00:02If you are a beginner wanting to learn how to use Copilot Studio, you have come to the right place

00:04Welcome to my complete tutorial for beginners

00:06I'm going to take my time, this one, it's a long video, it's broken up into sections so that you can get to the place you need, if you already know some of the basics, and I'm going to take you step by step through everything you need to know to build your own co-pilot agent, as we're calling them now, with Copilot Studio

00:24So what we're going to be covering in this video is, in the first section, how to get started with creating a co-pilot by describing it or by using one of the

00:30Built-in templates, how to add knowledge, and there's lots of different ways you can add knowledge into your co-pilots, from public websites, uploading documents

00:42We'll talk a little bit about other data sources like SharePoint and Dataverse in there too, the importance of building out topics and controlling the way the conversation flows, which is called conversational orchestration, why it's not just all AI all the time, and there's good reason for that, how to use entities and variables, which are some really clever things that help the co-pilot understand what's going on in the conversation

01:03Actions, your co-pilot agent can take action on your behalf by going out and using flows to retrieve things, or by getting other pieces of information for you

01:14We'll talk about the options you've got for publishing your co-pilot, and then some of the other things that you can do with it

01:22There's a lot in here

01:29If you get value out of this, please, please help me out by liking, commenting, sharing this with people, helps it reach more people, and YouTube will help get it out to all of those other people who want to learn this as well

01:36I am actually rating Copilot Studio as, at the end of sort of late part of 2024, as the most Strategic and useful skill that you can learn in Power Platform, you'll see that as we go through, but essentially this is the tool set now that you use to build out your own Copilot agent to extend Copilot for Microsoft 365, or any of the other first party Copilots

02:00With all of the new things that have been announced coming through in the Wave 2 of the Microsoft 365 Copilot, this is getting even closer to that experience and more relevant, and the road map of where Copilot agents are going, which is more that I want to cover here, is a very bright future of interesting things that we're going to be able to do

02:18I would really encourage you to come on this learning journey with me, and I hope you get a lot out of it, so let's get started with part one, building a co-pilot

02:26Everything happens here

02:34You just want to go to your browser and go to this website: copilotstudio.microsoft.com, and sign up for a free trial

02:38Now, everything I'm showing you in this tutorial, I'm working on a free trial, so you've got no problem

02:43You don't need to pay for anything until you're ready to actually publish your copilot agent and put it out into the world

02:47You will need a work or school account to sign up for that free trial

02:53Now, this is the main Landing page, but when you first come in and sign up through that trial process, it usually lands you here, and this is Copilot helping you build a Copilot

03:08There's a fair bit of this kind of experience going on in here of Copilots helping you build Copilots, but that's good

03:14So it says, "Hi, I'm here to help you build a Copilot

03:19In a few sentences, how will your Copilot assist your users?" So the scenario that I'm going to work through here is using a website that is local to my part of the world here that's called The Better Health Channel

03:25This is a series of sort of articles and pieces of advice on health

03:32And Health Services, it's a really good website because it's very information rich, so I'm going to be using this scenario of creating a co-pilot that can help people get advice on healthy living, so let's start with that

03:44So we give it a definition here, uh, so how will your copilot assist your users? It will help them find information about healthy living

03:57Now, you can put whatever you want in there

04:02I will show you where all these pieces go

04:06This setup is actually helping you put together step by step, essentially something that sort of starts by connecting to a website, you don't have to start this way, you

04:10Can actually sort of jump straight to this button here that says "skip to configure," but if you're a beginner, this is a really nice way to get started

04:22So, do I have any instructions for how your co-pilot should assist, for example, your specific tone? I'm going to keep this one fairly general, but you can add a fair bit of personality in here, so be friendly, provide clear responses, do not use medical jargon

04:43You can put things in here like, "use emojis," maybe we should have done that, we can come back and change it, and I'll show you

04:52"Use emojis," um, you can, you can say, "use very casual language, make sure you summarize." Everything in bullet points, so you can actually put things in there that define exactly how the co-pilot is going to respond

04:58That's called instructions, custom instructions

05:04Where should the co-pilot find important information? Now, at this stage, this is starting to add knowledge

05:10The only knowledge that we can add in at this stage is the public website, but we will be able to go in and edit this co-pilot to add more things

05:16If you want to create a co-pilot that doesn't use a website, you can just skip this step

05:22You can just say no website, so I'm going to put that whole website link in there, and it's going to think

05:29About that one a little bit more, are there any topics or tasks Copilot shouldn't help with or talk about? So let's say we don't want it to actually recommend any specific medical services or health practitioners, so we'll just put that in there

05:37So what I've done now is given this some guidelines

05:46I've given my Copilot guidelines about how I want to set up the Copilot

05:50I've given it a purpose

05:55I've given it some instructions about how it should respond, where it should get its knowledge from, and something that I don't want it to do

05:59It confirms all of those things, so that is good

06:04We can just click create when

06:09We're done, and it will go ahead and set that up, so this is going to put all of those things inside the Copilot

06:12It's going to actually deploy a Copilot for you, just on the language

06:16I'm going to go back and forth a little bit here between calling it a Copilot and a Copilot agent

06:22I've always only called it a Copilot

06:27Just recently, the language from Microsoft is starting to call these Copilot agents, so if I flick back and forth, essentially the same thing

06:31Copilot agent is just the more modern term, and I haven't quite caught up with that yet

06:35All right, so we have a Copilot here ready to go

06:39Let's have a look at

06:42Where all those things went, when I described the purpose of what I was doing, that's helped it to choose a name, which is fine

06:46It's brought all of this in here into these instructions, so this is taking what I asked it to do in the first place, its purpose, and be friendly and provide clear responses, and avoid using medical jargon

06:58So all of those pieces I've put in there have been combined into this piece that's called the instructions, which is that sort of, it's like an overarching prompt that this co-pilot agent has

07:10Every time it produces a response, it uses that as a guideline

07:16We can change that as we go, so that You see what's going on, we've got knowledge in here, so it is connected to this website

07:22I don't know if you can hear the birds chirping outside

07:28Where I live, it's getting towards the end of the day, and they're rainbow lates, in case you want to look up what that bird noise is outside

07:32And if you can't hear it, then I can hear it, and I'm mildly distracted by it

07:36We have got in here the Better Health Website that we just put in, and the other thing here is that the general knowledge is enabled

07:40That means it can use the large language model to sort of figure out answers for things that might not be just in there, we'll Have a look again later at what the difference is between those things, so we can actually give this a test run here

08:01So, if I come in here and have a look at something that's on this website, you can see we've got some information here about sleep and managing sleep and fatigue

08:06So, let's give this a go

08:12We'll come in and say, "How can I manage fatigue?" And this, as I said, it's only created it 2 minutes ago

