Build and Use Agents in Microsoft 365 Copilot: Complete Tutorial (2026)

Hello and welcome to this complete guide

on co-pilot agents. I'm going to take

you from the beginning through to

everything you need to know to be

somebody who can use, build, and work

with agents in Microsoft 365 Copilot at

work. I'm going to take you through both

the free version of Copilot Chat and the

paid version of Microsoft 365 Copilot so

that you understand what you can do

depending on which license you have. and

I'm going to take you through the skills

you need, the concepts you need, and

some of the out- of- the-box experiences

and how to get the most value out of

them. This is not a quick guide. This is

designed to be comprehensive. I'm going

to go at a pace that you should be able

to follow along. Please pop any

questions you have into the comments

below on this video, and I'll do my best

to answer them. And if you get value out

of this content, don't forget to like

and share it with others to help it

reach a bigger audience. And if you're

not already subscribed, you know what to

do. Let's get into it. Before we get

started, a couple of things. This is

working always with a work license. So,

we're talking here about having access

to the free version of Copilot Chat that

comes with your Microsoft 365 license at

work or the same situation with an

additional paid add-on license called

Microsoft 365 Copilot. None of what I'm

showing you here is possible with a

personal or consumer license. This is

very focused on agents at work. This is

also dependent on what your IT admin has

allowed or blocked in your organization.

So, if you can't see what I can see,

you're going to need to talk to your IT

admin about where they're at with

allowing agents. Maybe get them to watch

this video if they need to catch up on

what's possible here. So, I'm going to

start with a user who does not have a

paid license. This is somebody who's

logging in at work with their Microsoft

365 license and they have access to the

free co-pilot chat. Now, if you've got a

paid license, stay with me here because

we're actually going to build the

knowledge. You have access to all of

these things and more, but we're going

to start from the ground up with the

concepts here. So, don't jump away cuz

you'll miss some you'll miss some really

good stuff here. Now, the way that you

can tell whether or not you've got a

paid license, couple of things. If you

don't have the paid license, when you

log in, there will be nothing at the top

here. You just straight into Copilot

Chat. And you will see at the bottom

here that your license says Copilot

Chat. And you've got a little

information icon that explains that with

the free version of Copilot Chat,

everything you're doing here with

building agents is based on what you

upload or what you provide to it or web

sources. You can't automatically connect

to your teams, emails, all your

documents in SharePoint. That's the

stuff you have to pay for. So, if you're

seeing nothing at the top here, Copilot

Chat, and honestly, an upgrade button

because Microsoft wants everyone to be

on that paid version, you are in the

free version of Copilot Chat. If you see

some different things, if you see this

work and web at the top and M365 Copilot

at the bottom, that means you have the

paid add-on license. somebody in your

organization has paid extra for you to

have that and you will actually have

much more sophisticated experiences. So

I'm going to go through here the concept

of what an agent even is and this

applies equally to no matter which type

of license you've got. So I'm in my main

co-pilot chat experience here and let's

say I need some help with my

productivity. Not confessing at all that

this is a real use case but you can read

between the lines as as you like. So,

what I'm going to do here is just put a

prompt into my main chat experience. I'm

just going to copy and paste so you

don't have to watch me type here. I have

a big piece of work to get done, but

it's difficult to get started. So, I

keep ticking off smaller, easier tasks

instead. Anybody else? So, what this is

going to do, I'm just putting this into

the chat. This is just going to use AI

to give me a response. This is a secure

chat for my work scenario, but it's

actually not referring to anything. This

is just a general AI kind of chat coming

back at me. Uh Amy is the loggedin demo

user I have here who doesn't have a

license. That's why it's calling me Amy.

I'm obviously Lisa, not Amy. So, it's

actually just sort of coming back with,

you know, the cute little emojis and

it's giving me some general advice and

that's fine. Now, the difference between

just putting in a prompt like this and

building an agent is where you want

something that is very specifically

designed to handle certain types of

questions or certain types of tasks in

the way that you want. So, for instance,

this is just advice of a general nature

if you like that you're getting from the

AI response here. But let's say that I

want it to respond to me in a certain

way every time. I want it to have a

certain tone of voice. I want it to give

me some, you know, inspirational

encouragement, those kinds of things.

So, I could absolutely put all of that

into a prompt, but my prompt might start

getting quite long. And I want it to be

much more focused in the way that it

answers. So, the sort of the the path

here going from when is it a prompt to

when is an agent is to start to think

about is this something I want to do on

a repeatable basis, something I want to

come back to over and over again. That's

one reason you might want to create an

agent. Can I describe a specific task

and the way I want it done or a specific

kind of question and the way that I want

it answered and I want to do that in a

lot of detail and I don't have to type

that in every time and then do I want it

to refer to certain types of knowledge

in order to answer that question. So

these are all the types of things you

should be looking at. So let's have a

comparison here with what happens. This

is just in my chat experience in the

agent experience. So I'm going to come

across here and say new agent. Now there

are two ways that you can build an

agent. I'm going to show you the easier

way first and then the way that gives

you a little bit more control as we

build up to it. You'll see here there

are two tabs, describe and configure.

Depending on how your thing is set up,

I'm not sure where it will land.

Wherever you are, I want you to click on

that describe tab if it's not already

the one that's open. If you were already

there, then that's one less click that

you have to do. So, what I'm going to do

is actually give it a description of

what I want it to do. So, this is going

to be a productivity agent here. Help me

with my everyday productivity. When I'm

stuck, I want to come to you and get

pragmatic and practical advice on how to

improve my habits and my output. You'll

notice in my prompt here that I'm giving

it something more specific. I want you

to always begin by encouraging me and

then I want you to respond in no more

than three sentences and then I want to

finish with an inspirational phase

phrase. So what happens here is that it

is going to interpret all of this and do

the configuration in the background and

set up an agent for me. So let's take a

look at what it's done. Firstly, it

gives me a bit of a sort of confirmation

back that it understood. It has given it

a name, everyday productivity coach.

