Hello everyone. Creating AI agents used

to be a difficult job. You had to host

them, scale them, create APIs, and

manage all the infrastructure just to

make one agent work. But now things have

changed. With Azure AI agent service,

you can build powerful agents without

worrying about the infrastructure.

Microsoft runs the platform for you. So

you only focus on the agent logic and

the tools. And this is huge because most

of the AI projects start as a quick

demos. For example, you create a

chatboard or a single assistant. But as

soon as you need to create an agent

which has to do more things like query a

SQL database, search the web or search

the internet or send an email and all

this complexity creates the roadblock.

you need the orchestration, scaling,

monitoring, enterprise level security

and that's where the people usually get

stuck. So now using the Azure AI agent

service everything is changed. It gives

you a managed foundation where you don't

worry about the underlying

infrastructure. So hosting, scaling,

security as well as the observability

are all builtin. You just design the

agents, tell them what they should do

and connect the tools. And here's the

part I really like. Actually, it's all

inside the Azure AI Foundry. That means

you don't need to know the coding to

start. You can just go into the Azure AI

Foundry portal, create an AI agent

through the UI, connect it to the

knowledge tools or the action tools and

even build a multi- aent setup or multi-

aent application where agents talk to

each other. So now creating an agent

it's just a click away and it's very

simple to get started. So if you want to

just test and play with it there is a

playground which is available in Azure

AI foundry. You can run the

conversations trigger the actions see

how the agents respond. So all this can

be done without even deploying into any

production or any application. So now

let's understand the main building

blocks for the AI agents in the Azure AI

agent service. So the first one is the

knowledge tools. These let your agent

access the data things like the

documents in Azure azure storage or in

Azure AI search service. So you don't

even have to create a Azure AI search

service. You have to just upload the

document and it'll do the indexing

vectorzation and everything and your

agent will answer your questions based

on not only the general knowledge but

from the document intelligence. Now the

second one is the action tools. Now your

agent has should have the ability to act

and this is done by the action tools

where now you can query a database call

a logic app or hit an API or send an

email and finally the multi- aent option

instead of just creating one giant agent

you create the small specialized agent

and then connect them under orchestrator

and finally get a desired output from

the team of the agent. So let's talk

about the bigger picture. So Azure AI

agent service gives you a no code manage

environment inside Azure AI foundry. You

can spin up the agents quickly, connect

them into your tools, test in the

playground and scale them into the

production with all the enterprise

control built-in. So in this video I'll

show you the look and feel of Azure AI

agent service and then actually build a

set of agents, connect them together and

run a complete workflow setup step by

step. So we'll be creating a small real

world project which is like a property

management system. So I'll provide more

details later but before that I want to

show you how to deploy Azure AI agents

in Azure portal. I'm logged into Azure

portal now and let's start with the

creation of Azure AI foundry.

So let's go to Azure AI foundry create a

new one. I'll create a new resource

group RG- AI agent

and let's name it as Shelender AI agent

hub01

East US2 region

and

let's name the project as Challenger AI

agent project.

So we'll

keep everything default and then deploy

the Azure AI foundry hub as well as the

project.

So it's running the validation now once

the validation is done

and create. So now the deployment has

started. I'll pause the video and we'll

be back once it's done.

So the hub and the project is deployed.

Now let's go to the resource

which is Azure AI foundry hub and let's

go to Azure AI foundry portal.

So I am in the Azure AI foundry portal

and specifically in the project which we

have created.

So on the left side if you'll see there

are there is an option of agents. So

when you click here

it'll ask you to deploy the model first

because this is a new project we haven't

deployed any model. So it will just

provide us the models which are

supported. So these are the models which

are supported by Azure AI foundry hub.

These are all Azure OpenAI models. So

let's go with GT40.

Confirm.

I'll just customize it a little bit.

increase the token size because we are

going to create a multi- aent

application.

So the model is deployed which you can

see here if you go to the models GP4

model is deployed and as soon as you

click on the agents an agent is created.

So there is one agent which is created

and

you can try in the playground you can go

to the playground agents playground.

So if I'll write something

so it's just responding like a normal

chatbot.

Okay.

So now let's check how this agent works.

So there is a model which is deployed.

It's using that model. This is the agent

ID which is created. You can change the

name to anything. Then you provide the

instructions. These are the system

prompts which you are providing. So how

the agent will behave. System message or

instructions.

This is the description for you. It's

just so that you know what is the

purpose of this agent.

Then there are knowledge tools, action

tools and connected agents. So knowledge

tools when you upload the files you get

the AI agent search service by default

which is managed by the Azure AI agent

service itself and you don't have to do

anything there.

You can create your own Azure AI search

service and that will take care of

everything. You can connect to Microsoft

fabric shareepoint

bing search if you want to do the

internet search, trip advisor and other

application which you can connect.

