Make.com AI Agent Tutorial — Build Your First Agent in 20 Minutes

In this video, you're going to build your very 

first AI agent, even if you've never heard of that term before. You've probably used an AI 

chatbot like ChatGPT. They can answer questions and have conversations. Now, picture that same 

chatbot, but connected to all of your tools, so you can actually do things. Send an email. 

Update a spreadsheet. Post to Slack. And here's the big difference. An AI agent doesn't just 

respond to you -- it can also take initiative and handle tasks all on its own. That's the 

power of an AI agent, and that's exactly what we're going to build step-by-step. Hi, I'm 

Kevin, and let's dive in. To build our AI agent, we'll be using Make, and you'll find a link 

right here at the bottom of the screen. I partnered with them to show you how this works. 

If you've never heard of Make before, it's a visual platform where you connect your apps and 

services with simple drag and drop blocks. Now, with AI agents, these flows can think and 

act on their own. Instead of writing code, you sketch out how you want information to 

move, and the agent figures out the rest. Set up an account and I'll see you on the 

other side. Before we dive into the steps, here's the scenario we're going to build. At the 

Kevin Cookie Company, customers email us with lots of different questions. Some are simple, like what 

are your hours, or do your cookies contain nuts? Others, though, are bigger issues, like someone 

asking if we could ship 1,000 cookies by Friday. Our AI agent will take care of the easy ones all 

on its own, replying with the right information. And when it sees a special request, it'll escalate 

that message to Slack so a human can step in. Now, this is just one example scenario, but by 

the end, you'll have all the essentials that you need to create AI agents for whatever your needs 

happen to be. Once you finish signing in to Make, you'll land here on the main My Organization view. 

To create an agent, over on the left-hand side, under My Team, let's click on AI agents 

beta. Right near the top of this page, let's click on Create Agent. Right up on top, 

let's click on Create a Connection. Here, I can choose the provider. This is the brains 

behind the agent, and in a moment, we'll choose the AI model working behind the scenes. Different 

providers have different strengths. Some are great at reasoning, others at speed or cost. Pick the 

one that best matches what you want your agent to do. I'm going to use Make’s AI Provider since 

it works right away without any extra setup. I'll leave the connection name set to the default, 

and then down below, let's click on Save. Next, we need to give the agent a name. I'm going 

to call mine the KCC Support Agent. That way, I know what it does. Underneath that, we need 

to select a model. Now, we have a few different options. I'll go with large since I think that'll 

work well for most customer support tasks. Finally, we also need to enter in a system 

prompt, and this is where we describe exactly what this agent should do. Here, I'll 

type in some text. Now, if we go up, I want it to handle customer emails for the Kevin 

Cookie Company. If the email matches a known FAQ, well, draft a friendly reply. If it's complex, 

so let's say a bulk order, maybe a refund, a complaint, or maybe the request is unclear, then 

I want the AI agent to escalate it to Slack with a summary and also a suggested reply. And here 

are some guardrails at the very end. Never issue refunds or discounts on your own. Once you finish 

entering all this information in, in the bottom right, click on Save. This brings us into the main 

configuration view, where we can continue building the agent. Right now, my agent doesn't know how 

to handle customer questions. I need to give it some knowledge or context. Over on the left-hand 

side, you'll see the option to add context. This is background information the agent can reference 

to do its job more effectively. Right here, let's click on Add. Here's a quick preview of 

the file that I'm going to upload. It's the Kevin Cookie Company Frequently Asked Questions, 

or FAQ. It contains information like what are your store hours? Do your cookies contain nuts? Do 

you deliver? And lots of other questions. These are some of the most common questions that we get 

here at the Kevin Cookie Company. Let's go back to Make. Back within Make, I'll select my file, and 

then I can drag and drop it in. Right over here, I'll release it. And here we can see the Kevin 

Cookie Company FAQ. That looks good. Let's click on Upload. It looks like it successfully 

uploaded the file. Right down below, you'll notice a few extra options. We have 

something called MCP. This lets you connect the agent to outside tools and services. That's 

more advanced, so let's skip it for now. There are also system tools, which makes tools available 

to your agent everywhere. But in this video, we'll just assign tools later on. For now, over on 

the right-hand side, let's do a quick test. Let's ask the question, what are the store hours? If you 

recall that information is contained in the FAQ. Let's try sending that. Right up above, 

the AI agent pulls the information out of the FAQ into a nicely and friendly written 

response. Customers are going to love this. In the top right-hand corner, let's click on 

Save. Awesome. We just tested the agent, and what do you know? It can answer questions. Now, 

let's make it actually do some work. To do that, we need to set up what's called a trigger. A 

trigger tells the agent when to start working. In my case, I want it to respond whenever a new 

email arrives. So that new email becomes the trigger for the agent. Over on the left-hand side 

navigation, let's click into scenarios. In the top right-hand corner, let's create a new scenario. 