Fairly good so far. So good. Now, before

I go ahead and test, what I'm going to

do is switch across into the configure

tab. Now, if you saw my screen before,

when I was on that configure tab, you'll

notice that was all blank. I will show

you you can start there if you like. But

the way this works and especially with

all of the the way this is working

recently is just that two or three

sentences. If I'm kind of clear on my

intent, it is able to actually

extrapolate that into a set of very

detailed instructions for this agent.

Think of this as like writing a job

description. So this is now where I, as

a person at work, can have a whole team

of agents helping me with things. And

these are very specialized agents.

They're like subject matter experts. So

now this is my productivity agent. I

could have dozens of agents doing all

sorts of different things and they work

at their best when they are very

specific. So this is like I've hired a

little productivity coach, virtual coach

into my team. So it has a description.

Again, think of a job description.

There's usually like a purpose statement

at the top. So that's what that

description is doing, its overall

purpose. And then in the instructions

here, you'll see that we've got quite a

lot. It's pulled out quite a lot of

things based on the way that I've

described it. So, the approach is very

much coming from the way that I

described that I wanted it to work. It's

got a bit of a mission statement there

as well, which I like. It's guidance and

tone. Uh, we've got some examples. So,

it's it's kind of working well with

providing an example of how we want it

to work, limitations, and so on. Now,

you've actually got with this

description here 8,000 characters. So,

if you want to come back and work on

this or if you want to write something

yourself, I'll share with you a little

bit later some best practice guidelines

for this. You can actually go into a lot

of detail. Now, I can edit this from

here. If there's anything here I don't

like, I can change it. But let's see how

it goes. So, for now, we'll come back to

that knowledge piece if you took a peek

at that in a minute. I'm just going to

give this a try. So, I've got a test

window over the side here.

And I'm going to come back and put in

that exact same prompt. And remember

what it did last time. It gave me advice

of a general nature with pretty emojis

and kind of general advice. I've asked

it to behave quite differently here.

I've asked it to kind of start with some

encouragement, be concise, be pragmatic,

end with something inspirational. So I

hope you can see there the difference in

how it's responded and how much I've

been able to guide and control this. So

the concept that to get your head around

here is that even though we talk about

this as building an agent. So we come in

here and click new agent and and we talk

when we talk about this as building an

agent. Realistically what I'm doing is

refining the existing co-pilot chat

experience because I can do this exact

same thing in the co-pilot chat

experience and get a general answer. But

now I want to have a subject matter

expert that is doing it my way in the

way that I've described. And look how

much control I have over that. So that's

one example of what you can do. Now, if

I want to go ahead and save this agent,

I'm just going to click create. And that

will now add that into my experience of

working with Copilot Chat. And I can

very easily come back and use that

anytime I like. So let's just close that

down. you'll see it comes over here into

my agents experience. And if I hover

over that, I've got some options. Let's

just um sort of go back out of it and

back over here. If I click on pin, then

that means it's pinned into that

section. So, it's going to appear at the

top of the list if this is something I

want to come back to. Now, I select it

and it is sitting in there. And if I

wanted to go back and make any changes

to that, the three dots at the end here,

I can click edit and I go back into that

configure, make any changes I like and

update it. And you will find that as you

work with agents. And in fact, as you

work with any of these AI tools, don't

think of it as perfection the first

time. Think of it as an iterative

process. So have a go with that the

first time. Put in the description of

what you want it to do. Let it do its

thing and use it for a bit. And then

you'll find ah maybe that encouraging

tone is a bit annoying. Let's go back

and change that. Or let's go back and

say actually I want you to give me a bit

more depth or other things that you

might want to do. So if you come back in

here to edit, you can edit either by

directly changing those instructions or

you can actually come back into here and

describe how you want to update that

agent. So we'll come back to that in a

moment. The other way that you can use

it is that when you're in the chat, you

can actually talk to that agent, you can

at@mention it the same way you might

at@mention a colleague. So again, if I

want to have this I have a big piece of

work to get done, you know, how should I

do it? Instead of just coming in and I I

could switch across into my productivity

coach, but now that it's there, what I

can do is type in this at@mention and

say productivity coach. And now when I

put that message in there, it's actually

going to call on that agent to give me

the answer rather than doing it in the

general chat. So the idea here is that

you would have dozens of these and

you're calling on them to help you with

specific things and really tailoring

that experience of what you're getting

out of that co-pilot chat to make it far

more valuable than just that very

generic AI type of response. Let's do a

different version of the same thing.

This time what we're going to do is have

a look at working with a knowledge

source. So I'm going to come over here

again and say I'd like to create a new

agent. Again, we'll go back into that

describe tab. This one is now a little

bit different. What you'll see I'm doing

here is you should only provide advice

from the content in the knowledge

provided. And what I'm doing is putting

in a website here which is James Clear

who's the author of a uh a book um on

productivity that that's um that's very

famous and there's a lot of content on

his blog here like he's got a lot of

articles and things and let's say I'm

really you know into this idea of

working with atomic habits and I want to

do that. Now I could actually put in the

name of that and say follow those

guidelines but instead of that I'm

saying I want you to just use this as a

knowledge source. I don't want you to

kind of generate stuff out of your own

brain. So, not that it has an actual

brain, but you know what I mean. So,

what we've got here now is that it's

coming back and adding those things in

here. So, now we've got Everyday

Productivity Coach. We've already got

one called Everyday Productivity Coach.

