Build Your First Multi-Channel Customer Support Ai Agent using Make

1. Introduction to Msterclass: Hi. What if you could build a 24 by seven customer support team that never sleeps, never gets tired and work across every channel your customer use, like Telegram, chatbot, email, even WhatsApp. Hi, I'm Saifulla. I'm professional app developer and AI trainer. I have spent years bridging the gap between complex code and practical AI solutions. My goal is to help you master these AI tools without getting lost in the AI era, so you can build systems apps without any technical background. Businesses today are overwhelmed. They are losing leads in their inbox and ignoring customers on social media in which they can lose so many sales and which is very problem to founders and the companies. In this class, I'm going to show how you can build a multi channel ecosystem in which you can automate all the customer support queries in one place. That is med.com. This class, we are building a custom website, chatbot workflow that answers and escalates to Women sales team when needed. Telegram AI Agent, form submission AI agent, feedback collection AI agent, and advanced email handling AI agent. We also discuss how you can connect these five different type of AI agents built in this web.com to our e commerce website like this. You can add a customer support chat EI agent here. Okay. In the contact, this is a form submission AI agent. So this is the email handling AI Agent and feedback collection EI agent as well. And not only that, we will also discuss some moneymaking ways in which you can start selling these type of services to clients in the online. I will also provide this troubleshooting guide in which it will help you to solve any issues when you run the AI agentsnm.com. Not only that, you have final project, building AI support agents for WhatsApp and Slack channels using make.com. So by the end of this course, you can build any type of AIgen by using make.com. You need to complete this particular project. I will provide all the step by step process how you can create this particular WhatsApp and Slack channel customer support AI agents in order to complete this particular project in which you can gain the experience. By the end of this class, you will have a fully functional multichannel I customer support system, which is ready to deploy. No complex coding required, logic, creativity, and the power of meg.com, whether you are a freelancer or a business owner or a curious creator who are looking to learn the building AI agents in this AI era, you can join this class to build the real world skills that will set you apart in 2026 or beyond that. I will meet you in the first session. Let's dive into that. 2. 1.1 Basics of an Ai Agent: In this session, we are going to discuss some basics of EI agent like what is an EI agent and some differences between EI workflow and EI agent, and we also discuss some block diagram what actually EI agent do. Okay. Now let's start from the definition of AI agent. What is an actually EI agent is? You can see here an AI agent is an autonomous system that can look at what's happening around it, understand the information it receives and make a decision on its own to complete task or solve problems. That is very easy to understand, right? So it is basically a autonomous system that can look at what's happening around it. So according to our instructions, the AI agent will do the task according to the input receive and it will make decisions to complete that particular task. Now you can see, unlike regular workflows that only follows instructions exactly an AI agent can learn, adapt and choose actions to reach its goal without needing constant human guidance can easily try our AI agent according to our instructions, our own data, to learn, to adapt and choose actions according to its goals without needing constant human guidance, right? So we will discuss, so what is the actual AI workflow and AI Agent differences between in few seconds in that you can understand very easily. You can see here. And human guidance, it's like a digital assistant that works independently to help people with different jobs and challenges. Next, that is differences between AI workflow and AI agent. So you can see a structure, predefined sequential task or steps. So basically, this workflow is step by step process, we connect the different apps to do according to our logic. So when you come to AI agen, it is dynamic. We can change the instructions. Instead of changing the whole workflow, we can change the instructions in AIgen in which this AI agent plan its own actions to reach goals. That is simple. These are the steps. You can see the differentia. This is a AI workflow with start and it will go step by step to complete the goal, right. But in the AI agen, we can do a lot more things with EIgen instead of just connecting different apps to do one particular goal, to reach one particular goal, we can create an AI agent in which it will run according to our instructions input, it will choose the actions tools and the data we will provide to that particular AI agent. It will do the work to complete that particular task. You can see here AI agent, I choose actions smartly according to our instructions and the tools will connect to that particular AI agent. Okay, it will adapt to not to do the mistake similarly like that. Is it will take feedback, and again, it will choose the best action. Okay, the AI agent Smart ness is based upon the model we will choose OpenAI reasoning module or like that. And the instructions, we will write as a system prompt for the AI agent. That is simple. Now you can see the next autonomy. Low follows fixed rules paths. The AI workflow is just simple start to end step by step flow. There is no reasoning, but it a simple logic, we will just create AI workflow to done one particular task without any thinking without any reasoning. In the AI workflow, it will just work according to the step by step, but the AI agent is different. It will think on its own data. Okay, I will observe, it will adapt, it will learn from the mistakes. I will adjust the output, and it will choose the actions very smartly. Okay, like that. You can see the AI Agent hi makes independent decisions of the initial prog. Okay. When the input comes, it will take the decisions. Okay, what to do next. According to the, it will take action. But in the AI Verbit is step by step. Okay, it is dependent on the second step, third, step to reach the particular goal, but it is not according to the input, it will make the decisions. It will choose actions smartly and it will do the particular task. That is simple. Adaptability. You can see limited handless expected variations needs update for new scenarios. If you want to update some chips in this AI workflow, we need to change the each and every step according to that workflow and the application. But in the AI chain, we