Now comes the action tools as the name

suggests the different actions you can

take. So one is the code interpreter. If

you have some Excel file or some

documents where you have to interpret

the code, you can just upload it

directly here. It'll work for you.

Another one is the open API 3.0 to

specify tools. So that means if the

tools are not built here, you can use

the open API standard, use the API keys

and everything and it will work for you

perfectly.

Then the Azure logic apps. So these are

the different logic app actions or the

workflows which are already provided.

You have to just create them.

And then the connected agents and here

you can connect different agents with

each other. the temperature and topy

it's how how the randomness or the

response will be provided it's based on

that you can even enable the voice mode

however it's only available in voice

live APIs or the playground which I have

another video for this you can always

check that so now you have a good

understanding of how AI agents are

created in Azure AI agent service in

Azure AI foundry so

next step is Let's create a realtime

project so that you can understand how

these tools connect and finally how we

can create a multi-Agent application. So

I am going to create a real estate lease

renewal or property management system

which is a multi- aent system. One will

be the manager agent and then there will

be four different agent. SQL agent which

will do the SQL queries. Web agent which

will do the web search. document agent

which will look information from the

document and the email agent which will

send an email. So now let's understand

this in more detail. So there is a

property manager who has to check which

tenant lease is going to be renewed. So

the property manager will ask the

manager agent okay provide me the list

of the tenants whose lease is going to

expire. So the manager agent will

connect to the SQL agent get the list.

Then the next step will be because the

lease is expiring so there will be an

increase in rent. So the web agent will

look for the average rent for that area

currently. Then the document agent will

provide the options in case you want to

move out. You don't want to proceed

further or renew the lease. Then what

are the steps you have to take? Yeah,

how you can claim your bond and all

those details and the email agent will

collect all the information and send it

uh to the user or the tenant.

So now instead of having the multiple

people who doing this work, this will be

done automatically and there will be

only one property manager who will just

ask the question to the manager agent

and everything will be done by the

manager agent in the back end. So let's

start with the creation of all these

agents and then we'll finally create the

manager agent.

So now you can go to the agents.

So let's click on this agent first.

Let's change the name to SQL. So it's a

SQL agent. Straightforward name. And

I'll copy paste some information.

Let me make it bigger.

Okay. So, it's a SQL agent. I have a SQL

database which is already created and it

has a table named as tenant leases

table. So, it has multiple information

which it can pull. And now what we are

doing is we are providing the schema

that it has tenant ID, full name, email,

phone and all those details. So it has

to run the query and get the details.

But these are just the instructions.

What we have to do is connect to the

action tool. So click on the action

tool. Go to the logic app

and there is an option of execute SQL

query on a database.

So if you'll click here you have to

provide the name action

all those details you have to provide.

Then you have to authenticate

to the SQL server.

Then you have to provide the SQL server

details which will pro come in the

drop-down and then done. So all this

done all this thing is already been done

here. So I'll what I'll do is I'll just

quickly select the get lease expiring

and it has all the details already which

I was just showing you. So because I

have to create it create a connection

use the SQL credentials.

So that is already done in the

background. You can always click on the

connect and provide all those details

there.

So it's getting all the schema by itself

and create.

So let's quickly test this.

Try in the playground.

Provide me the list

of tenants

whose

lease is expiring

in September

2025.

So as you can see it's pulling out the

details directly from here. What I've

done is I've just used my email id plus

one used the multiple alias in the

database and it has pulled that

information. So and when you click here

view run info it has done the tool

calling. It went to logic app found that

information and provided to me.

So now one of our agent is working. So

SQL agent is working. Now let's create

another agent. Either you can create

from here or go to the agents create a

new one. You can use the different model

but we'll use the same model.

So now the another agent will be web

agent.

and I'll copy paste the system message.

So what is the purpose of this is it

just check the information in the web

based on the suburb as well as the

bedroom count and then provide the

details.

uh how much is the average rent right

now so that the tenant can decide when

we send an email tenant can decide so

this is the market price right now do

you want to increase or renew the lease

or you want to move out so then it will

be their choice

there is an option of using the banking

search however there is a limitation on

certain

accounts or subscriptions that you

cannot use the bing search because it's

a paid service and if there are credits

or visual studio subscription then it's

not allowed so I can't use it so what

I've done is I've used an alternative

Google SER API so you can look for it

SER API it's from Google you can create

your credentials and use it it's it's a

web search so I'm using this and for

this what you have to do is you have to

first create your credentials you have

to go to the management Enter connected

resources.

Create a new connection.

Go to the end. You can use the SER but

you have to provide the custom keys. So

what I'm going to do is I will provide

the custom keys here. I'll pause the

video. I'll just quickly provide it here

and add the connection. So I have

created a new connection sub API 1. So I

have provided the key sub API key and

let's go back to the project.

Let's go back to agents.

Select our web agent.