So, let's change that. So, I'm going to

call this one atomic

atomic productivity coach. And you'll

see that it is now grounded solely in

the content of this. But something else

has happened here. If I come down to

knowledge, you'll see that it's got this

website in there. Now, it's picked that

up from my instructions. But if you

wanted to do this yourself without

including it in the instructions, let's

say I came back to the other one and

said, "Actually, I want you to just use

this website." You can just enter that

website in there and add it as a

knowledge source. And the other thing

you want to do here is to say only use

specified sources. So if I toggle that

one on now, it will only use that. So I

should only get advice based on what's

going on that it can draw from that

website rather than the general answer

that it gave me before. So let's create

this one and have a look at the

difference. I could have done a test in

there as well. Let me just grab my um

same prompt again because I think it's

useful to kind of see a direct

comparison of the differences. and that

will also appear over here on the side.

Let's go straight to the agent from

there. And I'm going to put that same

thing in here. You'll see those starter

prompts. We'll come back to those in a

minute. So now what happens is that it

is referring to James Clear. It's his

website, his blog that I've used. And

now we're talking about the twominut

rule, the next specific action. So

you'll see this is a very different type

of response than what I got before

because before I said just use general

AI but answer me in a certain way. Now I

didn't specify those things but actually

I asked for it to give me advice based

on that particular knowledge source

that's sitting on the web. Other uses

for this if you've got say your own

website with a lot of jargon on it or a

glossery on it. You could create an

agent for new employees or even for

yourself that helps you understand your

organization's jargon. If you've got a

partner that you work with or another

organization you work with, if you're

say in a an organization that has a lot

of regulations and and has government

websites and things you need to refer

to, you could create an agent based on

that. I will give you a couple of

warnings about use using websites and

where the limitations are. So, let's go

back into this one, the Atomic

Productivity Coach, and just go into the

edit section here and take a look at

that. So, when you're using websites

here, you'll notice that this one is a

single website with a single level down,

jamesclar.com/productivity.

Let's say I had a website that was

further down the URL path than that. So,

I'm just going to show you a Microsoft

365 website as an example here. This is

a co-pilot learning site. So let's say I

wanted to create an agent that was

grounded on that's the phrase we use

where it's connected to that data and

just responds to that data. We say it's

grounded on that data. We've got see

here learn.microsoft.com 1/en

us1/copilot

1/icrosoft 365. We've gone more than two

levels down and you can't actually do

that. So if I come in here and say add

that URL

you'll see it gives you an error. The

website can't be more than two levels

deep. So that's always going to be a

limitation. You can go further up. So I

could go further up the chain here and

just do copilot. It will actually read

everything all the way down the

hierarchy, way more than two levels, but

the starting point can only be two

levels deep. So that is a limitation of

using the website. The other thing to

look out for on websites is that you

need the content to be on the website.

These are both good examples because

with this one, this is a blog. So the

the content is all just sitting here.

It's just text that you can that you can

read. Uh something like this again has a

lot of text in it that you can read. But

if you've got a website that has a lot

of codec coded elements like JavaScript

elements that are displaying tables that

aren't actually tables underneath or

displaying things in a way where the

substance of the words isn't actually on

the website or say it's an embedded

e-reader in the website, you're going to

have a problem. So basically what it can

read is the same thing that a search

engine indexes. So, if your favorite

search engine can't find that content,

your copilot agent also won't be able to

find that content. Let's take a look at

one final example here of working with

this style of agent, but using all

websites. So, when we made the first

agent, I just asked it to give me a

general answer. The second one, I asked

it to be grounded in or refer to only

that one website. This one, I'm saying

use websites for reference and provide

me with links for further relevant

reading. So this is one that's going to

be sort of a bit more uh giving me some

research and reading guides and things.

So if I have a look at the configure

option here, what happens is if we go

down to the knowledge, then it allows it

hasn't actually done this direct from

the instructions. You do have to toggle

this on search all websites. So you

remember in the previous example, we

just had that single website for the

James Clear blog and then I said use

only specified sources. In the very

first example, I didn't toggle this on,

which means it's not reaching out to the

web. It is just using the AI's internal

language, predictive language. Whereas

this one, I'm now saying, actually, I

want you to draw on all of the sources

for the web. So, if I come back in here

and do the same thing again for the

third and final time, this is a

productivity advisor. You'll see that

this is another different example of the

result that you can get because this

time I've asked it to go out and find

other things on the web. So now we're

getting kind of uh advice here around

different things including so you'll see

this one's come from the Harvard

Business Review. So again you can refine

this to encourage it to give you more

links but it's drawing on those sources

from outside. So that's websites. The

other thing that you can do here is work

with these suggested prompts. So, you'll

see that this one came up with some

suggested prompts to put on the front

page when you use it. Now, sometimes

that's incredibly useful. Sometimes it's

not incredibly useful. So, if you like

these ones, awesome. But you can very

easily come in here, you can edit the

title, you can edit the message, you can

delete them, you can add new ones. So,

you can do whatever you like here. You

can have anywhere between zero and six

of them. So that's entirely up to you.

Sometimes in the description, it will

generate those suggested prompts and

then do what you will to make them

useful for you. Now, you will have seen

as we went through and wrote those

descriptions that it's filling in that

configuration box for you with an

overarching purpose in the description

at the top and then a bunch of different

sections below. Honestly, as you're

getting started, that is far and away

the most easy, the most effective way to

do this. But here are some guidelines

and I'll put a link to the source of

where this comes from. There's a

Microsoft documentation site that has

some more detail if you want to follow

up on that where you can actually write

your own instructions. So instead of

starting from describe, you can just go

straight into that configure tab with

the blank box and just start typing in

whatever you like and create that from

scratch. So as you get more experience

and confidence and if you want more

control with this, these are the types

of things that you should put in there.

