How to Build an Outlook Inbox Manager in n8n (Step-by-Step, No Code)
In this tutorial Dr. Violet Howard demonstrates how to create a fully automated Outlook inbox manager using n8n’s no‑code workflow engine and a routing‑agent architecture. The video walks through every stage: configuring an Outlook "On Message Received" trigger, switching the output to raw to capture the full email (including HTML), cleaning the email body with an OpenAI message‑model node, and classifying incoming messages into three custom categories (high‑priority, billing, promotion) using a Text Classifier node powered by a free Flash model. Based on the classification, the workflow routes the email down distinct branches: moving the message to a corresponding Outlook folder, sending a Telegram notification, creating a draft reply for billing inquiries, or automatically replying with a polite decline for promotional offers. The tutorial covers credential setup for Outlook, OpenAI, and Telegram, dynamic field mapping (message ID, sender, subject, timestamps), prompt engineering for system and user messages, tool configuration (Outlook "Create Draft", "Send Email"), testing each branch, and finally activating the workflow for continuous polling. Throughout, the presenter offers practical tips, troubleshooting notes, and references to a paid community where templates and deeper agent‑building resources are shared.
Chapters
Today we're going to be walking through step by step how to build an inbox manager for Outlook. Yep, you heard me right, not Google. Today we're looking at Outlook. By the end of this video, you'll have an inbox manager up and running in your own n8n instance, and I think you'll be surprised how quick and easy it is to set up. So let's not waste any time and get straight into the tutorial. Okay, so today we're going to be using n8n to build an inbox manager, and we're going to be doing this using a routing agent architecture. So we're going to have one agent up front that's going to take our emails from our inbox, and then it's going to route off to a different logic path based on what type of email it is. So high priority, customer support, promotion, finance, and billing. In this example today, we'll just set up 3 different paths, but this is where you're gonna be able to really customize this thing based on, you know, the different branches you want. But the cool thing is we can customize what happens based on each email coming through. So we could either have the agent send us a message, we could have the agent respond to the message, we could have the agent create a draft, you know, whatever it is, we're able to get in there and customize it. And so whether you've been building with N8N for a while now, or this is gonna be one of your first builds, don't worry, we're gonna break it down, make it as simple as possible. So let's hop into it. Okay, so the first thing we need to set up is the trigger of this workflow. So what has to happen for this workflow to execute? We're gonna click on Add First Step. I'm gonna type in Outlook, and I'm gonna grab an Outlook trigger, which is On Message Received. So essentially, once we connect our Outlook account, every time we get a new email, this workflow is gonna fire off, and then we decide what will happen based on the type of email that comes through. So as you can see, I've already connected my credential, but you'll have to come in here if you haven't click Create New Credential, and it's super simple. All you have to do is click on Connect My Account. You'll sign in, you'll have to allow access for N8N to read, write, view, whatever it is, your inbox. And then once you allow that, you'll hit Save, it'll go green, it'll say, um, you know, account has been connected successfully. You'll get an email in your inbox that says a new app has access to your data, N8N connected to your account, and then you'll be good to go. So within this trigger configuration, there's a few things we can look at. The first one is a poll time. So this is looking every minute, it's going to search your inbox and see if you got new messages. You could obviously change the frequency. I'm going to leave mine at every minute. Then we have trigger on. So this is looking for a specific operation. Right now it's easy, all we have to do is look for a message received. And then we have an output option which is simplified or raw, or we could choose selected fields. What we're going to do is hit fetch test event and we're going to see an email come through, which is an example email that I sent myself earlier just to test out some stuff. And as you can see, this is simplified. So what we're getting is a body preview. You can notice right here that the email gets cut off. So this is probably not ideal when we're feeding it into our routing agent because we want the agent to be able to read the whole message and then interpret which way to send it. But you can see there's also other information we're going to get, like the subject of the email, the email ID, which will be important for later, the conversation ID, who it's from, all this kind of stuff. So I'm going to change the output to raw. Hit Fetch Test Event again, it's gonna pull in that same email, but now we have a lot more information coming back about that email. We have the received time, the sent time, attachments, importance, we have parent folder, we have all this kind of stuff, right? And here's the actual full