We're going to build from scratch. Right here, let's click on this option. Now, we're in a 

brand-new scenario. In Make, a scenario is just an automation workflow. Basically, a series of 

steps that run in order. This is where I can build out the workflow for my agent. The first step 

is to set up the trigger, and I want the trigger to be new emails coming in. So right here in the 

center of the screen, let's click on this big plus icon. Here, I see thousands of different apps that 

Make can connect to. With Make, you're not just limited to just one or two tools. You can link up 

thousands of services from Google Sheets to Slack to OpenAI and beyond. In my case, I want to watch 

for new emails. So right down here at the bottom, I'll search for Outlook, since that's my email 

provider. If you're using Gmail or something else, you can just type it in here and select it. 

Once I choose my app, right here at the top, I see a list of all of these different modules. 

A module is just an activity that happens within that app, like let's say watching for 

new rows in a spreadsheet, in this case, watching for new email messages right here at 

the top. I'll click on this one. This means that every time a customer email lands in my inbox, 

this scenario will automatically kick off, and my agent will know that it has work to do. If this 

is your first time using this service within Make, you'll have to start by establishing a connection. 

I'll go ahead and do that. Once you connect your account, you can configure different options. For 

example, do you want to watch all messages or just a specific folder? I'll click here, and let's set 

it up to look at just my inbox. You can also add other filters if you only want certain types 

of emails. Now, I'll leave it simple for now, and in the bottom right-hand corner, I'll click 

on Save. That way, whenever a new email arrives, the scenario will automatically start. Right up 

here, I'll set it to from now on, and then let's click on Save again. Now that we've added the 

trigger, the next step is to bring in the agent. The agent is the brain of this workflow. It's what 

will read the incoming email, understand it, and decide how to respond. So right over here, let's 

click on this plus icon. And here again, we see all the different apps that we could choose from. 

Now, we want to bring in the agent. So down at the bottom, let's type in Make AI Agents. And right 

at the top, we see the option for Make AI Agents, and we want to run the agents that we created. 

So, let's click on this option. Let's start by choosing the agent that we created earlier. Right 

here, I'll click on this dropdown, and you should see the agent that we saved earlier. I'll click 

on this option. Now that we've selected the agent, the first option is to assign tools. For 

example, we could let the agent send emails or post to Slack. Since we haven't built those 

tools yet, let's skip this step for now. Let's now look all the way at the bottom where we have 

this category for messages. Think of this as the input the agent receives. In this case, I want to 

pass in the customer's email details so the agent knows exactly what it's working with. Let's click 

on Add Item. When I click into the message field, over on the left-hand side, I see all the data 

that comes back from the email trigger. This includes everything from the subject line, the 

body, the sender's address, even the message ID. What I want to do here is map just the parts 

that my agent needs. So over here, I'll type in some text. I want the email message ID, the 

subject, the body, and the address. So right here, first off, let's find the message ID. Here, I 

see everything I'm getting back from Outlook, and there's the message ID. Let's now see if I 

could find the subject. Right here, we have the subject. If I look down a little bit more, we have 

the email body or the content of the email body, and here's the sender's email address. I'll make 

sure to select all of these, and they're all now inserted into message 1. This way, the agent 

has all the context it needs to understand the message and generate the right response. Finally, 

let's also include instructions on how the reply should be formatted. Right here, I'll click Add 

Item, and here we have an option for message 2. I'll type in some text, format the body content 

message to send to the customer in HTML. And here, I've also included some HTML tags that I would 

like it to use. I just want to keep it simple. We're now all done configuring the AI agent, so in 

the bottom right hand corner, let's click on Save. In the top left-hand corner, let's now give this 

scenario a name. I'll call it Respond to Customer Email. Then on the bottom toolbar, let's make 

sure to save it. I'll click on the Save icon. Over here, we can also define how often we want this 

scenario to run. I want it to run as soon as we receive an email. So instead of running at regular 

intervals, let's click on this dropdown, and right over here, we could select On Demand. Let's set 

that and then click on Save. And right here, we can now activate the scenario. Let's click on 

that. Then on the bottom, let's save once again. Now that our agent is created, we need to give it 

the ability to actually get things done by giving it access to tools. A tool is essentially a skill 

you add to your AI agent. On its own, the agent can only read and understand text, but with tools, 

you can extend what it can do. For example, we'll add an email tool which allows it to send emails. 