08:18It usually takes a couple of minutes to tee this up, but within only a few minutes here, I've managed to create my first co-pilot agent, and it's connected to a website, and it's answering questions

08:24Now, why is there a Great big long video

08:31Beyond this, you might, you might ask, realistically, this is a really good sort of party trick of being able to go look at this

08:38I've pointed to a website, here it is, it's all working

08:43There is so much more to this, and I can't stress enough how important it is to not just create a co-pilot, point it at a website, and go, okay, great job done

08:50It's a good exercise to go through for learning, but there are many other things

08:57You don't want a co-pilot, most of the time, that is just completely dependent on one website and just using generative answers, generative AI, to produce the answers to that website

09:04I mean, If that meets your needs, good luck to you, and thanks for watching, but realistically, there are always specialties in saying, but if someone asked this kind of question, I need it to answer this way, and I need it to refer to these other knowledge sources, and I want it to do other things, so we're going to go through all of that

09:23There is a skill in this, in the thought process of the actual design

09:28I'll try and talk through as much of that as I can as we go, but these are sort of learned skills, the more experience you get

09:34So, in the response here, fatigue can be managed in these different ways, and there's a couple of References, so you see it's coming back here with the fatigue fighting tips

09:40So if I click through, that's going to open that source in there

09:45Now, I am just working here in a test canvas

09:49The other thing you can do is if I click on here, we can open this track between topics

09:53I'm going to show you another experience as we go a little bit further into it

09:59If I refresh this one now, and if I got that one still on, uh, so you could ask it to say, give me, uh, three tips as bullet points for managing, let's spell it wrong because it can handle that

10:16I know how to spell, it can actually

10:19So this is the, the concept here is we've got this large

10:22Language model, it can actually sort of interpret and understand things, so there you go

10:26I've asked the prompt in a different way, and it's actually come back, so it's capable of doing that

10:29What we've got in the middle here is the canvas that's actually showing the flow of that conversation and where it went

10:35So when we come back in a little bit to the idea of topics, this is actually what's called a fallback topic, so this is basically just a sort of a default position within the co-pilot to say if of all of the other things I've told you how to answer, you can't find the answer, just fall back to this one at the moment

10:52This copilot is literally just a single fallback topic with that website, so that's all it does, but we can build other layers on top of that

11:05So let's take a look at this concept of knowledge and what this is doing in here

11:08We will come at the end to actually sort of publishing this all the way through here

11:12I'm going to work just in this test canvas

11:16This is the experience for building, but in the end, the idea here is that we are building a co-pilot that's going to be published out somewhere, either as an internal co-pilot for your users or as an external co-pilot on our website or various other channels, or you can

11:29Use this as an extension to your Microsoft 365 Copilot, or Copilot for Sales, or Copilot for Service, or Finance, or all of those things as well

11:35So wherever it's going, in the end it's going to go out into the world

11:41This experience is, uh, is just how we are testing it

11:48So let's come into this knowledge section here, and this is where we're navigating all of the experiences about adding knowledge into the system

11:54So at the moment, what we've got here is a website, and it's ready to go

11:57We can come in here and open the details

12:01Let's edit this one and have a look

12:05Now you'll see the knowledge description here is actually

12:08Both mandatory and blank, which is interesting

12:12When you go through the process that we just went through of the quick create, it skips over that part, but I want to come in here and show you this, 'cause this is actually really, really important

12:17Here, what is going on here is that this is a description of what this knowledge does

12:24So, at the moment, it just creates one topic that just goes off to that website, it doesn't matter, but you imagine you might add five or six different websites in here, or websites and documents and lots of other things

12:37This description is like a, it's almost like a prompt that helps the AI

12:42Understand what this is actually doing, so it's called Better Health

12:47So let's, at the outset, sort of tidy this up a bit

12:51There's the web page link, but let's call this the Better Health Website, because we can have a name that's not the website

12:56And then the description here is something like you want to say something about what it actually does, so this is providing information about, uh, healthy living and information about, um, health conditions and support services, something like that

13:06You will see later why this becomes important

13:16Let's just save that, because we're going to start using AI to help the

13:26Co-pilot, it determines which pieces of information it should use when, and so that description, and it's worth, I'm doing a bit of a quick job there, but it is actually really worth thinking that through and taking your time and refining that, can really improve the performance of what you're doing

13:38Before we go any further with adding additional knowledge sources and talking them through, I want to show you an example of a website where something doesn't work quite so well, because this one, you might realize, I've chosen because it works beautifully, it is a very knowledge-rich website, all of the information

13:56Here is on the page, there's a lot of information on that page, and the co-pilot is actually able to read all of that and provide really good answers, so this is a really good example of a website that works well

14:08One of the tricks or traps that people will fall into is if you have a website that has a lot of JavaScript or elements that are displayed as tables on the website, the co-pilot can't actually read those

14:19So this is an example here of, uh, Melbourne pollen, where I live in the world, we have a very high pollen count, there's a whole, a whole website, uh, dedicated to, you know, people who have asthma and conditions that

14:31Need to be aware of this

14:43Thankfully, at the time of recording this, it's all, uh, in the nice low category, but something like this, right? This is a beautiful table where we are able to look at what's going on in here, but if I grab this website, for instance, and let's come back into my, um, my bot here, I can add another website

14:51So this is just sort of showing you what you can do with adding more things in here

14:59Again, let's give this a description, so we'll say, "Provides the pollen forecast for areas of Melbourne and Victoria," and add that one in there

15:07We'll give it a second to spin up while I explain what's going on here, so as a

15:13Human, I can look at this beautiful table here and determine exactly what's going on

15:17This is designed for very good human readability

15:22I can come in here and straight away say, yep, East Gippsland is low, and see exactly what's going on

15:26If I have a look though at the page source, so let's click right click here and say view page source, and if I do something like, let's find the text for East Gippsland, which is a region in the state sort of near where I live, but if we go through here, this is actually the table

15:43So you'll see here we've got the forecast cell with the grid, but if you look at this, what I'm looking at as a human, On the screen is a beautifully laid out table that has one piece of, and another piece of information about the forecast

15:56This has now got East GPS land, and the fact that the forecast level is low is sitting, I don't know, it's kind of hard to see, it's in there because this is all formatted in a way that's designed to work for displaying it and not actually in a, it's not actually in a structured format or a structured table that an generative AI is going to be able to uh read, so let's close that one down and I'll show you what happens here, so if we come back into to our bot, and bot that co-pilots used to be called chat

16:10Bots or bots, I'm just a little all over the place with keeping up with the language

16:25Bot equals chatbot, equals actually no, an agent is more sophisticated than a bot