email that we're getting within the body content. As you can see, we have some kind of nasty HTML tagging going on here. So the agent can interpret this, but just for the sake of this video, I'm gonna clean this up, before we send it into our routing agent, just so the video makes more sense. But keep in mind, you don't have to. The agent will be able to read this and understand the content of the message, which is really all it's looking for, right? But anyways, so that, that's why we want to do this raw rather than simplified. You can also filter other things like you can only pull in messages that are not, that are unread or have certain attachments or within a certain folder. But for now we're just doing a general inbox manager, so we want everything. We're going to leave the filter and the options as default. So from there, like I said, we're going to try to clean up that, that text coming through before we send it off, just so it makes this tutorial easier to follow. I'm going to type in OpenAI, and I'm going to grab a message a model node. All we're going to do here is connect to OpenAI. Same thing, if you haven't created this credential, you'll create an OpenAI account. You'll have to load in like $5, and then you'll just hit like generate API key. If you need help, you can always click on open docs and it will walk you through. But once you have that API key from OpenAI, all you have to do is plug that in right here and then hit save. You'll go green. You're good to go. So I'm going to connect a model real quick, which is going to be 4o Mini. And then we have to set up a user message real quick and a system message. The system message is basically like the coding that goes on behind the scenes in ChatGPT where it says like, you know, this is your role. But the user messages, every time you talk to ChatGPT in that little text bar at the bottom, what you're saying. So what we're saying here is like, here is an incoming email, and all I'm going to do is give it the nasty body content of the email right here that has HTML tagging, and we don't want to read that. And then the system message, I'm just going to say like, this is me telling this node what to do, as if you were telling an intern to take these emails and do something with them. So I'm just going to say, take the incoming email and clean it up so it is more readable. What I mean by that is get rid of the HTML tagging, but don't get rid of any of the email content because we want to make sure that it's not changing stuff around, right? So now that it has its system message and instructions and it has the dynamically changing email that's coming in every different time, we'll hit Test Step. And on the right, we'll see that it's cleaning up the email, and now we're getting just the email. Although one thing it did is it output a subject. We don't need that. Lisa will come in here to the system prompt and say don't include a subject. So now I'll send that off again. It should have gotten rid of the subject on the right-hand side. We should only be getting the actual body content back, and now we're good to go. So I'm just going to change this node and say clean email just so we know what it's doing. And next we're going to move into the routing aspect. So here I'm going to click on this plus. I'm going to type in text classifier because this is an AI node that we're going to be saying, okay, you're going to be getting this text every time and what you have to do is classify it into one of these 3 paths and then send it down that correct path. So first things first is what text are you actually going to be classifying? As you can see right now, it's set to fixed. We want to change this to an expression because it's going to be looking at something different every time. And what we can do is we can just drag in the output of that clean email node and we just can put in right here. So it's referencing json.message.content, but what it's actually pulling in is the email that we got. So This is the variable, this is an expression, but this is the actual content that we're going to be having this text classifier node look at every time. Okay, cool. So now that it knows what to look through, we have to give it different categories. So like I said, we're going to be doing 3. We're going to be doing high priority, um, billing, and promotion. So we have 3 categories, and now what you can see is that we now have 3 paths leaving this text classifier node. So this would have been just one, but now once it sends it down one of these 3 circles, we can drag out logic different ways to do different things based on the type of email. So before we can do that though, we have to give a description so the AI knows what type of email goes into which type of category. So first we're going to do high priority. So I'm gonna paste in this description here. I'm going to, it's going to say emails that require immediate action, often involving urgent issues, escalations, system failures, or critical business matters. And then what can be really helpful is giving it some common words or phrases that are in here that may be found within a high priority email. So one thing to keep in mind here is, you know, you may not be setting up the same categories as I am. So when you're setting up your categories, you'll want to test different emails to go through, obviously. And if you see one not being labeled correctly as it should, then what you want to