Copilot Studio: Complete Tutorial for Beginners

If you are a beginner wanting to learn how to use co-pilot Studio you have come to the right place welcome to my complete tutorial for beginners I'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 co-pilot studio so what we're going to be covering in this

video is in the first section how to get started Ed with creating a co-pilot by describing it or by using one of the built-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 we'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 actions 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 we'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 there's a lot in here if 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 I am actually raing co-pilot 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 co-pilot agent to extend co-pilot for Microsoft 365 or any of the other first party co-pilots with all of the new things that have been

announced coming through in the Wave 2 of the Microsoft 365 co-pilot this is getting even closer to that experience and more relevant and the road map of where co-pilot agents are going which is more that I want to cover here is is a very bright future of interesting things that we're going to be able to do i' 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 everything happens here you just want to go

to your browser and go to this website co-pilot studio. microsoft.com and sign up for a free trial now everything I'm showing you in this tutorial I'm working on a free trial so you've got no problem you don't need to pay for anything until you're ready to actually publish your copilot agent and put it out into the world you will need a work or school account to sign up for that free trial now this is the main landing page but when you first first come in and sign up through that trial process it usually lands you

here and this is co-pilot helping you build a co-pilot there's a fair bit of this kind of experience going on in here of co-pilots helping you build co-pilots But that's good so it says hi I'm here to help you build a co-pilot in a few sentences how will your co-pilot 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 this is a series of sort of Articles and pieces of advice on on health

and 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 so we give it a a definition here uh so how will you co your copilot assist your users it will help them find information about healthy living now you can put whatever you want in there I will show you where all these pieces go this 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 can 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 so do I have any instructions For how your co-pilot should assist for example your a specific spefic 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 you 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 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 that's called instructions custom instructions where should the co-pilot find important information now at this stage this is starting to add knowledge the 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 if you want to create a co-pilot that doesn't use a website you can just skip this step you can just say no website so I'm going to put that whole website link in there and it's going to think about that one a little bit more are there any topics or tasks co-pilot 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 so what I've done now is

given this some guidelines I've given my co-pilot guidelines about how I want to set up the co-pilot I've given it a purpose I'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 it confirms all of those things so that is good we can just click create when we're done and it will go ahead and set that up so this is going to put all of those things inside the co-pilot it's going to actually deploy a co-pilot for you just on

the language I'm going to go back and forth a little bit here between calling it a co-pilot and a co-pilot agent I've always only called it a co-pilot just recently the language from Microsoft is starting to call these co-pilot agents so if I flick back and forth essentially the same thing copilot agent is just the more modern term and I haven't quite caught up with that yet all right so we have a co-pilot here ready to go let's have a look At where 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 it'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 jug and so 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 every time it produces a response it uses that as a a guideline we 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 I don't know if you can hear the birds chirping outside I where I live it's getting towards the end of the day and the the they're rainbow lates in case you want to look up what that bird noise is outside and if you can't hear it then I can hear it and I'm mildly distracted by it we 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 that 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 a test run here so 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 and managing sleep and fatigue so let's give this a go we'll come in and say

how can I manage fatigue and this is as I said it's only created it 2 minutes ago it usually takes a couple of minutes to 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 now why is there a great big long video Beyond this you might you might ask realistically this is a really good sort of party trick of being able to go look at this I've pointed to a website here it is it's all working

there 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 it's a good exercise to go through for learning but there are many other things you 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 I 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 there is a skill in this in the thought process of the actual design I'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 so 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 so if I click through that's going to open that Source in there now I am just working here in a test canvas the other thing you can do is if I click on here we can open this track between topics I'm going to show you another experience as we go a little bit further into it if 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 I know how to spell it can actually so this is the the concept here is we've got this large language model it can actually sort of interpret and understand things so there you go I've asked the prompt in a different way and it's actually come back so it's capable of doing that what we've got in the middle here is the canvas that's actually showing the flow of that conversation and where it went so when we come back in a little

bit to the idea of topics this is actually what's called a fullback 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 this cop it 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 so let's take a look at this concept of knowledge and what this is doing

in here we will come at the end to actually sort of publishing this all the way Through here I'm going to work just in this test canvas this 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 exter Journal co-pilot on our website or various other channels or you can use this as an extension to your Microsoft 365 co-pilot or co-pilot for sales or co-pilot for service or Finance or