16:39Forgive me, I'm going to try not to call it a bot anymore

16:45What is the pollen forecast for East Gippsland? And you'll see it's gone into that knowledge source, where we looked at before, and it says to find the pollen forecast, you can check local weather

16:54Okay, so that hasn't helped me particularly

17:03What this is doing, where it says AI generated content may be incorrect, what's happening, if I go back to my overview here, is that it's using this, so spoiler

17:12Alert, I've already told you it can't find the answer in that knowledge source, but I'm not really proving it very well here because I've got this enabled

17:19So this part is where you get this message, "AI generated content may be incorrect," means it's using the same way that if you go to just copilot.microsoft.com on the public website or Chat GPT, that large language model is able to answer things for you, that's what this is doing

17:37So you will have use cases where that is highly useful and desirable, and you'll have use cases where you say, "No, I want to lock down my co-pilot so that the only thing that it can answer." Is the knowledge sources that I give it, so if you set that feature to disabled instead of enabled, now it can't use its own general knowledge to be able to answer the questions

18:03Now I've said, you only know what I've given you access to, so let's see what happens here if we test the difference

18:09Um, I've actually got a whole other video on that feature in particular if you want to go deeper into it and showing you some of the things that it can do, so we'll give that a refresh and we'll ask the same question

18:18I am also going to give it a question or something it, it definitely can answer, so as much as it can't handle this sort

18:26Of table format, if I go into the sort of About Us section, this is a nice text based thing

18:34So let's come back in here and say, um, tell me about the Melbourne pollen service

18:40So this is something it can do because it can read this off the website

18:44So it's gone in to find that data source

18:47There we go, gives you the information, gives you the link

18:50Let's try the forecast again

18:54What is the pollen forecast for East Gippsland? I'm always worried doing, doing live demos

19:00There we go, can't find it

19:07We're going to have a solution for this later when we get to the section on actions

19:16I'm going to show you how you can make this kind of data

19:25Accessible in a properly structured table, but for now, just be aware that if you're using your website as a knowledge source, if you've got a lot of, if you've got a lot of JavaScript or a lot of beautifully formatted elements and things on the screen, the generative AI may not be able to answer that

19:34It's all done based on the indexing that happens underneath, and that's the same indexing that Google and Bing used for the search indexes, so if you can't find it on Google, or you can't find it on Bing, or you can't find it using copilot.microsoft.com, this isn't going to find it either because it's not accessible to

19:52The index, so something to keep in mind, if you've got control of a website that you're trying to connect to, that that will be a limit

19:57So let's go back now and have a look at some of the other knowledge options

20:03So the next thing I'm going to show you is that you can upload a document into here

20:07So I have got a document in here somewhere

20:11Let's take a look at my sample document here

20:15I have got a PDF that I've downloaded

20:19In keeping with our Healthy Living, this is the Australian dietary guidelines

20:23So we've got standard serves of things in here, so this goes through and tells you what a standard serve or vegetables fruit Grain, all of those things, looks like so we will just close that one down

20:34So let's come into here, and now I'm going to add knowledge, and I'm going to drag and drop a file here

20:41So I'm going to go to my other monitor and just grab that PDF in there, and again, let's be good about giving it a description

20:47So this one, I think I've got this one here already, information about, I'll just paste it in and describe it to you

20:52So I've put the description in there: information about the Australian dietary guidelines, including standard serves of all the major food groups, and recommended daily intake

21:00So I've given that quite a bit of

21:04Information

21:09Now, what this is doing is actually uploading the document into Dataverse underneath

21:14Don't panic, this is, it's not actually creating a database or anything in there, but it's using that Dataverse indexing to process that document

21:19So while it's sitting here in progress, it's processing that document and using that indexing to be able to understand it

21:23Obviously, the larger the document you've got in there, the longer that's going to take

21:27There's a range of different file formats you can use in there

21:31You can put in EBS, PDFs are obviously the obvious one, Word documents, so there's, there's a Bunch of things you can, you can put in there

21:36You were able to put Excel files in there

21:40I wouldn't recommend that

21:45It, it, it gives you a bit of a, uh, it gives you a bit of a sense of feeling like you're putting a structured data table in there, but Excel isn't actually properly structured, so this works best with chunks of text, unstructured chunks of text, so any document in that kind of format

21:57This is one where I have done a little bit of editing, because even though that ready pops up within about 30 seconds, it does take a little bit longer than that to actually work, give it a few minutes

22:03This is a good way to go away and

22:09Have a beverage of choice, or do something else

22:14Now, if I come in here and say, "How much is a standard serve of bread?", we will see that it's actually able to draw that response there from the Australian dietary guidelines, serving slice is approximately 40g

22:25So, if I click on that, you'll see it comes up here

22:36Now, this is different from the experience of connecting it to the website, because when we connected it to the website, it actually redirected to that website, whereas when you upload a document, it's taken that document actually into the co-pilot knowledge, and it will actually show the section

22:46So, we've Got here the one slice of bread, you notice you sort of lose all the formatting here because the original document looked much nicer than that, so again, you got to think about the use case for this, but it's very good if you've got knowledge that you want to be able to sort of upload and have the co-pilot refer to that, maybe doesn't or can't for some reason exist on a public website, and where it's an unauthenticated use case, putting it on SharePoint means that people would have to authenticate in

23:17Let's take a look at those other knowledge sources now, so if we come in here and say let's add knowledge, there's

23:24SharePoint, I'm not going to do these two here because my use case here is very much external, but if you'd like to see videos on the SharePoint and Dataverse knowledge, please let me know in the comments and I'll, I'll work on those

23:30But SharePoint basically means you can point at a SharePoint document library, so you put the URL in there

23:36There is a little bit more fiddling around there with setting up the authentication for the users to be able to access that, and same with Dataverse

23:49That one will actually allow you to connect to Dataverse tables, and again, it gives you the option to put in descriptions of things

23:55You can use that as a knowledge source, but again, there's a fair bit more in there that is, uh, a little bit more that I want to get into in this tutorial, so let me know if you would like to see those things

24:01There are actually other knowledge sources in here as well

24:06You can bring in enterprise data connections

24:12I don't have any in this environment at the moment, but you can connect to third-party data sources if those connections have been set up, so plenty more coming there

24:19This whole knowledge feature, at the time that I'm recording this video, is all still quite new, so watch this space, but the idea of kind of giving Your co-pilot knowledge is such a core concept of the value of a co-pilot in a business scenario, rather than just this general thing of being able to have a co-pilot that's out there just randomly answering questions from the website with the large language model, which is most of what we're used to, or having something that's connected

24:45The Microsoft 365 Copilot, for instance, is connected to all of your documents and your chats and your emails, and not everyone wants that all the time for every scenario, so the idea here is that you're giving it a very contained, controlled set of knowledge, and you can see

