited because you're going to go and try to get information or get it to help you with Excel. If your Excel spreadsheet isn't formatted in a way that it understands the data, it's not so great. And the other thing is, what can it see? So it doesn't train the LLM on your data. So the large language model that underlies this is OpenAI's LLM, and it's not training the LLM on your data, and it is respecting your 365 permissions. So what, what is actually What does that mean? That means that your data goes up to Copilot, looks at the public model and possibly the web if you determine that you want to use the web as, as part of the, the dataset. But it's not training on your data. It only sees what you personally can see. So depending on the size of your firm or organization, you're going to have permissions set up to be able to see certain things. So for instance, you can't see my personal Outlook unless we have set up a shared inbox. So if you can see someone else's inbox, Copilot can see someone else's inbox. But if you can't, it can't. So Bob can't see Alice's data. Which means that Bob's Copilot can't see Alice's data. So if you have set up, say for instance, in your firm you've got some bookkeepers, you've got an HR professional or a legal administrator, office administrator, or an IT group, and they have access to information within 365 that you don't see, or you've got, you know, just different kind of isolated bits of information that not everybody can see. As long as your permissions are set up correctly, then everybody can't— you can only see what you're allowed to see. The reason I'm making this disclaimer is because kind of early days when people are starting to implement Copilot, especially across larger organizations and companies, they would start using Copilot and it would start pulling up information that technically they shouldn't see, like HR information with salaries and things like that. That was because the permissions weren't set up up in SharePoint correctly to disable their ability to see that information. It's just that it— they didn't see it because it wasn't readily available from their screens. But Copilot is digging into all that information. So just be aware that if you do have sensitive information, if you have ethical walls set up, things like that, you need to make sure your SharePoint permissions are set up so that people can't— only the right people can see that information. And that's something an IT administrator would do for you and make sure that that was all set up properly. The other thing is, and this is the Ben Shore rule again, Ben is my friend who works for Microsoft, and he's been giving presentations on using Copilot all over the country, and I would have him do it here, but, and he is fantastic and very realistic, but he does work for Microsoft, and so he's probably not in a good position to tell you what I might tell you, which is sometimes it doesn't work. So we may see some examples of that right away. What can you use this for? It's good for creating and editing, asking and summarizing, brainstorming, but it's not always perfect and it's not really going to do the work for you. It's not, you know, sometimes it can do things for you and I'm gonna show you some examples of where I think it's really useful and some things that just doesn't do all that well. And that may be the fault of the end user, the wetware here, because I'm not prompting it accurately or effectively. And this, again, we're all learning how to use this and figuring out what our use cases are. My use cases are going to be different from yours. The person sitting down the hall from you is going to have a different use case than yours. So it's really a matter of figuring out what it can help you with. And then when you think, well, this would help a certain group of people, you can actually share your prompts. So that's another thing that we've been seeing a lot of is people creating prompts and then sharing them out in libraries so that people can see what has worked. When you're working with Copilot, you, you are having a conversation with the AI, and if you're not familiar with using artificial intelligence tools at this point, it— this may be this may be feeling a little weird. Like, I really felt weird waking it up and saying, good morning, here's what I'm going to be trying to do today. And, you know, it, it actually likes to have a little conversation with you. The other thing is, this is all an experiment, so just try stuff. Sometimes it will work, sometimes it doesn't. And then of course, you cannot just automatically trust what it tells you, so you do need to verify it. And one person just said creepy, and I agree. When I am having a conversation with it and asking it how, how it's doing this morning and asking it to help me, I do feel a little weird that I'm having a conversation with an AI and almost feel like I'm training it to be more humanistic, or at least try to act that way. But There it is. That's what we're supposed to be doing. The other thing is, um, the more back and forth, the better it gets. So just asking it a question and then having it go, I'm sorry, I can't help you, or I can't do that, don't give up. Keep having it back and forth. You can say, I'd like you to be able to do this for me. How is, how can I best get you to give me that result? And then Retry it. So there's a certain time commitment in this. While it does speed things up as you're learning how to prompt it, you may actually be slowed down a little bit, and you may be tempted to go back and do things the way you've always done them, which is also absolutely fine. We don't have time to spend all day trying to figure out how to get the prompt right. And so sometimes you go, I know how to do this. So speaking of prompting, It's still a thing. It's getting better in terms of we don't have to be as good with prompting, but there's definitely some tried and true ways to prompt the AI. And this, this holds true for not only Copilot, but also ChatGPT or Anthropic or Vincent AI or any of the tools that you're using where you're trying to get it to generate Information, the more you prompt it effectively, the better it works. So I'm gonna give you some examples here. So the better you prompt, the better it does. So if you write, write a blog post about probate, it's going to give you its best guess at what that is, but the more information you give it, the better