all of those things as well so wherever it's going in the end it's going to go out into the world this experience is uh is just how we are testing it so 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 so at the moment what we've got here is a website and and it's ready to go we can come in here and open the details let's edit this one and have a look now you'll see the knowledge description here is actually both

mandatory and blank which is interesting when 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 CU this is Actually really really important here what is going on here is that this is a description of what this knowledge does so 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 this description is like a it's almost like a prompt that helps the AI understand what this is actually doing so it's called better health so let's as the outset sort of Tidy this up a bit there's the web page link but but let's call this The Better Health Website because we can have a name that's not the website and 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 you will see later why this becomes important let's just save that because we're going to start using AI to help the co-pilot it determine 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 before 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 here 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 one 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 so 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 need to be aware of this thankfully 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 so this is just sort of showing you what you can do with adding more things in here again 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 we'll give it a second to spin up while I explain what's going on here so as a human I can look at this beautiful table here and determine exactly what's going on this is designed for very good

human readability I can come in here and straight away say yep East gips land is low and see exactly what what's going on if 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 Gibs land which is a region in the state sort of near where I live but if we go through here this is actually the table so 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 this 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 Bots or Bots I'm just a little all over the place with keeping up with the language bot equals chatbot equals actually no an agent is more sophisticated than a bot forgive me I'm going to try not to call it a bot anymore what is the pollen forecast for East Gibs land 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 okay so that hasn't helped me particularly what 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 alert 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 so this part is where you get this message AI generated content may be inre correct 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 so you will have use cases where that is highly useful and desirable and you'll have used 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 now 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 um 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 I 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 of table format if

I go into the sort of About Us section this is a nice text based thing so let's come back in here and say um tell me about the Melbourne pollen service so this is something it can do because it can read this off the website so it's gone into find that data source there we go gives you the information gives you the link let's try the forecast again what is the pollen forecast or east GS land I'm always worried doing doing live demos there we go can't find it we're going to have a solution for

this later when we get to the section on actions I'm going to show you how you can make this kind of data accessible 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 Ja 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 it'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 the 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 a limit so let's go back now and have a look at some of the other knowledge options so the next thing I'm going to show you is that you can upload a document

into here so I have got a document in here somewhere let's take a look at my sample document here I have got a PDF that I've downloaded in keeping with our Healthy Living this is the Australian dietry guidelines so 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 so let's come into here and now I'm going to add knowledge and I'm going to drag and drop a file here

so 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 so this one I think I've got this one here already information about I'll just paste it in and describe it to you so I've put the description in there information about the Australian dietry guidelines including standard serves of all the major food groups and recommended daily intake so I've given that quite a bit of information now what this is doing is actually uploading the document into dataverse underneath Don't Panic this

is it's not actually creating a datab base or anything in there but it's using that dataverse indexing to process that document so while it's sitting here in progress It's processing that document and using that indexing to be able to understand it obviously the larger the document you've got in there the longer that's going to take there's a range of different file formats you can use in there you 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 you were able

to put Excel files in there I wouldn't recommend that it 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 this 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 this is a good way to go away and have a beverage of choice or or do something else now 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 dietry guidelines serving slice is approximately 4 40 G so if I click on that you'll see it comes up here now 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 the co-pilot knowledge and it will actually show the section so 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 let's take a look at those other knowledge sources now so if we come in here and say let's add knowledge there's SharePoint 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 data ver knowledge please let me know in the comments and I'll I'll work on those but SharePoint basically means you can point at a SharePoint document

Library so you put the URL in there there 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 datae that one will actually allow you to connect to data verse tables and again it gives you the option to put in descriptions of things and you can use that as a knowledge Source but again there's a 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 there are actually other knowledge sources in here as well you can bring in Enterprise data connections I 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 this 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 the 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 that it's only answering questions based on that knowledge

that we're giving it all right that's part two done let's have a look now at part three which is how we used topics now topics are really important to control the flow of the information you don't necessarily want the generative AI taking full responsibility for answering all these things so this is the concept of conver ational 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 and things that you want it to