do is come back into here and give it some more examples of, you know, let's say you wrote an email and it didn't get labeled correctly. You could just put in this, in this prompt, like, examples of high-priority emails and give it some examples of, like, ones that you would typically get. The more examples and the more detailed your description for each category is, the more accurate and consistent it's going to be. And by the way, if you're enjoying this step-by-step style and the hands-on approach, then definitely check out my paid community. The link for that's down in the description. We've got a great community with members that are always, you know, asking questions about N8N. They're very dedicated to learning, sharing resources. Got a great classroom section with deep dive topics like building agents, vector databases, APIs, HTTP requests, and then of course deep dive on step-by-step builds where I'm doing step-by-steps of the videos I show on YouTube. So a great place to learn, and then also a calendar with 5 live calls per week. And then I've also got a free community. The link for that's also down in the description, and this is where you can come into YouTube resources and you can get the templates for free for all my videos. So this template you can find, you'll click on the post associated with this video, and then you'll be able to get the template right here and download that, and then you can import that straight into your N8N environment. So anyways, let's get back to the video and set up a description for billing. So what I said here was emails related to payments, invoices, subscriptions, financial transactions, or account balances. These emails often include due dates, payment instructions, or financial statements. And then once again, we're giving it some common words and phrases. And then finally, let's do promotion. So here we said emails related to marketing campaigns, sales offers, discounts, partnership opportunities, or ads. These emails often sent in bulk and contain promotional language, and then once again, common words and phrases that may be found in a promotional email. So now that we have all of these email categories set up and we have the text to classify also set up, we can hit test step, and it's going to fail here. And this is because once again, we're using AI to read through this email and determine which path to send it down. So in order to use AI, we have to set up some sort of AI brain. So I'm going to come in here and grab a Flash 2.0 model. I'm doing this because it's pretty powerful, it's, it's free, and it just seems like a great use case of categorizing these emails, especially if we're getting a ton. Don't want to be wasting credits on something that Flash can take care of with a free model. So now if I hit play, we'll see that it's going to output this item as high priority because we have one item down this branch. And if we click into it, we can see high priority, one item, billing, zero items, promotion, zero items. So we know that this is working as it should. Cool. So next, what we want to do is we are going to configure the logic for each type of email that's coming through. So we're gonna start off with high priority because we have a high priority email here. And the first thing that I want to do is, okay, if I know I have a high priority email, I want Outlook to move it into my folder in Outlook called High Priority, as you can see down here. Now, in this example, we're doing high priority billing and promotion, and so you have to create these folders in Outlook before you can set up this automation. So under these three dots, click on Create New Folder, and then you can set up the different paths you want. So as you can see, I have them already set up. So back in N8N, I'm gonna click on the plus after the High Priority branch. I'm gonna type Outlook, and I'm gonna go down to Message Actions and click on Move a Message. Okay, so this node is going to be moving this message into the High Priority folder. So that's what I'm gonna call the node to keep ourselves organized. You can see the resource, you can see the operation, and now we can see we're going to choose the folder, which as you can see, a dropdown, we're choosing high priority. And now we have to choose the actual message to move. So you can choose from a list, which will just have all of the emails in your email. But what we need to do is we need to choose by ID so that we can dynamically move in whatever email triggered this whole workflow. We can select that one and move it into the folder. So for the sake of this video, I'm going to send off a new email to my Outlook account that's called test. I'm going to hit send, and then we're going to move this one into a high priority folder. Okay, so I just refreshed my email. You can see we have the test email, which is this high priority message right here. And I'm going to go back into the workflow. I'm going to hit test workflow, and it's not going to work because this, this node is erroring beforehand because we haven't set anything up. So I'm just going to disconnect it real quick. Now I'm going to hit test workflow. It's going to pull in that new email it just got. It's going to clean it, classify it as high priority. And now let's link it back up here. And then once we're in this node, what we want to do is once again, we want to change this to an expression, not fixed, because it's going to be changing