can easily adapt handles new or unexpected situations. It will handle unexpected situations very well because it is a thinking model. It is not a fixed logic. I hope understand these points. For the decision making, it is embedded in the workflow, logic, step by step. You can see it is a step by step. It will work in the step by step process only when it will just it will go here. It will run just go here to reach the end goal, right? For the AII chain, it is internalized agent reasons and chooses actions in real time. According to the input from the user or any data, it will just choose the actions in real time, right. Like that. So we will understand what are the actions, what are the tools, what are the context, all those things in upcoming sessions. And for the memory and learning usually has no memory or learning between runs. As I said, it is a step by step process. We will just connect the different apps to reach a one particular goal. There is no memory in that there is no learning between runs like that, but in the AIA chain, we can also add a memory in which it will store our previous charts, previous data like the charge bit two. Then we can also add a memory to or AI agent in which it will give the personalized answers to particular user and it will give the smart answers. It will choose smart actions. It will also learn and adapt bytesel. For the AI agent, it can retain memory learn and improve over time. But the transparency, easy to audit each step. But for the Agent, full reasoning may be hard to trace. It is easy because it is a simple logic, we have been. But when you come to AIgent, sometimes we need to go in depth, how the Agent is working, what is the output, what are the instructions, we need to try, we need to test with particular AI agent in different scenarios in different ways to get the best output from the AI gen, which is some time taking process then this workflow. In the control, you can see for the AI flow controlled externally via workflow platform. For the Agent, control is internal to the Agent needs special Gadiils for management. This is a difference between AI Wflow and EIgen. I hope you understand these points. Basically, this AI Agent is an autonomous system which works according to our instructions and the tools we will connect to this particular AI agen, it will choose the actions or tools to complete that particular goal. We will discuss how you can create the AI agent in the make.com. Jumping into our next session, please make sure you follow tutorials step by step. Without any interruption. The backstep practice is just follow tutorials in your mobile phone and do the practical steps I will show from the next session, you can do exact things in our laptop, in our make.com account in which you can learn more faster than just simply seeing and learning in theoretical way. Let's dive into next session in which we will start the basics of make.com. Let's dive into that. 3. 1.2 Getting Started with Make.com Platform: Let's explore make.com platform. Come here, search for the make.com, you will get the link and just click here you will land here. This is the official platform that is make.com. Now you can find so many details about this platform, what they are for. Okay, you can check it out, all those things. So what is the features that make.com? We will provide to us like automation with AI, agentic automation, make AI agents, MCP server, make EI apps like that, you can see the solutions. It is work across all different aspects of automation like marketing sales, operation, customer service, resources, you can find the blogs, success stories, how to guide template library. And you can also search Make academy in which you can learn how to build something using this meg.com. You can find the details about all the things here. You can see you can explore so much things about all the may.com here, you can create the free account here. Just come here, get started free. Just click here, get started free. You can sign up for free account, and I have already an account. Another thing is, if you got any issue in creating this account, it is very easy. You can check it in the YouTube, all those things. And not only that, you can check the pricing, you can select any and start with the free account to test on this make platform. Otherwise, there is a one protip. You can go to YouTube search for the M pro account for free. You will get any videos they will provide the link. Just click here, you will get the free account for one month, that is P. That is you can get the 10,000 credits per month. You can see for 10,000 credits per month, you can just try out this Make platform to automate all your things. Just go and search for in the YouTube, you will get all the videos to help you in that to get the pro account per month for free one time. I hope understand this. Now let's come I already have an account you can see so this is a dashboard. After successful creation account, you will land here that is Organization dashboard. You can see how many credits you have left in your account and you can see all those things here. This is a pro account I have got. You can see you can add that teams users. And you can explore by yourself for more understanding purpose to use efficiently this platform in our workflows or building AI agents, EIAds like that. You can see the subscription Attils as well, you can see how many credits you use there, and in the left hand comer, you can see you can build the scenarios. This is the main part of our course is. You can build an AI Agent that is in the bit of click here scenarios. Though you can land here, you need to just click here open scenario Builder. You can come here, AI Agent, you can create Agent I agent from here, and this AI agent will be available in the scenarios in which we can connect our AI agent with the other apps to connect to automate our Dali task. Okay, I understand these points. Coming to the connections in the connection, when you try to connect to our apps like Google Sheets, Google Gmail, and different Telegram, Slack, sendsk for customer support handling, when you create all those things with the Make platform, they will show up here. Okay, even it can be open A, all those things, it will show here that we will see all those things in upcoming sessions. That is Webos Webooks is very most important thing. So this will help you to receive or to send the data outside of this platform. Otherwise, it can to call the webhook for different scenarios. In detail, we will see all those things when creating the scenario for our AI Agent. Okay. Now com in to the templates. We have different templates. You can try out by yourself to understand this platform, how it works, how the workflows works. Like you can see, there are a lot more templates you can explore, you can use in the platform in this platform and to automate inward daily task and just click here. Now, you can just