And what you can do is just go in here

and say new agent and just put all of

those things in. Here's my agent. Here's

the description. Here's the

instructions. Just start typing it and

and and work from there. Select what you

want in there for a website and and how

you want it to work. So you have

complete control over those things. Now,

there is one other last thing that you

can do here with the free version of the

license. Actually, two things. One of

the things you can do, let's go back

into our um everyday productivity coach

and I'm going to edit this. So, one of

the things that you can do here is give

it a new icon. So, it'll give you these

default kind of icons in pretty colors.

But if you edit this, you can browse

sort of standard little icons here. So,

you know, something like that might be

quite nice for productivity. You can

upload a file. So, if you've got a PNG,

it has to be quite small. I think it's

192x 192 pixels, no more than one um

than one meg. Or you can use an AI

generated image. Obviously, you can go

back into the main copilot or other AI

tool of choice and generate an image,

but this can actually now generate

something directly from the description

of the agent. Now, I will confess I'm

doing this live for the first time

because this is actually a fairly new

feature at the time that I'm doing it. I

am trusting in the process and I do

enjoy the thrill of doing something live

and hoping for the best. Looks like it's

going to come up with something fairly

[laughter] fairly fairly abstract. Oh,

here we go. Oh, no. It's cute. Oh, look

at that. Oh, see, look. It was worth

trusting in the process. It hasn't even

finished this, but I love it already.

Uh, while we're just waiting for this

little cute icon to uh [laughter] to

generate, don't forget if you're getting

value out of this, please like, share,

subscribe. All of those things really

help with uh with growing the channel

and helping this content get out there.

Oh, do we love that or what? That's I

mean, honestly, image generation is

getting a whole lot better. I'm just

going to apply that. And there it is.

It's so cute. I'm very glad I very glad

I did that. There's one more thing I

want to show you here that you can do in

your agent still in the free version of

Copilot Chat before we move into the

limitations uh where you need the paid

version. So, we go all the way down

below the knowledge here. There are two

capabilities that you can choose to add.

This is as simple as enabling a toggle

switch. And this is just a choice about

whether you want your agent to be able

to do these things or not. So, what I'm

going to do is switch both of these on.

The first one is giving it the ability

to create documents, charts, and code.

Now, for this use case, probably doesn't

really map across. This is more useful

if you've got an agent that sort of

needs to help you analyze and understand

things, but you are giving it the

ability to generate a document of some

kind. We'll have a look at that. And the

other one here is something where we're

going to allow it to be able to create

images. And again, if you don't switch

that on and the person using the agent

comes in and says create an image, it

just won't do that. So that's just a

decision there. Update at the top is the

sort of save to keep it going. Uh that

is the decision that you need to make

whether you want your agent to have

those capabilities or not. So let's go

back into the agent now and let's say

generate a spreadsheet.

showing me how much time I could save in

a week if I used proper

productivity

techniques.

So, we'll see we'll see whether it can

do this or not. If you put in a prompt

like the one I used before and then said

generate a word document with this

content, then it will actually literally

take the content and do that for you. So

what this is doing because I've added

this type of uh it's called code

interpreter even though the description

says create documents. What it actually

does is writes code underneath. You

don't have to know anything about what

it's doing. It does that for you and it

can actually kind of generate more

sophisticated types of things here. So

this has now given me a weekly time

savings calculator. So you'll see that

that is in fact an Excel document. Let's

open that file and

see what we've got. So here we go. So,

it's put some things together. Total

hours saved per week. It hasn't kind of

put the Oh, so it's asking me to fill in

the number of hours, hours saved per

week, and it's going to do that

calculation. So, it's basically created

a template for me. I could push further

with the prompting to ask it to actually

suggest times and things, but that's

actually sort of not a bad starting

point. If I come back into my So, again,

I can go new chat and stay within that

agent. And let's come back one more time

to this example we have about the big

piece of work. and what it's going to um

what it's going to to help me with. And

then I'm going to say, let's say, please

generate a fun image to remind me of

this. And that will trigger the image

generation capability that I switched

on. That will take a minute with the

magic of video editing. I'm just going

to pause and come right back when it's

done. And there it is. Not quite as in

love with that one as I was with the

icon, but you get the point. I didn't

write a particularly good prompt for it.

So, the better the prompt, the better

the image generation. Incidentally, off

topic, I'm left-handed. AI generates

every single image of somebody writing

with a pen with their right-handed. If

I've got any other fellow left-handers

who are watching and you've figured out

how to get AI to generate a left-handed

person, do let me know. One of those uh

little biases that's uh that's in the

training data that I haven't been able

to resolve or I haven't tried recently.

Okay. So, let's have a look now at where

we hit the limitations because pretty

much everything I've shown you now is

what you can do with the free version.

So you can write really good

instructions to get it to respond

exactly the way you want. Connect it to

websites, multiple websites like all the

internet or just a specific website. You

can have it generate documents. You can

have it uh generate images. But here's

where we hit the wall. So if I come in

here and say let's um let's start with a

new agent because it is a little bit of

a trick here that might mislead you.

When I go down into the knowledge

section, you'll see the only options are

to add a website or to add all websites.

That's it. This is the biggest

limitation between the free license and

the paid license. If I go into the

describe one here, let's say I'm going

to do something a little bit different

here. So, I have got a pets at workday.

Let's say we want to have something

where we allow people to bring their

dogs and cats into work. So, I can say

help me plan our pets at workday. Now,

there is an add content here which makes

it look like you can bring these things

in here. So, I've got my proposal for

pets at workday and I've also got a

budget file in here. Now, it looks like

I've added those as document sources.