We'll add a Slack tool so it could post in the channels. You can also add tools for Google Sheets 

or CRMs so you can log data or update records. Think of tools as giving your agent arms and 

legs so it can take real action in the world, not just talk about it. To add a 

tool, let's click back into scenarios, and in the top right-hand corner, 

let's create another new scenario. Each tool is its own scenario that the agent can 

call when it needs to take action. We're going to start by building a customer email reply tool. 

Then we'll add a second tool or another scenario to post a message in Slack. That way, the team is 

notified whenever the agent escalates something. Since this email reply tool will be used by the 

agent, let's start by adding inputs. At the very bottom, let's click on this icon. Inputs are the 

details the agent passes in when it calls the tool. Let's click on add item. To be able to send 

an email, first off, we need to know which email. So, we want the email message ID. That's the first 

bit of information for the agent to be able to use this tool. Then let's scroll down a little bit 

more. Before we send the email, we also want to include the agent's response to the customer. 

So, we need another piece of input from the agent, which is the email reply text. That's all the 

information that we need as inputs. So down below, let's click on save. Now that we defined those 

inputs, we can actually build out the steps to send the email. Right here, let's click on this 

plus icon. Down below, I'll search for Outlook, and here, let's click on show more. Now, I want 

it to respond to an existing message, and here I see the option for reply to a message. I'll select 

this one. Here, we can enter the inputs so Outlook knows exactly how to send the reply. Right up 

on top, we have one of the key items called the message ID. This ties the response back to the 

original customer email. I'll click in here. And then right up on top, let's click on this icon. 

And here we have all the different variables. Let's select the email message ID that we defined 

on the input screen. Then we could close this. Then we need to map the email body that my agent 

generated. That way, the reply includes the right content. So, let's click here. And once again, 

let's click on this icon. And there, I see the email reply text. Let's click on that. Once these 

are set, the agent can automatically respond to the correct email with the proper message. In the 

bottom right-hand corner, let's click on save. In the top left-hand corner, let's give this tool a 

name. I'm going to call this the email reply tool. Then on the bottom toolbar, let's save this. 

We can also define when this tool should run. I want the agent to be able to use it on demand. 

Let's click on this. And then in this dropdown at the very bottom, let's select on demand and then 

click on save. And let's activate the scenario. On the bottom bar, let's save once again. And 

that's what's involved in creating a tool. We're going to do this once more when we create our 

post to Slack tool. So just like we did before, over on the left-hand side, let's click into 

scenarios. In the top right-hand corner, let's click on create scenario. And then let's start 

by adding the inputs. Down on the bottom bar, let's click on this icon. I want the agent to 

pass in some key details from the email. So, let's click on add item. And the first thing 

I'd like is the customer's email address. All these other settings look good. Down below, let's 

click on add item once more. And I would also like the email subject. Then let's scroll down a little 

bit more, click on add item. And for the last one, let's enter in the email body. That way, when the 

message gets posted to Slack, the team can see exactly who it's from and what it's about without 

even opening the inbox. Once all these inputs are set, in the bottom right-hand corner, let's click 

on save. Now let's add the Slack module and map them in, just like we did with the Outlook 

reply. Right here, let's click on this plus icon. Right down below, let's search for Slack. 

I'll type that in. And I want to post a new message. So over here, let's take a look to see 

if there's an action that aligns with that. Here, we have the option to create a message. Let's 

select that one. And if this is your first time connecting to Slack, you'll need to create 

a connection. I'll go ahead and do that. Now let's finish setting up the Slack tool. Right 

up on top, I can choose the channel where I want it to post. I'll click on this dropdown 

and let's select from a list. Right over here, I could set it to a public channel. That sounds 

good. And right here, I could select which public channel. We have one for customer escalations, so 

I'll choose this one. Finally, in the text field, we can map in details from the email. So over 

here, I want to feed in three different pieces of information, the customer email address, 

the subject, and also the body. So over here, let's first get the customer email address. 