24:58That it's only answering questions based on that knowledge that we're giving it, all right, that's part two done

25:04Let's have a look now at part three, which is how we use topics

25:11Now, topics are really important to control the flow of the information

25:17You don't necessarily want the generative AI taking full responsibility for answering all these things, so this is the concept of conversational orchestration, where you can decide how you want to direct the conversation, including which knowledge sources you want to use for which parts of the conversation, or actual sort of hardcoded answers and variables

25:37And things that you want it to be able to do, so all of this lives in this menu here called topics

25:43So we're going to do a couple of important things

25:49First, lesson one, lesson two, and lesson three are topics that are useful if you want to play around with a sort of a demo

25:54These are things that ask you about the store opening hours for places in, uh, in Redmond, in Seattle

26:00What I want you to do is just switch those three to the off position, because if you're doing anything at all that overlaps with those, you're going to get some weird answers, and you just don't want them as part of your actual co-pilot

26:12The other thing that What we've got in here is to look at the difference here between custom topics, which is what these ones are, and the system topics, which are things that are sort of just functional, helping the co-pilot to work

26:26So there are things in here which are escalation topics

26:30I'm not going to cover that here, but you can escalate to a live agent in Dynamics 365 Customer Service

26:34We have got sort of things that happen if there's an error, or there's multiple topics matched

26:39Uh, we've also got something here that can help you with getting feedback at the end of a conversation, so the one that I want to, two that I want to show you here are

26:51These two at the start, so you'll notice it starts off with, "Hello, I'm Healthy Living Assistant virtual assistant." Just so you're aware, I sometimes use AI to answer your questions

26:57From here, this is sort of more talking to me as a creator

27:04This is not particularly useful for me as something that's facing out to maybe customers who want to use this

27:11You'll notice as soon as I have clicked on this, it's actually gone into this topic for me, but let me go back and show you where that is

27:18So in the system, it's this one here that's called the conversation start topic, and you can actually change this so it says, "Hello, I'm bot." Name, so that's the first idea of a variable

27:30We're going to be looking a little bit more at variables, but that's the Healthy Living assistant

27:33So if you change the bot name, that will change

27:36You don't have to have that in there

27:39I am going to remove some of this

27:42If you provided a website during creation, so we're just, that's sort of like an instruction to the maker, we're going to remove that

27:48Um, what can I help you with today? So you can put in whatever you want here

27:56I'm just going to use my little emoji keyboard to get an apple in here, in case you don't know, Windows period, Windows full stop is a way to get the emoji

28:03Keyboard, so let's pop a little apple in there just to add a little bit of color, and then what you can do is save that

28:06Now, this is a trap

28:10The save button looks like it's grayed out

28:13Oh no, what am I going to do? Click out of that, now we're good

28:16Always get caught by that, so we're going to save that, and now when I refresh my co-pilot, I get that new message

28:27So that is in topics, system topics, conversation start topic, or click on it, and it will take you straight there

28:34So let's create an example of a custom topic

28:42So let's say I want a topic here, which is controlling the conversation if somebody says, "What do I do in an emergency?" If I come into my actual bot, if you lose the test canvas at any point, you can just put that there

28:49What do I do in an emergency? Then it will give us an answer, and it may be good or bad, but we might want a real knocking around situation here that just says call the emergency number

29:01So we've got like quite a lot of stuff in here

29:08Maybe I want to control this conversation so that every single time, all it does is tell you this piece of information here

29:12In Australia, triple zero is 911, that's our emergency number

29:17So what I'm going to do is go into the topic section here, and I'm going to say add a topic

29:21Now, I can either create a topic from Blank, or I can use Copilot to help me build a Copilot topic, so we'll give this topic a name

29:33This is going to be "emergency," and create a topic to tell the user to call to dial triple zero for, whoops, for an ambulance or police

29:56Now, if you do something more sophisticated here, if you say you can put in sort of branching logic to say if the person asks for this, do that, so you can actually play around with that description and it will do some fancy things

30:00We'll show you that

30:05Couple of things that are really important here: firstly, the trigger has come through with these phrases

30:10Now, this is the old school way of doing it

30:18Going to upgrade this in a minute

30:27This is using phrases, so the idea here is that the co-pilot agent is going to be able to figure out which topic, where it should go in the conversation based on these trigger phrases, and therefore, whatever the person types, it's able to detect the intent using these trigger phrases

30:34And so you'll see it's automatically generated a few for me

30:41I could edit those phrases if something isn't quite right, and you can see here that it's basically interpreted what I put in the instruction and put that in there, so maybe that's good enough

30:45We've got some other things that we can edit in

30:50Here, but I'm actually very happy with that one, so I'm just going to save that, and we'll see the difference here

30:54If we refresh this and say, "What should I do in an emergency?" and you'll see in the canvas that it's actually triggered the right topic, and it's given the right answer, so it's now not using those knowledge sources

31:16I've got complete control over those topics, anything here when you're designing your co-pilot and designing the way you want it to work, anything that could get you in legal trouble of any kind, anything that needs some sort of like regulatory overlay, anything that's related to safety

31:29Anything that's really important to your brand, think of any major brand, and think about how you want certain questions to be answered

31:40If you're, um, Apple, and somebody says, "What's the best phone?", you might want that answered a certain way, and not just by generative, generative answers

31:46So, anything where you really, and, and think about this, because almost any co-pilot that you're using in a real scenario, you are going to want to control that conversation at some point

31:58This is a really simple example, we can do much more sophisticated things of the topics, which I'm going to show you, but the first thing I'm going to Do is upgrade this experience to something that's currently in preview, but is actually going back to why we put all those descriptions in, and something that's actually going to perform much better than this sort of, it's strange to think of this as old school now, because 3 years ago this was, this was state-of-the-art stuff, but we're going to go here into set things and generative AI, how should your co-pilot interact with people, this, oh sorry, we're calling it classic, not old school, I shouldn't call it old school, classic, I'm going to choose this generative preview and click save, and you'll see here that

32:14You can also choose how this is

32:30This is a temperature control

32:46This is a generative AI concept of, should the content moderation be very creative or very precise? I'm actually just going to set it at the medium level there, because I think that gives it a little bit more freedom to move

32:54So this now changes the experience, so that the description of everything is what is going to be used

33:01So this is actually using now the large language model to determine the flow of the conversation, rather than that classic experience that we've had in the past, which isn't as sophisticated at doing this

33:08So let's close

33:15This down, and now if I go into my topic, you'll see I've lost that topic

33:22Let's go back into topics and have a look at my custom topic called "emergency." This is now, do you see it's got a description, and I didn't do this