What actually happens here is that it

can not it doesn't actually keep them as

sources because these are files that are

sitting on my one drive on my sharepoint

and even though I can access them there

that's actually because when I run the

agent I can upload files but it's a bit

of a it's a bit of a trap because what

happens is that it'll actually say it's

giving you this warning now. So I've

tried to do something and it's saying I

can't add SharePoint files. I can't do

that. Um, and so then it's giving you

that warning. So, we'll go ahead and

switch across into the full version of

the license, uh, where you've got a paid

Microsoft 365 co-pilot license. So,

someone in your organization has paid

additional money and assigned that

license to you. And we'll take a look at

the difference in that experience. So,

as I showed you earlier in the video,

the experience when you're starting here

with a paid Microsoft 365 copilot

license, you've got a couple of clues

that you have that license. You have

both a work and web tab at the top here.

And down the bottom here, you will see

that you have a Microsoft 365 co-pilot

license. Doesn't want you to upgrade now

because you've already got it. And

you'll see here the key difference

answers are informed by your work data.

So this as a starting point in the chat

is connected to everything that you've

done in your Microsoft 365 world of

work. This is called work IQ. And this

is essentially what you're paying for to

be honest with the Microsoft 365

license. It knows your context at work.

It knows where you sit on the org chart,

who reports to you, who your boss is. It

knows your collaboration, who you work

with, the documents that you work on in

SharePoint and one drive. It doesn't

reach other systems unless someone's

explicitly connected it to those

systems. It has all your Teams chats,

Teams um conversations, your emails. So

it's got the full context of everything

you're doing at work as it stands and

that's the value of what you're paying

for. That's a concept called work IQ

which is to which is which is how we we

understand that. So what happens here is

that when you're building an agent in

the paid version that work IQ stuff

becomes available to you. So let's come

in here and have another go at this

bring your pets to workday agent. We'll

start by selecting new agent. Before I

go in and put that description in, just

want to scroll down and you'll see

something very different from what we

saw with the free version. The top

section is the same new agent

description. The instructions are the

same. Down the bottom here, these

capabilities and suggested prompts, all

the same. Where the value of the paid

license comes in when you're working

with agents is in the knowledge section.

So, you'll see here that you can add

files, meetings, chats, emails, and

websites. This allows you to have an

agent that is grounded on your work

data. So, what we saw earlier was that

you were working with websites, but you

couldn't upload any documents. What we

can do here is actually have it pointing

to all of the context of what I work

with. Now, this is the huge value of the

paid Microsoft 365 copilot license.

Whether you're in the chat or building

an agent, this is referred to as work

IQ. It has the context of everything

you're doing at work in the Microsoft

365 world of work. It knows who my boss

is, who my staff are, my collaboration,

who I collaborate with the most, the

documents I've worked on. It has access

to all of the documents that I have

access to in SharePoint and one drive.

It has access to everything I'm doing in

Teams. It has access to my email. So,

all of that is very, very rich. It's all

secure. So, you're bringing AI into that

work context, that work IQ. So, this

allows you to go much deeper and much

more contextual with the things that

you're doing at work. And that is

honestly what you're paying for over and

above having the free version. You can

also bring other data sources in here.

I'm not going to spend too much time on

this now, but you can see I've got

Confluence there. I've got another video

if you're interested in how to do that.

You need to be an admin and set up all

the right permissions to allow people to

use that. But the main thing that I want

you to focus on here is that in the paid

license, we've got access to all of

these things. You've also got this

reference or chart and profile info

option that wasn't there before. So,

let's work through this. If I come back

into this describe section here and say,

"Help me plan our pets at workday." Now,

this is something where I've given it a

fairly fairly basic prompt, but it's

going to take that and turn it into a

set of instructions. Again, if I was

doing this for real, I would be much

more specific with what I want it to do.

But have a look at what it's been able

to do just from that particular single

phrase. Right? Given me a whole lot of

various things in here. Now what I can

do now because I've got this as an

existing project and I can share this

with other members of the project team.

We'll come back to that in a moment.

I've got a PowerPoint deck. I've got a

budget. I've got a bunch of planning

documents that help me with this pets at

work planner. And I want to work in the

context of all of that stuff that I've

done at work. So, what we can do here is

that if you go into this box now,

instead of it just being a website URL,

you've actually got a whole lot of other

options. So, across the top here, I can

go into files, and this will show me my

recently used files. If it's something

you've recently used, you're making it a

whole lot easier for yourself cuz those

things are there. You can actually go

into SharePoint site. So, if you wanted

a whole site, you can come in here. Just

be aware of this. If you choose sites,

that is giving you the content on the

SharePoint site, not the documents. If

you want the documents for a SharePoint

site, you want to go in and choose

SharePoint files and folders or browse

for a specific one. Come back to that in

a second. So, you're going through

individual files that you can upload a

whole SharePoint site if you've got a

bunch of sites. You can go into Teams

chats and bring those things in there.

If you wanted to create an agent, let's

say you've got a group of colleagues

who's got a really long group chat with

a lot of answers and a lot of content.

You can actually create an agent that's

pointing to that specific team's chat if

you want. Uh you can also use it to

refer to meetings and the contents of

meetings. You've got a bunch of

different things that you can do there.

What I'm going to do here is just upload

these files. Now, if I didn't have them

in the recent files list, you can find

them by starting to type the name of it,

but you have to remember the name of it.

So, that makes it a little bit harder.

Uh, if it's in the recent list, it's

just a whole lot easier to grab. But, if

you want to browse for those files or if

you want to add a whole folder, this

little cloud icon at the side here will

open up the cloud drive so that you can

then go in and work with browsing those

things. So, let's say this is on my

communication site, uh, and I've got a

whole thing on responsible business and

I wanted to add every document from that

folder. That's how you can add a folder.