every time. We don't care about the text classifier node right now. We don't care about cleaning email. All we want to look in is the trigger, and we want to grab the message ID of the message that we got that fired off this workflow. So now every time that it gets down to one of these nodes that moves a message into a folder, it'll be pulling the actual message. So all I want to do now is hit test step. We will see that we get a success message. If I go back into my Outlook, we can see that that message just disappeared and it is now showing up in the high priority folder within our Outlook. So that's how that works. Okay, cool. Back into Outlook, this node's taken care of, and now it's just about figuring out what logic do you want to happen after you get a high priority email. In this case, let's just say we want to get notified. We don't want the workflow to draft an email or to send anything off. We just want to get a notification. So I'm going to add a Telegram node and I'm going to go down to send a text message. You'll have to connect your Telegram account. You'll click on Create New Credential. You can open docs and it will walk you through how to set up your Botfather and get your API access token, which you'll put in right here. Super simple. So follow that documentation if you don't already have this credential set up. And then all we have to do is send a message. So you have to get your chat ID. So the best way to do this is by coming into N8N, opening up Telegram, and open up a trigger, which is On Message Received. And then all you have to do is hit Fetch Test Event, message yourself in Telegram. You'll send a message in here, and then when it comes through, you'll see a message ID and you'll see the text coming through. So you'll just grab that message ID and then you can just plug that straight into here. So I'm going to grab my message ID. So I just saved that in a Word doc for now. So I have my message ID in there, or sorry, my chat ID. And now I want to also come in here and turn off the append and an attribution. So that when we get a message, it doesn't say this was sent by Telegram or by N8N, sorry. And then we just have to configure the text. So this is probably going to be dynamic, right? So I'm going to click on expression and I'm actually going to open this up full screen so we can look at what we want to do here. We want to close out everything because really all the information we need is going to be coming from the Microsoft Outlook trigger. So our message is going to be a mix of expressions and things that are variable and also things that we're kind of hard coding into here. So I'm just gonna say like, you know, new high priority email. So we're able to hardcode in new high priority email because obviously it's only gonna get to this note if it was labeled as high priority. Anyways, what we want to do is just say like who it was from. So once again, we're gonna get down to the Microsoft Outlook trigger and just look for the information we want. So let's come down here to the sender. So it's gonna be from, and then we can just drag in the sender name. We can say like, what else do we want to add into this? We can add maybe like, you know, the subjects, just in case they want a little snippet of what the email is about without having to actually get in there yet and look at it. So we'll give the subject, and then we can also say like, you know, maybe what time it was received. So date, and then we'll just say received date right here. And this is kind of a messy format, so what we actually can do is is, you know, this is basically going to be triggered immediately as soon as the email comes through. So I'm just going to basically come in here and do two curly braces. I'm gonna type in now, I'm gonna do.format, and then we're just gonna leave it like that. So it's just going to be getting the who, who sent an email, what the subject was, and then what the date was that you got it. Obviously all this kind of stuff can be customized, but that's what we're gonna leave it as for now. And then we will hit test step. I'll open up my Telegram and you can see this is what the email look— or this is what the text message looks like. New high priority email from Nate. Here's the subject and here's the date. So obviously there's more stuff that you could do, but in this case we were just going to leave it as simple as that for a high priority email. Okay, now we're gonna be taking a look at the billing branch. So what I did is I sent myself an email. As you can see, we have a payment confirmation, we have a billing email right here in the inbox, and what we're gonna do is set up the logic over here. So I'm gonna hit play on this Outlook trigger. Let's just make sure we're getting the correct email. We can see the subject is the one I just looked at. We'll have the AI clean it real quick, then we'll have the AI classify it, and then we'll set up the logic for the rest of the billing flow. So it correctly labeled this one as a billing email, and all we want to do first is put it into the billing folder. So instead of actually adding an Outlook node, I'm just going to copy this one, paste it down here. We're going to click into it, obviously just change this to, um, you know, billing folder. We're going to change the folder that it's moving to from Higher Priority to Billing. And the reason we're doing this is because we already have the message ID of the