33:33I didn't edit the video here at all

33:43It's taken those trigger phrases and converted it into a description, but from now on, because I've got that setting enabled, it's always going to use the description of what's going on in the knowledge sources in here to determine which thing it should trigger, so this is why you want to be careful about doing those descriptions to help out, and if you find as you test

33:54out your agent, that it's not finding the right topics, having a look at the descriptions of how you've set up those topics and knowledge sources is going to be your first point of troubleshooting

34:06Now, let's have a look at the next level of sophistication in creating topics, which is part four here, entities and variables

34:12So I am going to create another topic here based on this piece here, where we are offering guidance of where to seek support for mental health services for children and young people, and if we have a look at this content here, there's quite a lot on here, but you'll see that there are different

34:30Services here for children sort of up to 12 years old, and then for pre-teens and uh and youth

34:36So there's a uh service here called Headspace, which is designed for ages 12 to 15, and this other one here, Triple P, which is really designed for up to 12 year olds

34:49So I'm going to use this to show you a build of a conversation that can understand entities, and we'll come to this

34:55Entities, by the way, here does not mean data database entities, if you're familiar with that old language in Dataverse

35:01This is actually the concept of a a piece of known or structured data that sits in unstructured text, so if I in that piece where I Was just looking at it, says this is for 12-year-olds

35:15The age can be an entity, so that could find, the AI can actually look for a specific type of piece of information in there

35:20It's looking for the age

35:26You can also do this for things like looking for colors, or names of cities, or names of organizations

35:33I'm going to use the standard one here, but you can also create these entities for your own language to help the co-pilot identify those specific pieces of information in large blocks of unstructured text

35:39This is, this is quite clever and highly underrated

35:45This one is, uh, this is a good thing to know about

35:51So what am I going To do, I'm going to create another new topic here, and we will create this one from a description again

35:57So this one is going to be "support services for mental health" is the name of the topic, and I'm going to give this one a bit more of a description to helpfully show you what is going on here

36:10So create a topic to guide the user to different mental health support services based on the age of the person who needs support

36:19There will be options for children, young adults, and adults

36:26Let's see how we go with that

36:33So you can see there, I'm implying that there's going to be some kind of branching logic, let's see if it can

36:37Actually, do this

36:40So what this should do for me, here we go

36:43So it's actually given me a trigger, so this tool can handle queries like this

36:47So you'll see it's written that description based on what I asked it to do, and then it said, "What is the age group?" and it's given me multiple choice options

36:53So it says children, young adults, or adults, and then which one is it

37:00So if we use this topic, it will actually give you those options where it says, "Is this for..." Actually, let's, uh, let's do this, so I'll show you what this looks like

37:12So what happens is now if I say, uh, "mental health support," it should automatically trigger that

37:20Topic, because we've got that in there, and then it's going to give me these options, 'cause it's multiple choice: children, young adults, or adults

37:24Now, I didn't finish building this out

37:28If I click children, nothing will happen because, um, oh, here we go

37:32It's actually kind of filled this in with its own sort of suggestions, and, um, interesting

37:37So, the AI has actually built out the whole topic for me, but I want more control than this

37:44I actually don't want it to use these multiple choice, but you can use multiple choice in there, so I am going to edit this topic, and in fact, what I'm going to do is go into my topics

37:57I'm actually Going to delete this one, so let's go in and find it

38:05Support services for mental health, um, we're going to go in here

38:14I'm just going to grab this description because the description, the description is good, and I don't want to type it again, but I want to show you what we can do for creating a topic from scratch, so let's copy that

38:23So we're going to go back into topics and support services for mental health, delete, delete

38:28Okay, I'm just going to call it something slightly different so that I don't confuse things here, so we're going to create a topic from blank this time, and the title of this topic is going to be Mental health support services, just so it's not exactly the same, and then I'm going to use that lovely description that we had before as the things, the same concept

38:50What I want to do here though, is to ask a question and not give it multiple choice

38:56I want it to pick up on the age, because I want to show you these entities

39:01So as you're building out the topic, these are all of the different things that we can do in here

39:07You can send a message, which is just giving a response, ask a question, which is what I'm about to do, but we've also got a bunch of other sort of more advanced things here

39:13We're going to come back to the, uh, call

39:18And action, you can redirect to another topic or finish with a survey

39:27There's all sorts of things you can do in there

39:36So this is the, uh, the one I'm going to show you: ask a question, "How old is the person who needs support?" Now, instead of this multiple choice experience that we had before, what I'm going to say is, I want you to identify in here a particular type of information

39:45So this is where we've got different things, so not multiple choice or options from a list, age

39:49So age is one of the entities that we've got in there, so now it's able to find that in the conversation

39:54What I want to do now is save that user response as a

39:58Variable, and that variable means I can use that in other ways

40:02I could use that to push it out in an action to send to some other system

40:06I can use it to repeat back into the conversation, which I'll show you

40:10So what we can do here is, if I click on this one, this brings up the variable

40:15Now, you don't want to go naming things V one, because as soon as you got more than one, it's going to get confusing

40:20Let's leave VAR as the prefix and call this one H, and you'll see that you can set this variable to be working just within this topic, or I might make this one Global, which means anywhere in this agent if the conversation leads down

40:37This path and the age is determined, it can be used elsewhere as well, which could be a useful thing in this scenario

40:44Now what I'm going to do is add this conditional branching that we saw before, but I'm not doing it based on the, I'm not doing it based on multiple choice, I'm actually doing it based on that number that it's found

40:56So we're going to add a condition, and the condition is going to be based on the age, and I want that to be if it is less than or equal to 12

41:16We can rename this condition, so this can be child, even though we don't have the options that we had before, so that's what's going on there

41:23Let's add

41:31Another condition in here, go here again, add a condition, because I want, I want child, teenager, or young adult, and then just everybody else is the sort of, if neither of these things is true

41:38So if the age is greater than or equal to 13, and the age is less than or equal to 25, because remember that service for young adults, I'm just going to call this one teen and young adult, and then this one here is going to be, we'll just leave that as all other conditions, that's what's going to happen

42:04Let's just save this before we lose it

42:09So what we've got so far now is it's going to trigger the topic, how old is the person, and then it's going to

42:13Find the answer, so now what we want to do is continue this conversational thread

42:17So if it's a child, this is the part I didn't do before, we want to send a message

42:24So for someone who is now, we can bring the variable back in here, variable age, so we can repeat it back, years old, we recommend the Triple P service

42:33Now you can put more things in here

42:41I'll show you one example of something being a little bit more fancy

42:45So let's just take a look at the Headspace website

42:49I'm going to link to their services page

42:53I'm actually going to get one of the images from here, if I right click on this and you find something where you can say