Or if you wanted to upload them from

your desktop, if you don't have them in

there, you can upload them using that

little upload arrow. So, you've actually

got a lot of options here. We're going

to come back to the email one in a

second. So, just remember when you're

creating an agent with the with the full

version of Microsoft 365 Copilot, when

you're in the main chat,

it's got access to that whole work IQ,

everything you're doing at work. When

you create an agent, you're actually

narrowing down that experience and being

more specific about what you want it to

do. So, the starting point is that if

you don't add any knowledge, it doesn't

use any of those things. So you're kind

of starting from blank and saying which

things that Copilot has access to do you

want this agent to use. So this again I

think I said at the start it's like we

talk about building an agent as if it's

adding something. Really what we're

doing here is focusing the attention.

The main co-pilot chat already has

access to all of that. When you build an

agent, you start from blank and say

these are the bits that you as my

subject matter expert on this particular

topic should refer to and should have

access to. So if you don't add anything,

it won't use all of that stuff. The

co-pilot chat will the main chat

experience will with the with the full

license, but not if you don't add it to

the agent. Just take a minute on that.

It does take a bit of of sort of getting

your head around it. So now what we've

got is um let's let's also allow this

one to create documents and charts

because we might want to be able to do

that. And I am going to create that

agent and put it in there. Now let's say

for this one that I am actually working

with a team. It's likely that let's say

I'm on a committee and I've got a pets

at workday committee that I'm working

with. When I create an agent to start

with, it is private only available for

me. But I can choose to share it with

other users. And then I can choose who I

want it to be available to. If I choose

anyone, then I can share it with anyone.

And some depending on your

organization's permission, that option

might not be available to you. I've got

a system here that's allowing me to do

anything. If I say specific users, then

I can add specific users in there, or I

can go in and say I just want a certain

security group. So this is my demo

account, which is um Miranda. I'm going

to go in and add me, my actual main

account, as somebody that I'm going to

share it with. Uh, and then what you do

to share it is copy the link. Select

continue. Now, the important thing to

understand is that me, in this case, the

Miranda version of me that's logged in

and created this, I have access to these

documents as the user. If I create this

agent, then this is going to affect what

the other person can have access to. So

if I share an agent and the underlying

data sources that other person doesn't

have access to them, they won't be able

to get to them that way. So what this is

doing is giving access to those

knowledge sources. So again, this is

going to really depend on whether or not

you as the loggedin user are allowed to

do that. Your IT admin will have set up

permissions to do those things. But just

be aware that it especially if you

create something let's say I create

something on my email and then share

that that is not giving someone else

access to my email and you can't

actually grant permission to your email

but this is what we've got with these

documents. So in this case the other

user is me and has access to everything

but Miranda sends me that link. I paste

that link into my browser and I get this

experience here where this agent comes

through to me and I can just click add

and that will add it into my list of

agents. So the next thing I want to talk

to you about is the out ofthe-box

agents. You can get there to the agent

store by clicking on this all agents

button on the lefthand side menu. Most

of these agents are available to you on

the free version of Copilot Chat, but

there are two very special ones in here,

researcher and analyst, which are only

available if you have got that paid

Microsoft 365 copilot license. So, we'll

have a look at both. Now, there's a

whole lot of different things in here in

the built by Microsoft section. If you

scroll down, there are some really good

ones in here like prompt coach, idea

coach, writing coach. The prompt coach

is actually one of my favorites. If you

like any of these, you can click on them

and just click add and it will add it

into the list. And this has got a whole

lot of again in the background a whole

lot of instructions to help you improve

your prompting skills. So if you go into

the main co-pilot chat and just say,

"Help me improve this prompt." Again,

like we saw right at the start, it's

going to give you that advice of a

general nature. This one has a bunch of

instructions behind it that are using a

framework of how to write a good prompt

that's actually going to give you a very

structured way of doing that. That is

much more valuable than just asking the

general chat. Again, it's the subject

matter expert on how to write effective

prompts. So, there's a bunch of things

in there that you can play around with.

I want to turn your attention now to the

researcher agent and the analyst agent

which are honestly very very high value

agents that you get with that paid

Microsoft 365 co-pilot license and they

are sitting in this agent section if

you've got access to them. So let's

first take a look at the researcher

agent. This agent uses what's called

chain of thought reasoning. It's a much

deeper reasoning. So you've seen with

the other agents we've used so far they

sort of take a question and just respond

fairly quickly. This one actually kind

of goes, how should I solve that

problem? What should I do? And it works

with everything on the web and

everything in your world of work. So,

it's got access to all of that work IQ

as well as everything on the web. And

this is like your research assistant.

It's going to go out and do some

research for you in depth. So, what

we're going to do here, you want to give

this a problem to solve rather than sort

of a specific question. I'm preparing a

pitch for pets at workday. Please put

together a briefing paper to help me

understand everything I need to know.

And this is something where if you've

been tasked with a new project, uh some

kind of presentation where you need some

background information, it's actually

quite good as a learning tool as well.

Help me learn about this new subject

from a specific level or a specific

point of view. These are all really good

ways of using the researcher agent. I

find especially people who are in more

sort of senior executive type roles,

this one is often the highest value

thing to work with in in Microsoft 365

Copilot. So, it's actually giving me um

you're preparing a pitch and you want a

briefing paper that covers these things.

So, it asks me some specific things. Is

it a specific industry? Would I like to

focus on small, medium, any of those

kinds of things? So, what I'm going to

do here is just give it a little bit

more context. you you can just not

answer these questions if it's not

helpful. So, I'm just giving it a little

bit more. I'm also very deliberately

choosing I'll press the button and then

show you what it does. I'm going to say

I want short rather than long because 1

to five pages is still quite

comprehensive. Long is like a really

deep research paper. But you'll see I'm

giving it specific things. I'm asking

about the Australian workplace and I'm

doing that to prove I'm in Australia. I

know most of you aren't, but I'm doing

that to prove how tailored this can be

to my particular work context that it's

not just giving me generic information

that it can really kind of be very

specific. Now, look at what it's doing.