it gets. So we want to think about it in terms of giving it a goal, giving it some context, telling it what our expectations are, what sources we'd like to use, and just keep going back and forth with it and asking it to refine and improve. So here's, here's kind of the, the example of that. Write a blog post about probate. So write a blog post about probate in the state of Illinois for an audience of seniors who aren't very sophisticated. So we're, we're identifying, you know, who our audience is. And then we're going to give it some expectations. What, what tone do you want it to write in? And include information about designated beneficiaries, appointing executor. One of the things I like to point out a lot is that it's drawing from a huge corpus of data and it can sound smart, but the more information and contextualization you give it, the better it is. And that's something that, you know,, and so you're not excused from providing the context, the information, the points that you want it to make. And I find that the, when I'm using my own data as the source of information, it, it does the best job. So, um, then you can give it sources. So I can say, use this website, um, this document, you know, and so you can put this thing called, uh, it's a contextual IQ. So you can see you write a blog post, keep the tone in this way. Whoops. Be sure to include information about these things and then prioritize the information from these documents that we have in our OneDrive or in our SharePoint and also from the web. And so the way you do that is you go through any type of front slash and it gives you a picker and you can pick recent emails, files, Teams chats, all sorts of other things. And you just put that slash and then you start typing in what you want it, which means that you have to kind of know what information you have in there. And so, um, sometimes I will put a bunch of different files in a folder to start getting it prepared to help me contextualize. And then once you get the result, you don't have to stop there. You keep, you keep conversing with it. Say, ask it to make that a little bit longer. Sometimes I'll give it a word count. I'll say, write a 1,500-word blog post on blah, blah, blah. And then it gives me an 800-word blog post. And so I'll say, no, can you make it longer? I really need to have this be about 1,500 words. Or can we be a bit more casual or conversational? Can we add this section? And then you can actually say, how would you improve this article? What do you, you know, what do you know from the vast and gory information that, you know, to make this better? So you just keep prompting it, you keep— and again, this is to generate a whole bit of text. There's a lot of shortcut things that you can do with this that I find I use more frequently, but Those are kind of the tips. And then the other thing is when you want to have it change the conversation, rethink, you know, do a new activity, start with a fresh chat because it's actually working in context. So it's building off of that whole thread of conversations you've had with it. So if you continue to add new questions, but you're changing the topic, it's actually looking back at that, um, the context that you've been having the conversation. So when you wanna start a new activity, you have to start a new chat. The other thing is you can use metaprompts. You can actually ask it to help you write your prompts, and this is a good way. I mean, that's where you get some of the best outputs is when you say, I need you to help me. You can act as my prompt engineer. How do you know? Here's what I need you to do. And then this is an example. How would you describe the voice and writing of this article? And then it'll show you what that is. So there's a couple of different elements to this, but one of the bigger elements that weirdly isn't built into the apps itself, it's kind of a standalone Copilot chat, which you can actually start out with. It's helpful when you go into Outlook and use Copilot or go into Excel and use Copilot, but you can also start with this screen called Copilot Chat, and I'll show you where it is. And it gives you a bunch of suggestions about what you might want to do with it today. Within that, you can choose to access just your Microsoft 365 data and things that you have permission to see and/or search the web. So it's got the built-in search, um, and you just toggle between work and web. And so here's an example. This is on the web. Give me a concise summary of news about personal injury law in Canada. And here it goes, and it gives you— cites the sources, and it actually puts the, um, the link to that source in the, uh, the text that it gives you. So then this is the difference again between give me a concise summary of about personal injury law in Colorado versus giving it more context, a better prompt. I'm an attorney in Colorado. Give me a concise summary for each news item. Suggest ways I might communicate that news to the clients. So if you're putting together a newsletter for your clients and you wanna give 'em an update on Colorado law that has to do with a certain, um, practice area, This is a great way to use these kinds of tools. It searches the web. It helps you think about how to communicate with your clients. You can get it to write in language that is easy for them to understand. And so again, this is a good use. You know, we are really not talking about using Copilot to do deep research and legal, like legal research drafting, things like that because it's not really designed to be a legal research tool. So that is kind of a limited use case. And there's certainly better tools out there that, you know, Lexis Westlaw, Vincent, Bloomberg, Dscribe— there's tons and tons and tons of legal research products on the market that are designed to help you do legal research. Copilot is not one of them. And then it has agents, and we'll— I'll go into agents and, and show you those a little bit. This age, you know, the promise of agentic AIs that actually can do something for you. So help me plan a Hawaii vacation. It's limited in its ability to actually do this right now, and you still have to go through and do a lot of work. That said, you can build your own agents based on templates. Within the, the Copilot Studio, and I'll show you what that looks like. And then the other thing is that if you are using SharePoint, you can write SharePoint agents. So if you've