go with the start guided setup or create a new scenario from the template. It will directly import this whole template inward scenario in which you can try it by yourself and you can check all those things. Now, other part is you can click in the three dots that is more. You can find the data stores in which you can store your data, right? It can be your previous chat or AI Agent logs, all these things you can store in this data store. Don't worry, we will also add these data stores in our AI Agent workflows in upcoming sessions to easily track our users data name or all those things. In the keys, you can add the EPA keys, you a private or public keys in which you have the control our keys in which you can use in different workflows, web hooks, all those things. They will show up here. Okay. Now, come to the device where you can access this make into mobile apps as well the data structures. These are some different things. These are the more advanced, you can add your data structure from the different extensions like formats like JS and XML, CSV. In the custom maps you can build custom app in which you can integrate this custom map in workflow. In this course, we are not going to see these custom maps because these are not related to our AIHen. If you are looking to learn more about this, you can use that documentation, otherwise, you can follow in the YouTube aspet. Not only that you can come here right side, you can check the view profile. You can go with the Flip program, your profile when you click the profile button here, you can see the organization that is your organization. You can see this is the preferences and time zone options. This is simple. Now, APA access. This is the most important spot. In the API access, you can and add it to occur. So this is the most important that is MCP server. So this MCP server is better when you create AI agent from other or to call your what flows to other AI agents. For example, you can build an AI agent in the OpenAI from that OpenAI agent, you can add this MCP server, Make MCP server to the OpenAI agent to call your tools that you will create in the Seniouse. Okay? This is the most advanced part, but it is very easy to set up all those things. Okay? Next, in the twoFA you can set a password for the enabling that is you can do all those things here. Well upd problem, very simple. Now, come to our main part that is dashboard. Okay, now from here, we will create a simple AI agent that works like a chat GPT chatbard. So that we have done in our previous NITN customer support AI agent sessions. If you're followed my previous course or sessions multichannel customer support AI Agent using NTN, you will understand very easily in this platform as well. Okay, I understand these points. For more understanding, you can check all those things. You can see the organization settings. You can give the name for your organization. We can select. So we have already select this. You can change the time zone country, all those things from here as well. So this is a simple overview of this make.com. So for more understanding purpose, just go and create a free account and you can explore by yourself for more information about this make platform. In the NES share, we will create an simple AI agent which works like a Chat GPD. Okay, let's dab into that. 4. 1.3 Basics of Make.com : Before creating a simple chat, let's understand the basics of this particular scenario builder. Now when you click here, Let's add button, you will see the different apps you can add here like gold Sheets, data store, tools. Then a lot of you can do the name here. You can change the name from here directly. You can click on that. You can see the Canva options and you can see how much credits are used to run this particular workflow. You can check from this as well. Bottom tab, you can see you can get the two buttons that is run once. If you click this arrow, you can run different scenarios run this butterflow according to any instance like every 15 minutes like that. So let's start by adding some trigger. So what is the trigger? So there's a trigger is, it can be anything. Any tool you can use, for example, in this case, you will take it as a webhook. So for the webhook, you can check this webhook, mail custom webhook as a trigger. Okay. Now, what is the output action that is webhook response. We will take this as a custom webhook. We'll just close this to understand. This is a trigger. So what happens whenever the user or anybody call this particular webhook URL and sends the data to run this particular whole workflow, it will get to the message. This is called trigger. What is an output? If you click here, if you select this webhook. Now this webbook options only have this webhook response, which is called a output model. This body can use any any of this particular in between model, you can add a IGN or any workflow. This is a simple workflow. This is a output and this is a trigger. You can say this is a model. This is the whole scenario. These are the models and or the trigger and this is the output. If you click the settings button, you have the different options that is set up a filter. If you click this, similarly, you can add a different model from here as well as take any wool sheets. I'll just take for you can just click here. You can see here, we have linked this whole apps automatically. Very easily. If you see this, these are in different positions, if you check this one, this is not in the better alignment. You can click here that is auto align. It will automatically align this workflow in better way. You can check this explained flow like if when we run this particular whole workflow, we can easily see the explain flow. And you can see the scenario inputs and outputs you can add. So we will talk about in sessions. So in the settings options, you can check the scenario settings that is sequential processing. You can see this is. So this is not important so much, but according to the clients requirements or the business requirements, you need to see this all the things. Data is confidential or not. So you can check this one options. You can see this test data and the information about this particular setting, okay? So this is you can also add a note here, right click on the model, you can just add a node from here. You can write anything this is a trigger like this. If I click here, we have successfully created a trigger. You can see we have just added. When you click here, it will show this particular this one. You can delete from here. That is simple. We can add the note as well if you want the previous version. So this is a basic knowledge of this one. If you click here, expand a toolbar, we have different flows that is flow control, tools, parser, make AI apps, and the add model button as well. Okay. These are some basics of that. So if you want more, you can learn more about this particular scenario. Build you can click on