Crafting a pitch, organizing key points,

it's thinking out loud here. And you can

expand this box to see the full history

of what it's doing. So, it starts to

work through how it should do these

things and any problems it finds or any

any rethinking. Oh, now that I found

this, I might also try that. So, it's

actually quite interesting to sort of

watch the thought process go through

here. Now, because this does take a

while, I'm going to show you one that

I've prepared earlier. Exactly the same

prompt, but it's got it sitting in

another open tab here. So, I've done the

same thing. Uh, actually, look it. So,

if you switch tabs, by the way, it

actually kind of encourages you to go

back and gives you a little

notification. So, it actually didn't

take as long. Oh, no. It's still it's

still going, but that's the experience

you'll get if you go away and leave it.

So, here is my report. Pets at workday

briefing, benefits, challenges. So, now

what I've got is some nice icons that I

can work with and sort of copy those.

We've got a table with an executive

summary, the benefits, and if I hover

over these

references, this is the ABC. This is the

uh the the official um broadcaster here,

like the Australian Broadcasting

Corporation. So, that is a very reliable

source. Uh the New Daily, I'm not

familiar with this one, I must confess,

but it's a.com.au, which is my area. Uh

pound pause. So this is something that's

clearly some sort of site that is

relevant to this topic and again locally

here. So I can see that it is actually

bringing things in from the specific

context that I asked for. And you go all

the way down like this is honestly

really really high value stuff to get

across something very big very quickly.

And then I've got options here to say

well let's open that in word because I

want to take it outside um and continue

working with it in Word. And then I

would need to go ahead and save that as

a document. But you'll see it's brought

that in. So I can go ahead and work with

that there. Or if I wanted to work with

it in here and make some edits. You've

also got this option to edit in pages.

And that opens a sidebyside editable

version of this. So now I can come

through and I could make some changes

and do some things. I could actually

share this with somebody else. This is a

Microsoft loop component. So, two of us

could be working on this in real time

and then I can take it out into Word or

wherever else I want it to be. So, we'll

just close that down. But either way,

you've got ways of taking that output

and putting it somewhere else. The

analyst agent is the other one that's

very high value, and this is designed to

help you with data analysis. Now, what

I'm going to do first here is add some

documents. I have actually downloaded

these from a a public website that is

about university admissions in

Australia. But grab whatever data you

can find. Hopefully, you've got a real

work scenario here where you've got a

spreadsheet with a whole lot of stuff or

just find some other kind of public or

government website and grab some

statistics. If I show you what this

looks like, this is a series of little

PDF documents here that talk about the

sort of application by gender and and

where they're from and age and so on for

applying for university study in

Australia. So, what I'm going to do is

come back into the analyst agent and

select add content. And if I choose the

paperclip, add work content, that allows

me to go to my SharePoint files. But

actually, I just downloaded these from

the web. So, I've got a whole folder

here of demo documents that I've

downloaded. So, I've actually just got

these three tables here. I'm going to

choose all three documents and add them

in there. And then my prompt here is

going to give it some context, which is

actually really important for getting a

good result. These files show

undergraduate applicants and offers for

tertiary study in Australia. Here is

what the data is and how it's arranged.

And then I'm giving it a problem to

solve. what trends or outliers do you

identify? Now, it goes through this same

chain of thought reasoning process. So,

it's actually kind of going through and

you'll see I've actually given this

these are PDF documents. They're not

Excel spreadsheets. So, it's actually

able to read that from the PDF document.