got a document library in SharePoint set up, you can actually write agents to do certain things, and the agents will you know, notify you when a new document has been sent to this folder or things like that, which you can also do through Power Automate. Or you can create an agent that says when this file is loaded to this folder, rename it. Or, you know, depending on what your use case is, you can create these agents. The other thing is it gives you a whole prompt gallery. For you to be able to identify good prompts and useful prompts. So you can just copy and paste prompts and use them. So you can just click on any of these prompts. These are the Microsoft prompts. Well, another thing that I mentioned, save your prompts. So within the Microsoft prompts, you can type job type and select from the app. So if you wanna see Outlook prompts or you wanna see Excel prompts, you can sort by app, the task you want to do, the job type, um, and then if the job type is legal, then it starts suggesting a bunch of different options. Analyze case law for risk management. Again, be extremely careful with this. This is not a legal-specific product, and so your mileage may vary, but if you already have the data in 365, then this may actually be quite helpful because you can start mining information and leveraging content you've already written, contextualize by saying, look in this folder and that kind of stuff. And that could actually be quite useful. And then within the prompt gallery, the other thing you can do is save your prompts. And also save your prompts that work really well to team prompts. So when you're going through and you've created a prompt that works really well, you can mouse over it and just click this little, um, kind of bookmark icon, and that will save the prompt. And then again, you can go in and view prompts, um, to see what's working for people. If you want to share your prompts, those are where they save them. And then you can also share the prompts to team prompts. And so that's, you know, sharing it there. So if you find that you've created a prompt, set a reminder in 60 days to check on improvement in the data, include these people, and then you share to the team, and that shares it in the, say, in the shared prompts. There is a whole Microsoft Copilot Academy. So, and, and in this slide deck, I've put links to all these things so you can go and see it. But if you want to learn more about using Copilot and you wanna get some training, they provide free training. You can kind of watch videos and demonstrations and tutorials. And again, but what I found is the more you intentionally use it,, the better it is. So that's probably the biggest challenge for me is to remember to actually use the thing. Um, it, it's constantly popping up and suggesting that you use it, but in our day-to-day world of trying to get our work done, sometimes even though you've purchased this tool and you think, oh, I, I should really be using this, you don't always, um, actually actively use it. So We are going to walk through this. A couple of disclaimers. This product is still in beta and many, many, many— well, I would almost suggest that all of the products that we are purchasing in the GenAI space are still in beta. We are the beta testers. They're constantly changing, constantly improving, constantly sending out new versions, new updates. And so while we are paying for them, they aren't perfect and they're still learning. And that's just one of the things. The reason that we don't have a lot of free tools, and there are a lot, but things like this aren't free because it costs them a ton of money to generate. And they, you know, as you use it, it gets better and then they continue to make it better. But for instance, I had Copilot create the slide. And I told it highlight the beta status of Microsoft 365 Copilot, emphasizing the importance of user feedback and refining the tool. And then use the quote from 2001: A Space Odyssey, "I'm sorry, Dave, I can't do that," to add our humorous reference. So the image that it provided for me was not this "I'm sorry, Dave, I'm afraid I can't do that" from A Space Odyssey. It just put some generic and weirdly spelled graphic in there. And when I went to go ask it to improve it, I said, thanks, you referenced the quote regarding I'm sorry, Dave, but it doesn't appear in the slide. Can you add the quote in place of the image? And it basically said, I'm sorry, Dave, I can't do that. So it says, I understand I can't make changes to your presentation directly. Here's some other things I can do. So Again, it's a lot of trial and error, um, and sometimes it can be a little frustrating because you think, oh, I would really like it to be able to do this for me, but we know that we can do things the old-fashioned way. And so you just go back to grabbing an image off the web. So looking at— and this portion of the slide deck, by the way, I created from an outline. So it looks a little different from the previous slides because these slides are all created from an outline in Word that I uploaded into PowerPoint using Copilot and had it generate the slides. And I didn't really make any major changes to it. And that's the kind of thing that can speed time up pretty dramatically. So the first thing I wanna look at is the 365 Copilot chat. And that's the thing that we looked at, and that's kind of, if you think about it, it's, it's your hub. And so I'm gonna bounce over to that. Where you find Copilot Chat once you've got the thing installed and you've subscribed to it, when you go to Microsoft365.com, then over here, if you're, it'll start you, it may start you out in Home, and Home is you see they've renamed it. I'm not even going to try to figure out what Microsoft is doing with file naming conventions, or rather not file naming conventions, but product, um, titles. Anyway, now My365 is called Copilot, um, and One of, one of the attendees said, thinking Copilot's attention to permissions, could it be possible that it did not pull the graphic you needed because the Fraser image was copyrighted? So that's a good question. If I went into an image generator and told it to create an image with that, that text, it may create the text. It probably would not give me that actual image from the copyrighted movie still. So it— but how well it's actually ignoring my request because it was copyrighted, I'm not entirely sure. But let's hope that it actually