the right, you can add module paste, any type of butt flow, and you can show the module ID, show router root, let it snow. So if you click here, it looks like some beautiful visual like snow. Okay. I'll just click on. Again, it will just go on. So this is a bubble. So this is a simple message. Okay, you can see the value must be be empty, like that. Okay? These are some basics. And not only that, you can Edit this name as you want. The other point is immediately as data arrives. If I click here, so we need to save this one. I will show the warnings. I'll just go with the Ignore warnings, save anyway. If you click here, we can change the scenario run settings immediately at regular intervals or once every day. This settings all depend upon the business operation or according to the client's requirements or business requirements, you can select this one. According to that, we will configure this webhook and the whole automation, and we can optimize very well according to the requirements. These are the basics of scenario builder of make.com. There is one point that is scenario inputs. Directly, you can click on the scenario inputs and outputs. You can write any input structure. For example, instead of webhook I will just simply delete a model here again. This is a Google sheet. Instead of this Google Sheets, what I will do, I will remove this one as well to understand and to explain you in better way. So this is a trigger we bond, right? So what I will, I will just take as a scenario. Search for the scenario. You can see scenario, start a scenario, written output, called subscenario. I'll just go with the start a scenario. So add scenario inputs. Let's add so we have directly come here. Add a item. Let's create a simple input like a name. So our description, we will just take otherwise, you can leave it as empty. I'll just save this one. Now, we have used this as a trigger, right? This is the input value, so we can manually. If I run this one, for example, I will save this run ones. Notice, ask me to provide input for this. I need to provide a name here. According to the scenario workflow, we can easily configure the scenario inputs according to our requirements or any data we required to run this whole scenario. You will learn all those things in the later sessions, we will discuss each and everything there. Now what I will do, I will just remove this one. This is a scenario which act as a trigger. So what I will do, you can add any type of AI agent or any type of app or any type of model by just clicking right click here, add a model or just click on here, another model. You can select any type of apps. OpenAI, HTPP scenarios, all those things. For example, if you use this OpenAI here, you can write the name. Instead of that, for example, add this open E agent. I'll just giving an example how you can understand this scenario inputs very well. If you click this OpenAI, so you can click on this generate response. Basically what we have done, we have just added this one. Now, we need to connect the open ear connection so we will discuss each and everything in the net. For the model, we just take it as simple. You can take anything. I'll select this petive 4.1 system. Let's do the prompt. We can take from here. Instead of cinary inputs, what I will do, I will rename it again, here, I will just give the name as a prompt. Enter prompt. No Discretion let's keep it as empty. We will just take this as the name of this one. Let's save this one. Now what happens here we are giving the input. By manual, instead of adding this webhook to connect with the exnal chat UI, you can add this scenario to check whether it works. So these are scenario input. That is prompt. This prompt is sent to this particular AI model. So in this text prompt, I can take this name or the name we have created, which we have created in this scenario. Okay, the name has changed, but you can select this. I will automatically configure it. So maximum tokens can keep as simple. I'll just save this one. Now, it is working now. What happens here when I run this? So it will ask me to provide a prompt. For example, Okay. Before that, we will add a output. So what is an output? Output, it should be a scenario, right? Scenario. You can click on this, you can return output from these scenarios. So what basically happening, you need to add a scenario output as well. So for the name, I will just take the AI response. Okay, just click on the Save. So we have successfully connected this scenario input and output. In between that, we have just added a simple chargeability to give the response for our quotient through scenario input. And this response is sent to this particular scenario output. The output can be used anyway. So you can add this whole scenario to any AI agent as a tool. So what I will do is just click here. So instead of that, what I will do just come to this scenario output. Come to this one. In this one, you need to write a description like a written a I response. That is simple. Let's keep it as simple save this one. What happens here? I will just save this one. Now when I click on this on demand, now I'll just click on the run once again. Now, it is asking me to enter a prompt. So this is a scenario input. I'll write Was AI. Okay. When I click this, it will run. You can see it is working. Now, you can see we have got an issue. Let's see what is this. So you can see it is true, output format. In reasoning. Okay, let's check this. Let's configure whether it is in the connection system, prompt name. So it is updated now, just click on Enter prompt. Let's save this one again. Now, come here and show run once again what I will do? I just choose this run with existing data. We have previously just call it. You can run once again. Can see it is running now. Now let's see it is successfully run. Now, let's check our output. You can see we have got nothing. Why? We have just now configure very well. Let's check into here. We want credit user here. We have got a response from AI. You can see it is a input. You can see input a output come to this new conversation, text, two, it is a prompt what is an AI from scenario input. Now the output is come to the result. You can see AI stands for. You can check the output as well here in this type of model. Now let's jump into this scenario outputs. Let's run this model only, not is working. So we have gotten some issue you can see missing value frequened parameter AI response. So let's check. So this is how you can just come here, select the response from this previous AI agent, that is, you can take the response raw result or in the result, you can check, click on the result. Save this one. No it will run Let's say this. Run once. So I will just take the question as a What is an AI again? Not is running. Let's check we have got an AI response. You can see we have got an AI response. Basically what we have done, just click right click on this, and