original message coming through mapped. So as soon as I drag this in, it now has that connection. Now you can see this is green and we have the message ID correctly. We'll hit Test Step. We will get a success message, go into our Outlook, and we can see if we open up the Billing folder, that message just got put into there. So Saved us a lot of time. We already have that configured. Now it's just a matter of, okay, now that we have that billing email in the right folder, what do we want to do? In this case, like I said, you can customize it however you want, but in this case, let's say that we're going to add an AI agent that is going to respond to this email as a draft. So I am going to change this to billing agent. We're going to rename it there. We're going to change the user message. Right now it's looking for connected chat trigger node. Which just means it would have been looking for this thing right here that allows us to chat in our interface of Any10N. And because we don't have that, we're not using that as the user message. We are going to change this thing. So same way we set up like the OpenAI node with a user message and a system prompt. That's exactly what's going on here. Here's the system prompt. Here's the user message. And what we want to give it as context is the output of the clean email node. So I'm going to change this to an expression. I'm going to say, here is the email. And I'm just going to drag in the content of the clean email right here. So as you can see, that is what it's getting to respond to. Cool. So now we just want to set up the system message. So I'm going to tell it, you know, here's an overview. You are a— I'm just going to say billing assistant, whatever that means. Your job is to respond to the incoming emails in a professional manner. So we're just wanting to define sort of like its job, its role, its end goal, and you know how it works. So what we're going to do is we're going to give it access to a tool to send to create drafts. First thing, I want to give it a chat model. So in this case, because we already used Flash 4.0, Flash 2.0, I'm just going to give it 4.0 mini over here and we're going to add a tool. I'm going to type in Outlook. We're going to rename this tool Create Draft. And we have the operation send message. That's not what we want to do. We want to change this to a draft and we have operation now create draft. So that's perfect. We have, we'll have, we have a couple of things to fill out. So we have the subject of the draft we're making and the message of the draft we're making. And all I'm going to do is click on this button that lets the model define this parameter, which is basically just saying, okay, the agent, you know, this main agent up here, I'm going to get this email. I'm going to read through my system prompts, understand I need to respond to it by creating a draft, and then I'm going to send it to this tool where I will automatically fill in the subject, I'll automatically fill in what the message is going to be, and anything else I need to configure. So in this case, we're creating a draft and we want to be replying to, um, the original sender that sent us the email in the first place. So Outlook, this is where it gets a little bit different from Gmail. In Gmail, when you reply, you can send up a reply ID so it automatically responds in that thread. That's not how it works in Outlook, and you may think it's the reply to, but it's not. This just puts a reply to email, as in like when you send an email, um, usually there's like a reply to like this email, you know what I mean? So that's not what it is. What we want to do is just add a to. So unfortunately it won't put it in the same thread, but we can have it automatically send to the person who sent us this email that triggered the workflow. So I'm going to close out of Text Classifier. We want to get back to the trigger where we can get this information. So every time we get an email, right here under sender, we'll have the email address of the person who sent us the email. So that's what we're going to map in here. So that is pretty much set up. I just have to change this to, um, the new credential. I was using the wrong credential. And yeah, I think that's it. We will basically test this thing out now. So actually, first we want to give it a system prompt about its tool. So I'm going to say tools Let me make this bigger so you guys can see it better. So tools, I'm going to say create draft, which is use this tool to send an email in response to the one you receive. So maybe that language isn't optimal. We'll see. We're going to hit play on this agent and we'll see it think about it using its brain. And then it should hit this tool to create the draft. There we go. Once it's done, it's going to tell us you know, here's the response. Let's actually go into Outlook. We will click on drafts. So this one was a test from earlier. Here's the draft that it just made. As you can see, we have the re— so it isn't— it is a, um, it actually looks like it is going to be put in that thread. We have the emails responding to, and then we have the message ID. Actually, no, so it's not actually— well, maybe it is because The, the subject was generated by AI, right? But if we click into the inbox manager, all it's getting is this email. Okay, so I don't think it's really responding in that thread. I mean, we can easily just find out by hitting send. So here's the inbox that