Let's again go back to action tools. Now

because ser API is not directly

available. So what we have to do is open

API 3.0.

So let's provide the name as web search.

and next.

Now in the connection we have to choose

the connection and then provide the

details

which is the schema.

So this is the schema and you have to

define what API key you have defined. So

this is this is how I named it API key

and you have to just define the schema.

Next create tool and it's created.

So now the web search tool is created.

Let's create another agent

the document agent.

So let's name it as document agent and

system message for this document agent

is that it just look for specific

information that how to get the bond

when you are moving out or what are the

things or the other formalities that you

have to take care and for this what we

are going to do is we have to add the

knowledge tool upload the file we'll

directly upload the file here renters's

guide which is so it's Australia IA

Victoria renters guide. So we'll upload

it here and it will provide information.

So now we have three agents created.

Let's go with another one and let's name

it as the email agent.

And there is a big system message for it

because we want it to send a proper

email.

So what it's doing it's it's using the

email tool and then getting some details

for example name email current rent

average rent and address and create an

email using all the details from uh the

SQL from web from document and send a

proper email to the tenant.

Now let's

let's go to the action tools

and in Azure logic app there is an

option of sending an email which is send

email using outlook. So you have to

provide these details go next

authenticate you have to connect however

I have already done this and you don't

have to do anything else just connect it

to your office 365 email and it's done.

So I am going to use directly which I

have created and this creates a

consumption logic app in the back end.

So click on email tool

and create.

So now it should be sending an email.

So now four agents are created. Now the

last one

which is a manager agent. Let's name it

as manager

and

the information provided here is system

message provided here is that first it

has to get the detail from the SQL agent

then search the web search the document

and then send email and it has to follow

a proper flow or process so that it will

take care now because this is the only

the system instructions and how will it

connect to the other agent and there is

an option of connecting agents. So now

when you have to connect the agent you

have to provide some information how you

will call that that agent.

So now start with the SQL agent we have

to provide a specific name. So let's

name it as SQL lease agent. And what it

does is it fetches the information

from tenant leases and

provide the list of the details of the

tenants. So add it.

So let's add another one and this time

let's add the web agent.

Let's name it as web rent agent and it

finds the average weekly rent.

It's getting added as you can see it's

saving your data and it's and let's do

the document one.

So it's a document tenency agent and

it provides some steps and guidance for

the tenants during the lease renewal.

And the last one will be the email

sender agent.

And it sends email using the name,

email, current rent, market rent, and

the renewal steps.

And this is the email one. And let's

add. So all the four

agents are added now. So now it's time

to test it whether it's working or not.

Let's see.

So let's try in the playground. So now

we are in the playground and the manager

agent. Let's start with hi.

whose

lease is going to expire in

September

2025.

So now it should provide me the list and

then I can just proceed with the one

which I want to.

So now there are two persons Ammon Gupta

and Daniel. So let's go with ammon. So

would you like to proceed with any of

them? I'll just say ammon.

Nothing much.

And here

it did the tool calling and it connected

with the SQL. Now other things are

happening in the background. So now it's

referencing. It will take bit of a time

couple of seconds

or a minute also. And because now it's

going through all the agents and then it

has to send an email.

So I'll pause the video and we'll be

back once it's done. Whose lease is

expiring. So Aman Gupta I just provided

the ammon and it's saying so these are

all the details. Do you want to send? I

say yes.

And now it will send all those details.

So let's click check the run.

It it did the multiple tool calling

here.

So the connected agent

so SQL ease agent. So first it went to

the SQL then

the web

and then finally the document one. And

now it's asking us to send the email. So

when we said yes, it should be going to

the email agent and then just sending an

email.

So let's see sent an email directly to

Aman Gupta and I didn't do anything

special it just got the got the list

provided the name and it has sent the

email. Let me quickly open my email and

show it to you. So I got an email I have

created an alias as you can see plus

one.

So and I got an email. I know the

formatting is not good but I can use the

HTML or markdown or anything. So it's

completely your call how you want to

format it. Now what it's doing is it's

telling first it's lease renewal notice

for the address and then it's going to

renew below are the details current

weekly rent average weekly rent in your

area and the proposed new weekly rent

because there is a 10% increase and some

details from the documentation. So all

that information collected together and

is provided here. So that's great. And

if you want to deploy this in your

application,

you have to just check this code and it

will help you deploy this. This will

give a good understanding but then you

have to create a chat application so

that it it works in a loop. So now to

summarize this video, I have provided a

deep dive information on what is Azure

AI agent service, how you can create AI

agents, how you can use the different

tools with your agents and then create a

multi- aent application where all the

agents work together based on the

information provided or instructions

provided by the orchestrator agent. And

I have created a real world small

project which is a property management

project where all the agents work

together and provided an information and

sent an email to the tenant. So that's

all I wanted to show in this video. I

hope you liked it. Please like and

subscribe. Thank you so much.