you can give this particular result from the previous AI OpenAI model that is result. Select this now it will automatically generate a response. This is the power of scenario input and scenario output. Basically, what we have done instead of adding the any different triggers to check whether it works or not, you can use the scenario inputs and output as a triggers, and you can test it out this whole scenario with our manual data and we can add when this particular scenario works very well, you can add this whole workflow to external apps or you can go for the live position. Okay. And there are a lot of more use cases. Not only that, you can rename all these apps. You can have very different things like let it know. You can check each and everything as just click on the right. You can rename it like a chart and GPT like that. You can rename this particular model as well. Click on this again, rename it that is input values like that, you can ream this particular app names, marvel names as well. Not only that, you can just click here. You can import a blueprint or explode Blueprint. Blueprint is nothing but this whole scenario file. It convert into the JCNFle it will provide to us. You can use it, you can share with your colleagues, users or clients to test out in their systems in the mag.com. You can import from the file just click here, you can choose the file that you have doloaded from the mag.com. Otherwise, you can export this whole scenario by clicking here, and even you can copy this blueprint, just to paste it wherever in the new scenario builder. I will just paste here only to show they can see it is working perfectly. If you check here, we have copied this whole scenario with this trigger. But in the next time I have pasted this whole scenario, but this trigger is not shown. Why? Because for one scenario, we can only put one trigger. This whole scenario is triggered by only one. We cannot use a number of triggers in one particular scenario. That is the point. I will just click here, I will delete this just by clicking here, Dital modo Ditm and Diltal model. You can do so much things not only that, you can click here for Auto Alive. You can see this Aligned Auto. You can explain you can click here airplane Cone button to explain the flow how this workflow is working. You can see it is working like this. It is showing this is the working of this flow. This is a scenario inputs and outputs. You can click on the scenario setting, you can change according to your requirements as well, and you can attach a node for each type of model as well we have used here. Just come to this click on this particular model, you can add a node. You can give the node like it is input data like this. It is well when you provide this whole workflow to your clients or anyone, you can explain this what this particular model does and what is the output of that. You can add this node, as you can see, we have got the icon here as well. Okay, you can jump into the previous sessions by adding this one. Okay, you can pick up the versions that you have created in the earlier. I come to this flow control tools, there are a lot more things you can check with the logs, a logs that is preparing what we have done in the previous, this is the one. You can share it as well click here. You can create the public scenario page. You can share this. They can change the settings, you can change the settings like name description as well, and you can save the changes and you can copy the link, you can share to your colleagues like that. This is a simple Things you can follow, you can check the credit usage. If you click here, you can see we have user so many credits you can see this EI app model is user one credit, that is simple. It will show the credit user as well, show model ID, show router order. This is how it works. You can try out by yourself. You can add. We have just learned some basics, how the scenario inputs and outputs work. If you are looking to learn more about this scenario inputs and outputs, so you can check the documentation as well. Otherwise, you can go yout videos about scenario input and output for different use cases. Let's jump into our creating first simple chat AI agent. Let's dive into that. 5. 2.1 Building Custom Chatbot Ai Agent: In this session, we're going to create an simple AI agent which works like hat GPT chat board. Okay, let's do that. So just come here. In this dashboard, come to click here AI agents. Now you can create AI agent from here. You can start from here or here. L just go with this create AI Agent button here, no, you need to create an connection here. I'll just create a connection. Now you can find the different providers. You can see make AI provider OpenAI, anthropic Cloud, Jenney AI, Michal. There are a lot of AI provider by companies you can use by yourself. You can add an OpenAI. You need to get your OpenAI from OpenAI platform, and you need to create this connection in the make.com. I will just go with Make AI provider because we don't need our extra setup like OpenAI. We can use this Make AI AI Agent by itself to test it up. Now I will give the name here. I will Make provider or make AI agent. So to track our agent very easily. Now we have just create the connection with Make AI Agent provider. Now we need to give the name for our agent. Now we are here to create a simple AI agent which works like hatchiPit chatbot. Now I will give the simple name that is Chat or let's give chat AI. This is a simple chat AI Agent name that I have given. We can select and model according to your preferences. Now you can give it a small medium lodge. Now I will just go with the lodge GPT five mini for better reasoning to our prompts and all those things. Now as for the system prompt, I will just give a you are a helpful AI agent because we are here to create a simple AI agent which works like chat GPT. We will just give this U or a helpful AI Agent. You can give the assistant like that. Okay, now I will just save here. You can see we have successfully created AI Agent. You can see the example you are a customer service agent. We have three different features of this make AI Agent like context, MCP, tools, which are very most important. In upcoming sessions, we are going to see how we can add the context MCP and tools as well for our AI Agent to automate our task or to improve our AI agent intelligence, all those things by providing our own data to AI Agent in the upcoming sessions. You can test from here. Come here, just try hi. Now you can send a message. Now you can see, it is thinking now, you can see it has just given Hello, how can I help you today? Okay. This is simple. This works like a simple chat Gi pity chatbot like that. You can see here this is 1.04. This is a credit that it is taken to give the response to me. These credits are just decrease in my accounts. You can see here, 900,998.96 credits. This is how it works. For every operation