it came to. It was sent as an individual email, so it didn't come in the same thread even though it has re. Um, that was just the AI being smart enough to make a subject that was in response. Anyways, That's— we're getting off topic here anyways. So that's how it works, right? But maybe we still want to get notified that this happened. So in that case, what we can do is we'll copy in this Telegram message, we'll hook it up because we already have the chat ID configured. But now what we want to do is just change the message. So all we can do is say like, you know, new, new billing, you know, inquiry. I don't know why it's capital. We can just say new Billing Inquiry Received at, and then we can put the time by doing $now. We can format it so we can get the time. And so instead of looking for year, year, year, month, month, I think if I do hh:mm, there we go. And then how do I get like AM? Okay, there we go. So we have 4:48 PM. So now it's saying like, you know, you got this new billing inquiry at this time. We'll say a draft has been created and is sitting in your Outlook. Although maybe let's say like who it was from. So we'll say from, once again, we'll close out of all the stuff. We want to get down to the Outlook trigger. And at the bottom, we remember we can get the from address, but we can also get the from name. So I'm gonna drag in the from name right here. So, new billing inquiry received at 4:48 PM from Nate Herkelman. A draft has been created and sitting in your Outlook. So that's what we want every time we have this process taking place. We can see we got that message in Telegram, and that is the end of the process for billing email. So, that was high priority, that was billing. Let's set up the final logic for a promotional email. Also, one more thing I wanted to point out is if you didn't like the way the email was structured, if you wanted to change like, you know, the sign-off or any of the formatting, obviously you could do that within the system prompt of this agent. You could give it like rules on how to output, how to format an email, and then also you could do some of that stuff in here. So in the message description, you could say like, you know, sign off as, you know, Bob from ABC Corp or something like that. You could also give it rules on how to make a subject. So that's where you can add these little descriptions to make this stuff even more customized for how you like it. But anyways, what we're gonna look at is I just sent an email to myself in here. So as you can see, it's pretty promotional. Boost your workflow. There's not even a sign-in or an introduction. Hey Nate, 20% off, all this kind of stuff, right? So we're going to go back into our Outlook Inbox Manager, hit play over here, make sure we're getting the correct email, which should be 20% off. Perfect. And now we're going to click clean email so it creates human-readable. We're going to have this text classifier node, as you can see, it outputted as promotion. So if we click into here, promotion branch, one item. And now we're just going to set up the logic for promotion. So the first thing I'm going to do is copy in this billing folder because we know that we just want to move it to a promotional folder. So I'm going to come into the folder section and change this to promotion. And that's all we have to do besides change the name real quick. So promotion folder, we will drag in this item down here because now we will be getting the correct mapping of the ID. Perfect. And now what do we want to do for a promotional email? Maybe we just, we don't care if it's a draft because it's promotion. We can just respond every time with something like, not interested, but thank you. So once again, we're going to add an AI agent to do this. We are going to, um, connect it first. Oh, it is connected. It's because it has to look for, oh, we didn't run this. Okay. So we're going to run this previous one. So with the, So then the agent actually has data to look at. First of all, we can see that this is true. We'll go into our inbox and make sure that it got moved to promotion. As you can see, there's that email we were just looking at in promotion now. Anyways, back into here, let's look at what the agent's going to do, which is we have to first define the user message. So once again, we're going to do the same thing pretty much. Here is the incoming email. We're going to drag it from the clean content node. So now it's getting the email. And we're gonna add a system message that basically just says, um, you will receive promotional emails. Your job is to politely decline all offers. So should be as simple as that. Use the send email tool to do so. Okay, so that's a really, really simple system prompt. We're gonna keep it at that. I'm going to connect it to the same chat model because you actually can have two agents using the same model. It's no issue. You don't have to. It's just, you know, if it, if you think you don't like the aesthetic of that, but it keeps things simple, right? So then we're going to add the tool, which is going to be an Outlook tool. We're going to be sending a message this time. We're going to change this to send email because that's how I just defined it in the agent system prompt. And now we have the to, the subject, and the message. So once again, I have to go to the bottom down here and looking in the Microsoft Outlook trigger. For the message. So there's the email address— or