42:57Copy image link, then what we can do is add a message in here, so let's click here, send a message, but instead of just having that very basic message like we had before, you have actually got the option here

43:15If you go in, you can actually use different kinds of cards, so you'll see here that you've got the ability to add images in there, videos, adaptive cards, which is a whole other skill set, if you've got that, that you can do even more sophisticated things in here

43:22I'm going to use the basic card again

43:28If you're interested in more detail on all of this topic stuff, then let me know, we can do more videos on this

43:32I'm trying to

43:36Give you the highlights all in this one video here

43:40So, uh, for teens and young adults, we recommend Headspace, and then what we can do here in the media section, come in and put "Headspace" is the title, and you'll see it's building it out as I work through it

43:58So we'll go "Mental health support for young people," and then the image URL, which is the one I just grabbed, we'll just paste that in, and then we can put the text in there as well, that should work, and you can also add buttons in there too

44:18Now, we can, for all other conditions, add a message here that is saying refer to some other service, or we could use this topic to say

44:32Let's refer to a particular website, so let's go back to the Better Health Channel here, and I'm going to find Mental Health Services

44:37Now, this will only work if the URL is only two levels deep, I should have said that at the start

44:44Your URL that you put in here will only work if it's two levels deep, but once you've got that in there, it will find everything below that

44:51So I'm actually just going to use this particular subsection of the site, so we're going to go back into the topics, and I'm going to use Advanced here and say generate answers, and then we are going to take the input, which is we need to find the system

45:13Input, so this is the activity

45:19Text, so this one here, activity.text, is basically whatever the person has typed in here

45:26This might not work so well because I've got this question, but we'll go with it, and then we can put the data sources in here and say I want to add knowledge, I want to add a public website, and I want to put that one in there

45:35Let's pretend I've done the description

45:43I think I've, I think I've explained that well enough, and now what we've got is something that is going to use generative answers within that topic to go to a specific page, so see how you can use that knowledge to actually sort of embed it in

45:52A particular place that could have been a document there as well, so lots of different things here in controlling how this works

45:57Let's give this a test run now

46:01Uh, I'm not sure how well my generative answers will work with this question in here

46:05I'm not going to worry too much about that, because mostly what I'm trying to show you here is the entities concept, so you'll notice now, remember when I had my track the conversation before, this has now got me a different option, because I switched over to that new generative AI experience earlier

46:14I've now got this wonderful thing called a conversation map, and this is now

46:23Going to show me the sort of inner workings of how the agent is deciding which topics and which knowledge sources to use

46:32Let's give this a refresh

46:38I want to show you the variable and the entity filling first, so the first thing we're going to do here is to say, uh, "mental health support," and it should trigger that topic, and in fact, it should show me on the side here what's going on with that, so it can see that it has triggered the topic, and it's asked the question, "How old is the person who needs support?" So let's try out our beautiful card

47:05Let's put in someone who is 18, and there we go, we get the Headspace with the link

47:12And that card that we showed, the other thing about the entities, is that it can identify it even if the person doesn't just give a straight answer

47:19So, in the previous one, we just put in the number 18, which looks fairly easy, but let's say the person comes in and asks for mental health support, and then it will ask, "How old is the person who needs support?" "My child is 6 years old." So, I haven't just given it the number, but it's actually able to find that piece in there, and then go ahead and do that, so it can extract, and this is useful for things like emails or order numbers or any of those things that you want to do

47:52Extract that information, and it can also just know in the conversation if it's already got that information, and you can allow it to skip this question

47:58So, if we have a look at the options for this question, you can actually say that the question can be skipped if the person has started the conversation with, "I need mental health support for my six-year-old," then it already knows that, and it can skip the question

48:07So, you can do some quite clever authoring with entities

48:15Again, if you want more on that, I can go deep on entities, but that should give you enough for a beginner tutorial to get started with that as well as

48:23With variables that you can use to put back into the conversation, and when we get to the next section on actions, you can also use variables there as well

48:36So let's take a look, part five here, at what we can do with actions, which is this next one here, create your first action

48:43This is about allowing the co-pilot to do something, to go out in the world and get some information

48:46So there's different ways that we can create actions here

48:49You have got some popular ones

48:53We're going to use the weather forecast one, because that's the one that's easy to use

48:56It's an external one that doesn't require any kind of sign up, so if you're Learning, anyone can access this one, but you can do this if you've got APIs into your existing systems that have been set up, or other APIs you can use

49:08We also have got the ability to use Power Automate flows, so I'm going to show you that concept as well

49:15And if you're a Power Platform person, and you've been working with connectors and custom connectors, those are all available to you here, so you will find yourself on familiar ground

49:22So let me show you the concept first of using this API action, so this is the MSN weather forecasting tool, which actually gives you a great deal of information about what's Happening with the weather, so back to our sort of thread about the pollen counts and things, maybe we want to check the UV index as well

49:39So this one is going to reach out and get that information

49:47What we're going to do here is switch this from user authentication to co-pilot author authentication, so that the user doesn't have to do some kind of login

49:54Again, that's there so that if you want an authenticated source for the user to log in, that will do that, but for this one that's publicly available, that just messes it up, frankly, so we want to have that in there, and then we've got different options to configure, so

50:06Again, we've got the action name, the display name, remember the description, and why that's important, but it's pretty much there with the connector, so that's okay

50:17So let's click next, and then what we've got is inputs and outputs

50:27So the inputs here, we have got the locations, so this is looking for the city or the place where the person is coming from, and we've also got units, which is the measurement system, are we talking Imperial or metric units? When we have got the outputs here, there's 39 different outputs, so there's a lot of different things in here

50:38We can come in and edit the outputs, and let's say I don't want The moon phase code, I can delete that one and move all of those things around, and you'll see we've got a lot of different things in here

50:50We've got the wind direction, the wind speed, precipitation, a whole lot of, a whole lot of things in here, so I'm just going to leave them all in, but you could actually really narrow this down to just provide you with the things that you want, and we're going to click save and next, so that is all done

51:05Get the forecast, and let's finish that

51:10Now, we can actually change this up a bit

51:14I'll show you how it works, and then we can, uh, then we can give it a second to, to come in, we can actually come in

51:19Here and change it so that it knows that it's only working in certain units, but let's just, uh, we can just say, "What is the forecast in Bendigo today?" That's another city in the state where I live

51:33We'll see if it can handle that, that's a fairly obscure reference for it

51:38Um, all right, and it's got, saying we're missing some information, so it's actually saying here, so this is good, this is a good error, I definitely did this on purpose

51:50So what this is doing is saying the idea of the plug-in here is that this action needs certain inputs, it needs the input of the location, so you can see in this canvas on the side here that it's got