between AI apps or AI Agent response, it will just decrease the credit come back to AI Agent. Now, I will send this chat AI again. This is a simple chat PT like chat AI agent. Now, this is a simple. Now when you see this one we have just created the simple AI agent works like hagipi in the NTN in the previous course. We have just created the AI agent in the NTN just by taking the chat messages from the users in built chat AI in the NTN itself. We have just given the simple memory, open AI brain to this AI agent. But in this make.com, we don't have the built in chat section like this here to integrate this chat bought in our website. We don't have like that, so we need to create chat section outside of this make platform. So I have just used the ChatPT to create simple AI chatbot. Before that, we need to create an webhook to call this. I will explain each and everything. Come here. So this is in the Enten, we have just create the chat message, okay? Which is available to the user directly by the NTN platform. So when the user send a message here, the agent will use the open Ibrain and simple memory to store users data, like name, all those things. Then it will just give the answer back to the chat. Okay? Now, in themge.com, we don't have the inbuilt chat functionality like NTN, so we need to extra setup outside of this make platform to send or receive the response from AI agent in the chat. To do that, we need to create an simple chat GPTUI in the outside. Then we need to connect that chat UI from outside to the Make platform. We can do with methods by using the Make API or using the webhook. The best part is, it is very easy to set up the webhook instead of ABI. Just click Save this Agent save. Now before that, you can check the documentation to learn more about AI Agent. Now you can click at the Agent settings. You can rename your agent name as well, you can select the other model and you can control the output tokens as well. You can keep according to preferences. Now you can just call the steps per agent call. You can see what is this main maximum number of steps that agent will do before stopping when solving for single call. That is very easy to understand, and this is a maximum number of agent runs in red. So up to ten, it will remember your history. It will remember you previous chat history up to ten, even can add so many I'll just keep as a ten simple. Okay. Now you can save this one. Let's jump into scenarios. Now, you need to create an scenario from here. So before that, we need to create an folder. Okay, you can create a scenario without an folder, but to organize my scenarios, AI agents with close, I will create a folder. Okay, I will just give the name as customer or multichannel's give this name. Multichannel customer support. AI agents. Let's save this one. Now, we have the folder. Okay, just click here folder. Now we have the folder. Now we need to create a scenario here. Just click here that is create a scenario button. We are in the scenario builder UI. I will give the name like chat, AI or simple chatbot. Simple chat. What? I'll just give you the name. I will just click the plus button. I need to select the webhook, which is very most important to connect our external chat UI to this make.com to send or receive the response from AIHN. I hope understand these points. But let's click here. If you don't find the Webooks you can search for here like webhook. You will find this one. Just click here. We have three different options. That is custom Mail Hook, custom webhook, and webhook response. This is the webhook. We need to send the message to the user. I will just go with the custom webhook. Okay. Now, what you can do, you need to create an webhook here from scratch, just click here. You can the name here. I will just give it the chat. I'll give the chat. This name. Webhook. Okay, that is fine. Now, you can add an API key as well to authorize your API. You can create an API click here, create a keychain for security purpose. Before creating this keychain, we need to test it out whether this webhook is working perfectly or not. Okay. Now, for now, I will just to go with this creative webhook. We have successfully create the webhook. So we need to call this webhook URL in order to check whether our external chat you are getting the response or not. Okay? To do that, I will jump into Chat TIPT to create a UI for ourselves. Okay? Now, I am in the Cha GBT. I have just created this particular UI by using hA GPT. Okay? Don't pet I will provide this whole code in the documentation. You can get you can run this whole code in our website or wherever you want. If you got issues in this chat UI, you can use that Cha GBT for hell to resolve your issues. Okay, I hope understand these points. Now what we need to do, come here. You need to copy this one. Come here, just copy address to clipboard. When you copy here, just come to chat UI. Search for this one, C webhook URL. Okay? Now you need to place your webhook that is you copy it from make.com, and I will paste it here. Okay. What you can do here, you can save this one. So run your code by self. Now I will just go with the preview button here. Now our make.com, UI hat looks like this. When I write any message, I will send a message here. Allow Let's try another time. Send hi. Now you can see it is accepted. Now webhook working perfectly. Now when you come here, you can see it is successfully determined. That means our webhook is working perfectly. Okay? You can enable that one setting here. Now you need to keep it as empty here. For more information, you can see here. Okay, you can see if it applies only if the scenario is not scheduled to run immediately. Now, we need to add an AI agent that earlier we have created. Now, search for the mai agent. That is you can find it here or make a agent. You can find the more options, so we need to go with the run and agent. Now you can select your agent that you have created. In our case, we have created like hat AI. You can add the tools from here come the thread ID, you can add in Thread ID as well. So we will talk each and everything about in the coming session. But the message, that means it is a prompt. Okay. So we have tried all the things in the previous while creating the chat by sending the high message like that, we need to create an message to send an EI agent to give the response for our query. Okay. So we need to provide this one message. This message will come from the webhook that is message which calls from this chat UI, because we have just paste our webhook, or to connect our make Make AI Agent and this exnally hosted web chat. When I send the Hissage, this hi is sent to the make.com in the webhook, you can find it here in the message. Now what I will do, will just drag and drop here. Okay, now our AI agent has access to our message. Now, just click in the save here. Our Make AI agent will give the response for my message. Now, click