be able to do so all of this lives in this menu here called topics so we're going to do a couple of important things first lesson one lesson two and lesson three are topics that are useful if you want to play around with a sort of a demo these are things that ask you about the store opening hours for places in uh in Redmond in Seattle what 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 the other thing that 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 is things that are sort of just functional helping the co-pilot to work so there are things in here which are escalation topics I'm not going to cover that here but you can escalate to a live agent in Dynamic 365 customer service we have got sort of things that happen if there's an error or there's

multiple topics matched uh 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 these 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 from here this is sort of more talking to me as a Creator this is not particularly useful for me as something that's facing out to maybe customers who want to

use this you'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 so 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 we're going to be looking a little bit more at variables but that's the Healthy Living assistant so if you change the bot name that will change you don't have to have that in

there I am going to remove some of this if 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 um what can I help you with today so you can put in whatever you want here I'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 keyboard 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 now this is a trap the save button looks like it's grayed out oh no what am I going to do click out of that now we're good always get caught by that so we're going to save that and now when I refresh my co-pilot I get that new message M so that is in topics system Topics conversation start topic or click on it and it will take you straight there so let's create an example of a custom topic so let's say I want a topic here which is

controlling the conversation if somebody says what do I do in an emergency so if I come into my actual bot if you lose the test canvas at any point you can just put that there what 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 nocking around situation here that just says call the emergency number so we've got like quite a lot of stuff in here maybe I want to control this conversation so that every single time all it does

is tell you this piece of information here in Australia tri0 is 911 that's our emergency number so what I'm going to do is go into the topic section here and I'm going to say add a topic now I can either create a topic from blank or I can use co-pilot to help me build a co-pilot topic so we'll give this topic a name this is going to be emergency and create a topic to tell the user to call to dial triple 0 for whoops for an ambulance or police now 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 we'll show you that couple of things that are really important here firstly the trigger has come through with these phrases now this is the old school way of doing it we're going to upgrade this in a minute this 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 and so you'll see it's automatically generated a few for me I 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 we've got some other things that we can edit in here but I'm actually very happy with that one so I'm just going to save that

and we'll see the difference here if 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 I'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 be get you in legal trouble of any kind anything that needs some sort of like regulatory overlay anything that's related to safety anything that's really important to

your brand think of any major brand and think about how you want certain questions to be answered if 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 so anything where you really and and think about this cuz almost any co-pilot that you're using in a real scenario you are going to want to control that conversation at some point this 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 you can also choose how this is this is a temperature control this 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 the medium level there because I think that gives it a little bit more freedom to move so this now changes the experience so that the description of everything is what is going to be used so this is actually using now the 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 so let's close this down and now if I go into my topic you'll See I've lost that topic let'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 I didn't edit the video here at all it'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 out 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 now let's have a look at the next level of sophistication in creating topics which is

part four here entities and variables so 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 Services here for children sort of up to 12 years old and then for pre-teens and uh and youth so 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 so I'm going to use this to show you a build of a conversation that can understand entities and we'll come to this entities by the way here does not mean data database entities if you're familiar with that old language in dataverse this 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 12y olds the age can be an

entity so that could find the AI can actually look for a specific type of piece of information in there it's looking for the age you can also do this for things like looking for colors or names of cities or names of organizations I'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 This is this is quite clever and highly underrated this one is uh this is a good thing to know about

so what am I going to do I'm going to create a another new topic here and we will create this one from a description again so 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 so create a topic to guide the user to different mental health support services based on the age of the person who needs support there will be options for children young adults and adults

let's see how we go with that so you can see there I'm implying that there's going to be some kind of branching logic let's see if it can actually do this so what this should do for me here we go so it's actually given me a trigger so this tool can handle queries like this so 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 so it says children young adults or adults and then which one

is it so if we use this topic it will actually give you those options where it Says is this for actually let's uh let's let's do this so I'll show you what this looks like so what happens is now if I say uh mental health support it should automatically trigger that topic because we've got that in there and then it's going to give me these options cuz it's multiple choice children young adults or adults now I didn't finish building this out if I click children nothing will happen because um oh here we go it's actually

kind of filled this in with its own sort of suggestions and um interesting so it's the AI has actually built out the whole topic for me but I want more control than this I 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 I'm actually going to delete this one so let's go in and find it support services for mental health um we're going to go in here I'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 So we're going to go back into topics and support services for mental health delete delete okay I'm just going to call it something slightly different so that I don't uh 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 what I want to do here though is to ask a question and not give it multiple choice I want it to pick up on the age because I want to show you these entities so as you're building out the topic these are all of the different things that we can do in here you 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 we're going to come back to the uh call and action you can redirect to another topic or finish with a survey there's all sorts of things you can do in there so 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