51:56The location as Bendigo, it's picked that up from the, the, the thing

52:02This is a little bit like that entity thing that we were just looking at

52:08It's looking for a location, and it can identify that, but it also needs the units, so it needs to know, do we want Imperial or metric, and it can't answer that

52:14It's missing information, but it's able to figure out what it's missing and ask for it

52:21I didn't build that, right? You didn't see me build the topic that said to ask for that

52:25It's just, that's just part of the, part of how it works

52:28I work in, I work in metric here in Australia, so we're going to go with that, given that it's my Healthy

52:32Living based on information, here now it's happy, it's got both of those things, and we should get, we should get the forecast

52:38Come on, you can do it, we can do it

52:44All right, I still haven't finished with this one, but that's a good example of how it comes together

52:49Let's give this an edit, because I've obviously skipped over something important here in the topic

52:59So, so let's go into my action, let's change a couple of things

53:08So the input, the location, is going to be dynamically filled with something that comes from the user's response, that's still true, but the units, instead of being dynamically filled, this time I want To say we're going to set this as a certain value, and then I'm going to choose the value, either Imperial or Celsius

53:18Forgive me if you're watching from the United States, I'm going to go with Celsius, so now we've got that set, so it shouldn't actually ask me for that anymore

53:25But on the outputs, there's one other thing here, so we've got all these outputs, but we haven't actually given it the space to respond, so we do want it to respond, that would be helpful, thank you

53:38And so now what we've got is the option to say the user will see either the AI dynamically generates a message, or I can create a message, so if I say I want to

53:46Create a message

53:52Now I'm back in the world of the same sort of thing I did with the topic that comes in there, and you'll see that I've got all these different sort of variables that are coming from the output response, so if I just wanted to have the wind direction, I could put that in there, but let's just let the AI dynamically create the message, and adaptive card is an option there as well if you have that skill set, so we are going to click save, and hopefully this will be a bit more useful this time

54:15Let's refresh this and ask again, what is the forecast for Bendigo today, and it should just give me that

54:24We'll See it pop up on the side with the conversation map

54:32It's going to figure out that it's got a location, it knows that I want the answer in Celsius, and it should give me all of that information

54:41Now, what we can do is use Power Automate to call something

54:48So remember before, when we had our lovely Melbourne pollen situation, and it couldn't read this information here? So what I'm going to do is create a spreadsheet

54:55So I have actually got something that's very basic

55:01This is Excel online, so this needs to be online, but you can use this with any kind of, um, data source here, but this needs to be, I want this to be an

55:08Unauthenticated data source, so this is a really simple table

55:15It's just got these values in it, so you can see I've just copied the values from the website

55:18They're all low

55:22I've just changed one of them to high because I want to prove that it's actually returning a different response and not just saying low every time

55:26The trick here, if you want to be able to use Excel online, this needs to be actually saved into your OneDrive, and it needs to be formatted as a table

55:37So if you don't know how to do that, you grab that, we go to home and format as table, and you just format it as a table and give that table a name, so you need to

55:45Make sure that you've got those things right, and then we can connect this up using Power Automate, because it can't read this on the website

55:52So if this was my real scenario, it would be relatively easy

55:58Maybe this could be connected to the data source where it's coming from, if I could authenticate that, or I could just put it on a spreadsheet and have some automation that just updates this spreadsheet every morning, and my Copilot is reading from this spreadsheet

56:10So let's get rid of that

56:16We'll go back into here, and I'm going to go back into my actions now and create another action

56:21Now this could also be Something again, you could choose whether this is just an action that sits there, or whether you have it in a topic where the topic says caller flow, but we've looked at topics

56:26I'm just going to show it to you from here

56:32Uh, this is actually one that I've prepared earlier, the get pollen forecast, so we're going to build it again because it's more interesting if you see it being built, so we're going to say let's use a flow, so I can use that existing flow if I've already got that in there, but I'm going to create a new flow that's going to open a tab here for Power Automate, and you'll see what we've got here is this input and

56:52Output, we want to run a flow from Copilot, take some kind of input, and then give some kind of output as a response back to Copilot

56:59So, run a flow from Copilot, has inputs in it, so we are expecting a text input, we are expecting the person to give us the name of the region

57:12So, we remember my spreadsheet here has a region and a pollen forecast in it, so that's as simple as that, that's all we need to do

57:19And then what we're going to do is, in between these two actions, is run a connector to Excel Online to get that piece of information

57:31So, the connector we need here, and the action we need here, is to get a row, so what we're going to Do is choose the location, so where is this coming from? I've just got it saved in OneDrive for Business, but you could have that in a SharePoint, uh, library as well

57:43The document library then is just going to be OneDrive, the file, so we look up the particular file

57:48So we've got a file folder here

57:52I've actually just got this saved on my OneDrive in a Copilot Studio demo, uh, folder, and then we choose the table

57:56So remember I said you had to format it as a table

58:02You'll get stuck at this point if you haven't formatted it as a table

58:08I've got my table in there called pollen forecast, and then which column am I looking for? So the user Is providing us with a region, so I'm looking for, so now it knows the columns in that table

58:14So we're going to choose the region column, and then the value this time

58:20Now, this value is what is the person putting in

58:26I want to get the value that they entered in this previous step, so we're going to use this little icon here to search for things that are going on in the flow

58:33So I want something from the runoff flow from co-pilot step

58:38I want that input that they've put in there, which is the region, that's it

58:45So we've now run the flow, asked them for the region, retrieved that, and now we need to output the other column, which is the Matching pollen forecast value, so we're going to respond to Copilot

58:57We're going to add an output, which is also text, that is going to be the pollen forecast, and the value that we respond with, we'll come in here and look that up, and that is coming from that "get a row" step, and it's this one here, pollen forecast

59:10You can see that's coming from Excel

59:17The other thing you want to do is give this a name

59:24I'm going to, because I already had the get pollen forecast in there, and like you normally wouldn't be doing this, but I'm just going to call it something different, "look up pollen forecast," and we're good

59:29Let's save the

59:35Draft of that, and then we're good

59:40Okay, let's publish it, give it a second, and then there's one more thing I want you to check

59:52So if we just go back here, what happens is you've got a connection reference, so this is connected to the Excel online business

59:56I've got a green tick there

01:00:00If it's the first time you're using the connector, just make sure that that's all properly connected and authenticated, otherwise you're going to get errors back over in your co-pilot

01:00:03So now, if we go back into here, it will say refresh, and we should see that third flow

01:00:11There we go, so there's the one I had before, but this is the one I've just

01:00:17Created, so now I can choose that, and this is now going to behave in a way that is very similar to when we set up that weather forecast, but I've done my own connection here

01:00:22So the input is the region, how will the co-pilot find it just from the user response, and now the output, we might want to edit this output, same kind of concept we had before, so we could put in here, how do we want this to respond