here. You can see the webhook. Now we will only get access to this option, webhook response, the body and just search for the make IIH. Give the response. What happened here? I will explain from scratch. This is a webhook which received the user message details from the AI chat we have created with the webhook URL. Now this chat will send our message that is high or our other query to the make AI platform here. Okay. Now our AI agent have the access to my message as a prompt here, messages, you can see here. We have just dragon drop from the webhook URL. Now EI agent will give the answer to my message, and it will send to this webhook. So this webhook have the body. The body is the response from the EI agent that we have earlier connected here. You can find the response in under the make AI agents. This webhook we will get the response in the chat section here. You understand these points very well. Now, let's send the message like tight. I'll send it here. So you can see, there is no scenario listening for this webhook. Why we have not enable this scenario here. Okay? Now you need to save this one. Now you can come here. You can enable this. You can publish immediately as data arrives, you have different options to do that. Now what I will do, you can click here run with existing data. You can select Before that we need to run once. Click here run once. No can see it is running now. Now it is waiting for data. Now I will test by sending the message form that is I will send high. But Now, you can see we have got the response from AIHN. That is how can I help you today? That is very easy, right? Now, come to that is simple. We have created the webhook and AIchN which works like a simple AI chatbot like Cha GBT. Okay, come here. You can make it immediately as data I. Save it is live. When I send a message from here, it will give the response directly here in the chat. Let's ask some other questions like what is AI explain win 40 words. Let's give this qui. Let's wait for a few seconds. You can see our AI agent has given the response that is artificial intelligence AI refers to computer system and algorithms that perform tasks. It is under the 40 words. Okay, it is working like a simple chatboard will give the answer for any question. Now we are here to create a multichannel customer support AI Agent, which works specifically for my product and services that are in online. I understand these points. We here to build a specific AI chat booard to work as a customer support agent for my company. In the next session, we will add our context background information like company details, product and services to this AI agent. Let's dip into that. 6. 2.2 Adding Knowledgebase to Custom Chatbot Ai Agent: No this session, we are going to see how we can create this particular AI h to work as a customer support A Agent for my product and services. Okay. And remember, before creating that particular AI HD, you can check this by clicking in the bubble here, you can see what is one operation it will take the one credit, and you can see what the session Edit has taken from that. Okay, you can see this the message we have retrieved from the chat UI, like that you can click in the AI Agent output. You can see it is input which comes from the chat UI. This is the output. You can see the response Agent response, how Hello, how can I help you today? You can check all the things from her as well. You can click head the bubble for more information for the operation, credit user, and the output input as well. Now, we will create the simple AI Agent for my product and services. Just click here, AI Agent. We will keep as it is, but we will add our contacts background. We will add our own data in the PDF format which have my FAQ about my product and services details, my company details, and simple customer support. I will provide this whole data to this AI agen and it will use as a knowledgebase, it will give the answer from my company data, not by own EI agents intelligence. So what happens instead of working like a generalizer purpose like ChangeB or other AI modus, it will work like a specific AI agent for my product and services. It will only give the answers from my knowledgebase. Now, come here, just click in the Make AI Agent. You can click here configuration. Let's jump into configuration. Now, you will land here that is system prompt, all those things you can change system prompt, all those things here. Now, let's add the context. You can see background information, data or additional context for your agent to reference. Now when I add my company details, product and services through PDF and I will add as a context, it will use as a background information as a reference to give the response to user, instead of working like a generalized purpose AI chatbot like Cha chibit. It will work for specific only for my product and services. Let's add our company data. I will upload my PDF fine. You can send company. You can see this company docs PDF. I will just upload here. That is uploading. No, you can test it from here. Before that, we will check this PDF, what my PDF contains. This is my simple PDF. You can see I have just generated with the perplexity to test out my AI agen. You can see this is a simple completes, ACM home appliances founded 25 headquarters, Machine this is a products the company will sell. This is a frequently asked questions. You can see how do I contact customer service, phone number, support, email as well. Products. This is our products information, other examples. This is a simple company data that have generated by using AI modules to understand the AI ten ecosystem. Now you can do by yourself as well. Now I will just come to this chatbot. Now, I will test how directly here. Now before that, you can write the system prompt to use this context. Now, you can copy this example because we are creating the simple customer service agent only. You can see this is example that has already given by meg.com. You are a customer service agent whose primary goes to help customers with their queries and resolve issues quickly. Always respond with a friendly and empathy tone. I will just copy this one. I'll paste it here. Now, what you can do and I will view this Stritly follow only use provided data as context to ser User query. Now, let's see. I have just given this six follow only use provided data as contacts to answer User Query. It will give the answer to the user according to my provided company data. Save this one. Now, I'll just give what is your company name like that? Let's see this one it is thinking now. Now you can see it is essential context retriever it will retrieve the answer from my company. You can see our company name is ACM home appliances. You can check this in my document itself. You can see here. Now it is successfully retrieving information from the provided company. You can ask any question like, I will just copy from this you