Remember, we configured this as one of the inputs that the agent will send to this tool. Right up 

on top, let's click on this icon, and here I see the customer email address. I'll select that. Here 

for the email subject, let's select this item. And right here for the email body, let's select this 

input. This way, whenever the agent escalates an issue, the team will see the full context right 

inside Slack without needing to dig through the inbox. Now that we've configured all this 

information in the bottom right, let's click on save. To finalize this tool in the top left-hand 

corner, let's give this tool a name. I'm going to call this post escalation to Slack. Then on the 

bottom, let's click on save. Here we could define when it should run. I want it to run on demand. So 

right down here, let's select that. Then click on save. Let's activate the scenario and then click 

on save once more. Now that we've built the tools, the last step is to give the agent access to 

them. Remember, the agent is the brain, but the tools are the hands. There we'll let it actually 

take action. So over on the left-hand side, let's jump back into scenarios. Back within scenarios, 

here we see a list of all the different scenarios that we've added. We have the two tools, and we 

also have the agent workflow. Let's click into the agent workflow. In the top right-hand corner, 

let's click on edit. Let's click into the agent module. Now that we're back in agent settings, 

let's connect the two tools that we just created. Right here at the top, we have the option to 

add tools. Let's click on this. Right up on top, let's select the first tool or the scenario that 

we just built out. Here we see all the different options. Now I see the email reply tool. I'll 

select this one. Down below, we also need to enter in a description, so the agent knows when to 

use this tool. I'll enter in the send email reply, sends a response back to the customer, tied 

to the original message, using the reply text generated by the agent. This looks good, so in 

the bottom, let's click on add. The agent now has access to one tool, so let's add the other. 

Again, let's click on add tools. And right here, we see the dropdown list with all the scenarios. 

And I'd like to add the post to Slack tool right here. Let's select that scenario. And here too, 

we also need to enter in a description. So, I'll type in post to Slack, shares customer 

email details in a Slack channel, so the team stays updated on new messages or escalations. This 

looks good, so in the bottom right, let's click on add. With those linked, the agent can not only 

understand customer emails that are coming in but can also respond and notify the team when needed. 

Now that we've added those tools, down below, let's make sure to save this. I'll click on 

this. Now that we have our tools and the agent is watching incoming messages, it's time to put it 

all to the test. Let's start by sending a simple email that matches one of our FAQs, something like 

what are your store hours? Down below, let's send that. Within Make at the very bottom, let's now 

try testing this out. Right here, let's click on run once. Right up above, I can see that it ran 

successfully. Let's now check the inbox. Here, the agent replies automatically with all of the right 

information. Next, let's try something a little bit trickier like here's a bulk order request for 

a thousand cookies with customized packaging with their company logo. Now that's not something the 

agent's going to find in the FAQ. So, let's see if it escalates it to Slack. This will show that 

the agent is actually making decisions on its own, not just following a rigid set of rules. Down 

below, let's send the message. Back within Make, let's try running this again. Here within Slack, 

I can see that there's a new message within the customer escalation channel. And here it calls out 

the customer request for a custom cookie order. Now someone in the team can follow up with the 

customer. This is exactly what I expected. Back within Make over on the left-hand side, let's 

click back into scenarios. And here we can see that all three of these different scenarios are 

active. Our agent will read and understand all incoming emails, and then depending on the type 

of email, it will either reply directly to the customer or it'll escalate to Slack where someone 

on the team can follow up. Now let's take a step back and look at everything we've built inside 

Make. Over on the left-hand side, let's click into something brand new called the Make grid. Here we 

can see the whole customer support workflow that we just built. Right over here, the email comes 

in. Then the AI agent reads the email and decides what to do. And then it either sends an email back 

to the customer with the answer from the FAQ or right up here it escalates to Slack. The best part 

is that as we add more automations, like let's say for example, order tracking, refunds, or customer 

surveys, the grid keeps everything visible in one place. You can see the common data sources across 

multiple scenarios. Now let's say for example, we built another workflow using this same exact 

agent for tracking refunds, we'll be able to see that right here. The grid makes it easy to 

spot the common dependencies and understand how everything fits together. That way, nothing 

gets lost, and you always know exactly how your automations are connected. And that's it. We just 

built our very first AI agent in Make to handle customer support for the Kevin Cookie Company. 

It automatically responds to common questions, escalates the tricky ones to Slack, and we can see 

the whole process clearly inside the Make grid. Now, this is just one example, but now you have 

all the essentials to start creating your own AI agents for whatever tasks you need. Again, if you 

want to try this out, use the link down below for a free month to a Make Pro plan. Please consider 

subscribing and I'll see you in the next video.