so this is where we've got different things so not multiple choice or options from a list age so age is one of the entities that we've got in there so now it's able to find that in the conversation what I want to do now is save that user response as a variable and that variable means I can use that in other ways I could use that to push it out in an action to send to some other system I can use it to repeat back into the conversation which I'll show you so what we can

do here is if I click on this one this brings up the variable now 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 let'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 this path and the age is determined it can be used elsewhere as well which could be

a useful thing in this scenario now 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 on multiple choice I'm actually doing it based on the that number that it's found so 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 we 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 let's add another condition in here go here again add a condition cuz 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 so if the age is greater than or equal to 13 and the age is less than or equal to 25 cuz 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 let's just save this before we lose it so 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 find the answer so now what we want to do is continue this conversational thread So if it's a child this is the part I didn't do before we want to send a message so 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 now you can put more things in here I'll show you one example of something being a little bit more fancy so let's just take a look at the headspace website I'm going to link to their services page I'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 copy 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 if 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 I'm going to use the basic card again if you're interested in more detail on all of this topic stuff then let me know we can do more videos on this I'm trying to give You the the highlights all

in this one video here so uh for teens and young adults we recommend head space 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 so 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 now 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 let'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 now this will only work if the URL is only two levels deep I should have said that at the start your 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 that so 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 anwers and then we are going to take the input which is we need to find the system input so this is the activity. text so this one here activity. text is basically whatever the person has typed in here this 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 let's pretend I've done the description I 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 a particular place that could have been a document there as well so lots of different things here in controlling how this works let's give this a test

run now uh I'm not sure how well my generative answers will work with this question in here I'm not going to worry too much about that cuz 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 I've now got this wonderful thing called a conversation map and this is now going to show me the sort of inner workings of how the agent is

deciding which topics and which knowledge sources to use let's give this a refresh I 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 let's put in someone who is

18 and there we go we get the heads space with the link and 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 so 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 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 it can extract 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 so 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 so you can do some quite clever authoring with entities again 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 with 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 so 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 this is about allowing the co-pilot to do something to go out in the world and get some information so there's different ways that we can create actions here you have got some popular ones we're going to use the weather forecast one cuz that's the one that's easy to use it'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 that we have also got the ability to use power automate flows so I'm going to show you that concept as well and 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 so 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 so this one is going to reach out and get That information what 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 again 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 again 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 so let's click next and then what we've got is inputs and outputs so 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 we 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 we'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 get the forecast and let's finish that now we can actually change this up a bit I'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 here 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 we'll

see if it can handle that's fairly obscure reference for it um all right and it's got saying we've missing some information so it's actually saying here so this is good this is a good error I definitely did this on purpose so 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 the location as Bendigo It's picked that up from the the the thing this is a little

bit like that entity thing that we were just looking at it'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 it's missing information but it's able to figure out what it's missing and ask for it I didn't build that right you didn't see me build the topic that said to ask for that it's just that's just part of the part of how it works I work in I work in metric here in Australia so

we're going to go with that given that it's my Healthy Living based on information here now it's happy it's got both of those things and we should get we should get the forecast come on you can do it we can do it all right I still haven't finished with this one but that's a good example of how it comes together let's give this an edit because I've obviously skipped over something important here in the topic so so let's go into my action let's change a couple of things so 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 forgive me if you're watching from the United States I'm going to go with Celsius so now we've got that set so shouldn't actually ask me for that anymore but 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 and 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 create a message now I'm back in the world of the same sort of thing I did with the 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 let's refresh this and ask again what is the forecast for Bendigo today and it should just give me that we'll see it pop up on the side with the conversation map it'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 now what we can do is use power automate to call something so 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 so I have actually got something that's very basic this 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 unauthenticated data

source so this is a really simple table it's just got these values in it so you can see I've just copied the values from the website they're all low I'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 the tricks here if you want to be able to use Excel online this needs to be actually saved into your one drive and it Needs to be formatted as a table so 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 make 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 website so if this was my real scenario it would be relatively easy maybe 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