Uh it goes through and navigates and and

works through all of those different uh

different options. And again, I'm just

going to stop this one actually because

I did this one earlier because it does

take a little while. So here's the

response we get these files. So it goes

through and does that for a couple of

minutes and then provides me with these

kinds of trends and insights. So then

what I can do is ask it to say uh please

provide please generate

a chart showing the trend

of total applicants

over the last 3 years. Now this actually

uses a code interpreter capability. So

it is capable of creating quite

sophisticated charts. This one, I

haven't been very specific about it, but

it can plot all sorts of things and help

you analyze data. So, if you've got a

real world scenario where you've got

massive spreadsheets and you're in

charge of having to kind of interpret

things and write reports and so on, this

one is also again kind of at that at

that manager level, executive level, a

very very high value uh thing that you

can that you can do. So, while it's

doing that, if I just scroll back up the

chat, you've got examples of these are

quite basic ones, but if you're asking

it for other things, it can do that. You

can also ask it to say, "Please give me

a commentary that assign that that goes

with this." So, I'm imagine I'm having

to present these results back to an

executive board or something. It's

actually helping me with kind of giving

me a summary and giving me some visuals

that I can that I can use in there. And

again, we'll just stop that from working

because it's it's done its job. These

ones, researcher and analyst, do have a

cap on usage. Uh, at the time of

recording, that is 25 uses per month

across both of them. So this is not

something I would say not every day, but

most of us have 20 working days in a

month. So you could actually use it

every day, but it's not the default

option. It's not something you're using

several times a day. You want to think

through that this is something that you

need that deeper research or that deeper

analyst, but they are extremely high

value things for those use cases. You've

also got the ability to create and work

with agents on your data in SharePoint

and One Drive. Again, we're in the paid

Microsoft 365 Copilot license here. So

this is just helpful if you're working

somewhere else. So if I primarily work

in SharePoint rather than the main chat

experience, let's say I'm in human

resources, I've got all of these

documents on all our different policies

here. You can click on open agents at

the top. And this has got a readymade

knowledge agent that is working with the

documents that are sitting in here. So I

can ask, can I take a day off on my

birthday? We've got a birthday leave

policy sitting in there. and I haven't

had to build or do anything. This agent

is just existing in SharePoint against

this document library and it's able to

answer questions based on that document

library. If you want to narrow it down

to something more specific, you can also

do that from here. But let's uh give it

a second to answer this question. So

yes, according to the birthday leave

policy and it gives me a summary of what

that birthday leave policy is including

with the references to where that's

coming from. Let's say I only wanted to

do something on the birthday leave

policy. What I can do is select that

document and up on my menu here, I've

got AI actions and I can create an

agent. Now, when I go in here, you will

see an experience that hopefully becomes

quite familiar based on everything else

that you've seen so far. We've got a

name, we've got an icon, we've got a

purpose, which is the description, and

then we've got the sources. So instead

of having to create the agent first and

upload or point to that document, I've

just done it the other way round because

I'm starting from the document and

behavior is basically kind of the

starter prompts and some of the

instructions. So all of those elements

are there and you can have that agent

experience sitting sort of directly in

SharePoint for users who are working

with that. So I can actually do the same

in one drive. I've got a document set of

documents here that are related to

Contoso. I've uh borrowed these from a

Microsoft Learn site. Uh and you've got

the ability to come in here and create

an agent or to select documents and work

with Copilot there. That's all

relatively new. Let me know if you'd

like a deeper dive on the One Drive

SharePoint stuff. Just giving you a bit

of a a whistle stop tour of those things

as well. One last way I want to show you

that you can use your agents in context

then is working inside the application.

So working with Word and the idea here

is these agents follow you around

wherever you are. So the agents we

created earlier um are available. So

let's say I'm sitting in word and I have

got my something about premium t here.

What I can actually do this I've just

selected the co-pilot button on the

side. I'll close it down to show you

from scratch. I'm in word online but it

works in the desktop as well. You can

expand or reduce this to give yourself

more space. But this is exactly the same

as this main copilot chat experience

here. You'll see you've got the work and

the web. And what I can do is use this

menu to open the navigation panel and I

can use any of those agents including

researcher or analyst or one that I've

got in here that's called te assistant

um where I might you know be able to

sort of put something in and I'm in in

the context of working in my document. I

want to go back to that agent and get it

to help me answer some questions and

actually ask it to pull those things in.

Now you'll see it's picked up on the

context of my open document as well.

Suggest tea pairings for different

occasions. It knows I'm in the Mystic

Spice premium chai tea over here. So

it's giving me all of these things. And

if I want, I can just copy all of that

and put it in the Word document. So

you've got these things side by side. If

I go back into my chat and refresh this,

then you will find the T pairings for

various occasions chat that I just had

is sitting there. So it's actually just

surfacing that whole co-pilot chat with

those agents in a different surface. So

you can start to use these agents

wherever it is that you want to work. So

up until now we've been looking at

working with agents that are very

contained within this Microsoft 365

co-pilot experience or the co-pilot chat

experience. You'll see these are all

designed for everyday business users to

just make that experience better.

They're very clear guard rails. They're

very clearly constrained. There's no

code anywhere. There's not really that

many choices. This is all about working

with language, creating these subject

matter experts, and refining those

instructions. But I hear you ask, what

if you want more? What if you want

something that will run on a schedule?

What if you want something that can

connect to other data sources and where

you have greater control and you're

talking about the language models

underneath and you want to kind of get

right into it? If you want something

that can run on a schedule, at this

stage, these agents don't have that

option. But if you're in a Frontier

tenant at the time of recording or if

you're watching this later, maybe this

is available live, you will find that

there are options in here for things

called the workflow agent. I've got

another video that talks you through

what that does. That workflow agent will

allow you to create scheduled

automations. But if you want a true

agent that runs on a schedule or that

gives you far more options for things,

that's where Copilot Studio, the full

version of Copilot Studio comes in here.

What we're doing with this agent

building experience in Microsoft 365

Copilot is effectively working with a

light version of Copilot Studio. Whereas

if I want to go into the full version of

Copilot Studio, then I've got a whole

other set of skills that I'm going to

need and you've got a whole other set of

capabilities. And this is much deeper.

We're shifting the user persona here.

Everything up till now is an everyday

information worker, business user.

Copilot Studio is getting into a low

code maker actually through into prodev

makers as well. You can start from

scratch over there. But if you've got an

agent you like here and you're hitting

the limits of this and you want to take

it further, you've actually got an

option in the three dots up the top here

to say copy to Copilot Studio. It gives

you a bit of a briefing on what that's

going to do. And we click get started.

And this is where you're starting to

work with environments, application life

cycle management, all sorts of other

things in here. You will have to have as

a user the permission to use C-pilot

Studio. Your organization will need a

license. But essentially what this does

is takes the work you've done so far

with all of the instructions and all of

that information that's in there and

moment in time copies it across. It

doesn't it's not connected anymore. It's

it's genuinely like cloning it and

taking it off on its own path. So you'll

see this is the T assistant copy. All of

those things that we had in there are in

there including the knowledge source.

I'm just going to click create on this

so that you can see what happens. And

this now takes this agent into a world

of many, many more things that you can

do. Now, forgive me here. This is my new

book. I wrote a book. This is how big

the book is on things that you can do

with Copilot Studio. I wrote this thing

and I can't believe how big it is. It's

like 550 pages. So, there are plenty of

resources online. I've got a tutorial

here and I'm updating more tutorials,

but just to give you a sense of how much

there is to know and honestly there's

still more than I could fit in one book.

So that's a reference if you want to get

deep into it in a very structured way.

But essentially what this will allow you

to do is connect to all sorts of

enterprise knowledge, take control of

the conversation with workflows and

topics, trigger the agent based on uh

specific trigger points of data points

that happen or running on a schedule

more than I can possibly do justice to.

So if you'd like to learn more about

Code Pilot Studio, check out my tutorial

here or there's a link in the

description if you would like to get a

copy of my book in old school print or

ebook format. Thanks for watching.