01:00:41So yes, we do want to respond to the user after running the action, and we've got the same options before, so AI can dynamically generate a message, how about we create the message this time, so we'll just say the forecast is, and then I

01:00:54Can just grab one of those variables, so we should be able to find the variable called pollen, the output there, pollen forecast, so then we should be good to go

01:01:06We'll just go next, so there's the action, we've got the input, the output, and finish

01:01:12Now, let's see what happens

01:01:20We'll give it a second to connect everything up, and hopefully, we won't have any errors in here

01:01:27If you are getting value from this, if you've come with me this far, give me a comment if you've come this far, not everyone will come this far, and please remember to like and share, and add comments on here to help other people find this

01:01:33I'm putting a lot

01:01:39Of explanation in here because I think people uh find that helpful, so please uh please add a comment if you've come this, if you've come this far with me, and uh we'll see how we go

01:01:45All right, I think that, oh, there we go

01:01:51All right, I thought I was going to have to talk for longer to do this in real time, so now we have got that one in there, so now I should be able to say, "What is the pollen forecast for East Gband?" and hopefully this time, because I've got that in my data source, ah, now this is where we need to go through another, another kind of connection here

01:02:12The first time you as a user use this, you do need to do this sort of

01:02:19Connection request, so we'll just come in here

01:02:25This is just one off, so we'll just say, let's connect that one, and I'm not sure if this is going to be a permanent thing or not, but it is something that we have to do at the moment, so this is a one off though

01:02:32Once that's done, you shouldn't have to do it again, just like building the suspense of this

01:02:45All right, come on, loading, there we go, connected, green tick, happy times, close that one down, and let's retry

01:02:52I don't like the retry, let's, oh, there we go, the forecast is low

01:02:57All right, but let, I mean that's working, but let's see it properly working here, so we go from the start here, what

01:03:03Is the pollen forecast for East Gibsland, and it should just come back and say the forecast is low, so there you go

01:03:07So that's actions

01:03:12There's many other different types of actions in there

01:03:17You can use all sorts of connectors

01:03:20There are some advanced actions with AI Builder prompts

01:03:23Again, let me know in the comments if you'd like additional videos on all of those things

01:03:30I'm going to see what people are most interested in

01:03:37Let's do the final thing now of sort of putting this out into the world, so we can publish this, but I'm just going to do one more thing first, part six here, publishing and settings, so my co-pilot's

01:03:41Name is fine

01:03:44I'm going to change the icon

01:03:48I've actually used Copilot to generate a little Apple icon here, because why not? Uh, and it needs to be in PNG format, so that's it, it's happy, that's good, we can just save that

01:03:52So that will, that will come up, something on a white background, it kind of comes in a circle format, so don't have something that goes right out to the edges, because it will get a little bit, uh, a little bit chopped off in there

01:04:05The other thing that you can do in here, and if you want to publish this, I'm going to publish this now to a demo website, and this is something that you'll need to do if you want to do this for yourself for practice is go into security, and you'll see that you've got these options here for the authentication, so the standard authentication setup, and this is done so that it's easier for you to create Copilot agents that are connected to SharePoint, but in this case, we're not doing that

01:04:29I want this to be publicly available with no authentication, so you need to go in here and say no authentication, and then you'll be all right

01:04:35So if you get stuck with putting it out there and it won't let you do it, please make sure you've checked that setting in there, but this is also where you're going to be able To set those configuration options in the real world for what you want it to do, and from there, we are going to close those settings and hit the publish button, and we will publish this Copilot

01:04:52So you've noticed I don't have to publish the whole way, we just do that when we get to the end

01:04:58While we're waiting for that to publish, let's just have a look at another one here with some of the channels that are available, so you can publish Copilots for internal or for external use

01:05:09So you can publish them into Microsoft Teams, you'll see here, you can put them into different customer service systems, third party customer

01:05:17Service systems, or you can be publishing them out to your customers for external use, obviously on your website

01:05:24You can have them working through Twilio, Facebook, mobile app

01:05:28There's a, there's a whole lot of things here, so I'm going to work with the demo website here

01:05:32We'll just, uh, close this one down and go back to the one that I had before

01:05:36The options are all available to me here because I've switched that authentication to no authentication

01:05:42The other one, not all of them were available because I hadn't, I was setting it to, uh, authenticated, so it's only going to let me publish in Teams, so I want to use the demo

01:05:49Website, and this is the easiest one that you can use if you're just learning because it's just there

01:05:55So then I can just sort of copy this, uh, this website link, and you can put different welcome messages in here, so try out our Healthy Living agent, and again with my kind of little apple in here

01:06:15And then conversation starters, you could put suggestions in here

01:06:22I'm not going to worry too much about that, so we will save all of that

01:06:28Let's grab a nice fresh browser tab here and put that one in, and there we go, we can try this out live

01:06:34So I can come in here and say something like, what happens if I'm diagnosed with diabetes? It's going to go out, so over the course of this tutorial, now we've got a co-pilot here that is connected to two websites, an Excel spreadsheet online, a document, it's got a topic in there, what do I do in an emergency? I don't know if we ever tested that one

01:06:43Uh, and so all of those things are in play

01:06:52Other things that you can do in here, part seven, don't run away, there's more here

01:07:02You can actually come in here and create based on a template, so there are a bunch of templates in here that you can come in and play around with if you want to see what those things do

01:07:13Some of these are coming soon, we have got voice capabilities

01:07:19Coming soon, we've got the ability also coming soon to create agents that can start to do things on their own more independently than what I've just shown you

01:07:25Make sure you're subscribed, 'cause I'll definitely be doing content on that when it comes through

01:07:31We've actually also got all of these co-pilots in here, so you've got a bunch of things that I've mucked around with, but you've also got within here your co-pilot for Microsoft 365

01:07:36So if you've got that paid subscription, this is the one where it's $30 per user per month, and it's inside all of those Microsoft applications, including inside Teams, you can use

01:07:49The skills you've learned here to extend that to connect to the other systems, and the brand new announcement of this piece here, which is creating a co-pilot agent inside that you might

01:08:03Now understand that this is familiar, we've got this field service agent that's sitting inside Teams, and look at this co-pilot Studio describe, that's right where we started from in this tutorial

01:08:15I've taking you right back to the start, this is why I think this is one of the most important skills you can learn right now

01:08:20Thank you, if you've come all the way to the end with me, I appreciate you

01:08:22Thank you so much for watching

01:08:25I've

01:08:27Got a playlist here with some more videos on Copilot Studio

01:08:29I'm going to keep adding to it based on the comments and based on the feedback of what people would like to learn

01:08:32Thank you so much.