any every morning and my co-pilot is reading from this spreadsheet so let's get rid of that we'll go back into here and I'm going to go back into my actions now and create another action now 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 I'm just going to show it to you from here uh 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 output we want to run a flow from co-pilot take some kind of input and then give some kind of output as a response back to co-pilot so Runner flow from co-pilot has inputs

in it so we are expecting a text input we are expecting the person to give us the name of the region so we remember my spreadsheet here has a region and a pollen forecast in it so that's a simple as that that's all we need to do and 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 so the connector we need here and the action we need here is the 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 one drive for business but you could have that in a SharePoint uh Library as well the document library then is just going to be one drive the file so we look up the particular file so we've got a file Folder here I've actually just got this saved on my one drive in a copilot Studio demo uh folder and then we choose the table so remember I said you had to format it as a table you'll get stuck at this point if you

haven't formatted it as a table I've got my table in there called pollen forecast and then which colum 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 so we're going to choose the region column and then the value this time now this value is what is the person putting in I 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 so I want something from the runoff flow from co-pilot step I want that input that they've put in there which is the region that's it so 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 going to respond to co-pilot we'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 you can see that's coming from Excel the other thing you want to do is give this a name I'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 look up pollen forecast and we're good let's save the draft of that and then we're good okay let's publish it give it a second and then there's one more thing I want you to

check so if we just go back here what happen s is you've got a connection reference so this is connected to the Excel online business I've got a green tick there if 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 so now if we go back into here it will say refresh and we should see that third flow there we go so there's the one I had before but this is the One I've just created 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 forec but I've done my own connection here so 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 so 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 for cast is and then I can 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 we'll just go next so there's the action we've got the input the output and finish now let's see what happens we'll give it a second to uh connect everything up and hopefully we won't have any err ER is

in here if 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 uh and add comments on here to help other people find this I'm putting a lot of 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 all right I think that oh there we go

all 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 the first time you as a user use this you do need to do this sort of connection request so we'll just come in here this 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 once that's done you shouldn't have to do it again just like building the suspense of This all right come on loading there we go connected green tick Happy Times close that one down and let's retry I don't like the retry let's oh there we go the forecast is low all right but let I

mean that's working but let's see it properly working here so we go from the start here what is the pollen forecast for East gibsland and it should just come back and say the forecast is low so there you go so that's actions there's many other different types of actions in there you can use all sorts of connectors there are some Advanced actions with AI Builder prompts again let me know in the comments if you'd like additional videos on all of those things I'm going to see what people are most interested in let'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 name is fine I'm going to change the icon I've actually used co-pilot to generate a little Apple icon here because why not uh and it needs to be in PNG format so that's it's happy that's good we can just save that so 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 cuz it will get a little bit uh a little bit chopped off in there the 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 it's easier for

you to create co-pilot agents that are connected to SharePoint but in this case we're not doing that I 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 so 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 public Button and we will publish this co-pilot so you've noticed I don't have to publish the whole way we just do that when we get to the end while 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 co-pilots for internal or for external use so you can publish them into Microsoft teams you'll see here you can put them into different customer service systems third party customer service systems or

you can be publishing them out to your customers for external use obviously on your website you can have them working through twilio Facebook mobile app there's a there's a whole lot of things here so I'm going to work with the demo website here we'll just uh close this one down and go back to the one that I had before the options are all available to me here because I've switched that authentication to no authentication the 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 website and this is the easiest one that you can use if You're just learning because it's just there so 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 and then conversation starters you could put suggestions in here I'm not going to worry too much about that so we will save all of that let's grab

a nice fresh browser tab here and put that one in and there we go we can try this out live so I can come in here and say something like what happens if I'm diagnosed with diabetes and 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 imerge urgency I don't know if we ever tested that one uh and so all of those things are

in play other things that you can do in here part seven Don't Run Away there's more here you 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 some of these are Coming soon we have got voice capabilities coming 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 make sure you're

subscribed CU I'll definitely be doing content on that when it comes through we'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 so 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 the 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 Now 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 I'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 thank you If you've come all the way to the end with me I appreciate you thank you so much for watching I've got a playlist here with some more videos

on copilot Studio I'm going to keep adding to it based on the comments and based on the feedback of what people would like to learn thank you so much