sorry, not the message, the email address. So I drag that in. Next, we can have the subject and the message be automatically determined, or we could have this be fixed. So let's say in the case we don't care if the subject is dynamic every time, we'll just have the subject say not interested. Okay, so the subject now says not interested. The message will be changing every time a little bit. And yeah, I think we're good to go here. We can maybe just say at the end, like, just to show you guys how this works, we'll say sign off. Um, sorry, my internet's a little laggy. Sign off all emails as Frank. Okay, so now the, the message is looking at this description when it's deciding how to make its message. It's going to know to sign off all emails as Frank. Um, cool. So let's hit play on this agent. It's going to think about it. It's going to send an email. And then we should get a success message. So I politely declined the offer from AI Tech Solutions team. We'll come into our Outlook. We can click on Sent Items, and here's the old one. Let's give it a refresh. We will see— oh, I know what I did. I didn't connect the right credential. So in my Outlook, I was using the wrong credential, as you can see. So now I'm going to change it to the right one. I'm going to run it again. It's going to do the exact same thing. There we go. We just got the same exact message. The email's been sent. We'll come in here. Now we have that not interested because we set the subject as not interested, and then we have it signing off as Frank. So everything here is working as it should, and maybe we don't even want a notification about promotional because we just want less noise as possible. So that is this inbox manager. We clean the email. Like I said, you don't need this step because the AI is able to read that HTML stuff. So you could easily get rid of this. Hey guys, one quick thing. While I was editing this video, I just noticed if you are to remove— like, if you follow the step-by-step and you remove that Clean Email node, just be careful because your agents that are looking at the original email, we referenced that Clean Email node output. So if you do delete that node, you're going to have to remap some stuff later on. You could also just remap the email that comes through from the Main trigger instead of from the Clean Email node, or you could remap that from the Text Classifier node. Either way, just, you know, I noticed if you are to delete that node and you were following this step by step exactly how I did it, your automation will break. So just wanted to let you guys know that you just have to remap some stuff, but it should be good to go. Anyways, back to the video. But what's important here is that we have the trigger coming in, we have the Text Classifier sending it through these different branches, and only that logic will take place. And then what you want to do after this is built out and you like the way it's working— and also just because I don't want to get annoyed, I want to change this to a promotion agent. Um, now what you need to do is click on Activate. So you'll activate this workflow. It's going to say your workflow will now regularly check Microsoft Outlook for events and trigger executions for them. These executions will not show up immediately in the editor, but you can see them in execution list. So we'll click on Got It. What that's saying is, you know, we saw the very visual breakdown of which way the logic was going. If you sit here and you send yourself an email to test, you're not gonna see it live, you're not gonna see it happen. But when you go into executions, you will see like, you'll see a yellow line right here that will say running, and then it will show. And then once it's finished, you can click into it, and then you'll see, then you will see this visual breakdown like this. But it's just all gonna be happening in the backend. But so if you activate it, as we know, it'll be looking in your Outlook every minute or however often you set this up. And, and then you have, you know, an inbox manager. So hopefully that was super simple to set up. Um, I know I only did 3 paths down here, but you can get in there, add as many categories as you want, and customize the logic as much as you want for your agents. It can also get really cool if you have like specific agents. So let's say you had a customer support section perhaps. You could have an agent, you could like give them access to like a Supabase vector store or even just like a Google Doc where you could have them read a Google Doc of customer support policies and rules in order to write that response. So these can get really, really tailored and it can really, really save you a lot of time. You know, the more and more you build out the details of what this stuff should be doing and the different categories. So that's going to be it for this one. If you guys enjoyed this sort of step-by-step style, definitely check out the paid community. Link for that will be down in the description. I do a ton of step-by-step builds in there, and that whole community is a ton of people dedicated towards building AI agents and learning, sharing resources, um, all that kind of stuff. So if you enjoyed the video, please give it a like. Definitely helps me out. I really appreciate you guys making it to the end of this video, and I really appreciate all the support as of lately. And I will see you guys in the next one. Thanks.