can see what is a varunty period of our products. Let's copy this one. I will send you this here, let's give this one. Let's check whether it will works or not. That is thinking now. You can see me home applies this product come with two standard warranty. You can see. It is also retrieve the phone number, support, and live chat on our website like that. Now it is working smoothly according to word data. Now, let's test out from our chat UI. Now, let's come to a scenario, we have created at here. Now we will just come and we will save this again. Now let's come to chat UI. Now I will send a message to AI Agent. Same question that is, what is the warranty pid for your products? Let's set this. Now you can see we have got the answer from our data provided. You can see all APMA products come with standard two year warranty. Extended warranty options are available if you would like longer coverage. If you want to purchase an extended warranty, you can contact us phone number, support, email. No is working like our specifically customer support agent instead of generalized purpose EI, like Chagbr. No, let's test out. I will ask a question like what is EI out of our EI agent contexts. Let's take what is EI. Let's see whether it will give the response or not. Et's wait a few more seconds so you can see. Now you can check. I don't have direct definition of AI in company documents. Now when I ask exact question like what is an AI explained in 40 words in previous session, it has given the answer that is AI refers to computer system and algorithm that perform tasks like that. When I add my context, when I add my context information and instruction strictly follow only use provided data as context to answer user query, now it will just refuse to give the user question which is out of company's data. That is what is EI. It has just been given I don't have direct definition of AI. Now this AI agent is working specifically for my customer support service for my product and services instead of working like generalized purpose like Cha JP chatbot. I hope understand these points very well, but providing good instructions and context, this AI agent will work like that. In the NST session, we will add a tool. Let's see this one. We will add a tool whenever a user asks a question. When AI agent is not get the information from the provided data, it will escalate to the human agent, or it will send the user query to the humans team, which will handled by the humans not AI Agent. We will see in the Net session. Let's dive into that. 7. 2.3 Adding Human Escalation Tool to Custom Chatbot Ai Agent: Now in this session, we will see how we can add human escalation to our AI Agent that whenever a user ask a question out of our company data, about the product damage or any discount, this AI agent will automatically send this particular user query to our team, Okay, which is handled by women's not by EI agent in which they can interact with user and they can solve the user queries as soon as possible. Now, how we can do that now this is a chat AI we have created. So before that, we will test our AI chen how it will work whenever a user asks a question, which is not in our company's data. Now I will ask a question which is not in the company's data. I will ask a question. That is, I need 90% discount on all products right now. What happens here? I'm just asking AIHN I need 90% discount. Let's write the discount here. On all products right now. Now, I'll just give this question. Let's see how it will works. Let's wait for a few seconds. Okay, now you can see. I understand wanting a big discount, I would help you to base on the company information. There is no record of any policy or promotion that allows 90% discount and discount on all products, and I don't have authority to grate such a discount. That is well and good. Now, you can see here it is simply saying that I don't have enough data to provide related to the 90% discount on all products. Because I'm not mentioned in my company docs. Now what happens, it is simply showing what I can do right for you, check available offers or promotions, escalation formal request to sales team. This is the name. This is our main thing here. And you can see suggest alternative that may reduce cost all those things. No, helpful contact. I have provided my phone number and email support as example in my company data PDF. You can see it is providing, so I need to contact to them here. We can set an automation. Like whenever a user asks a question out of our AI Agent knowledgebase, this AI agent send to the human agent or sales team automatically in which we can track our user, not to leave our platform. I understand these points very well. So to do that, just come to this particular AI agent. Now you can add from the as a tool. Okay? This AI agent will use as a tool to send the mail to company sales team, which handled by the humans, not by AI Agent. Okay, this sales team will see the query. They will reach out to the user directly. That is very easy to implement that in this make.com. We have already done in the NAN. That is, this AI agent will automatically send an email to the team, which is handled by the humans when the user asks a question, not in the company docs. We have done in the Ineren If you followed that, you will understand easily. If you're not, you can check it our Ender ten master class as well. Otherwise, you can follow from in them make.com, which is very easy. Now, let's jump into creating this feature in the AI agent. Now, let's come here. You can add as a tool here. Just click on the ad press button, you can add as a model or scenario. As we can see, as a scenario, you need to create and scenario separately, then you need to connect it. Or you can use as a model. Okay, let's jump into by creating the model because we are using only the Gmail here to send whenever the user asks a question not related to our company knowledgebase. Okay, let's create a new model. No, let's find further GML, you can see here. In your case, you can add anything like any email automation platform you are using for your business or you can use slag, Zendesk or anything you want. Okay, let's create an model. I will just select it as Gmail. No, we need to send an email, right? For every time the user asks a question not related to our knowledgebase, this AI agent will use this tool to send an email to company steam. That is simple. I can see you can give the name as GML, send and GML. I will give this name as a Humane escalation. That is simple. Now, you can use the description as well. You can see help your agent to understand when to use this tool, you can describe here. So we will write like call this tool where user ask query out of your knowledge. That is simple. Okay. You need to correct the Gmail here. Let's create an connection. You can give the name as well. I will just keep it as my Gmail connection. Just go with the sign in with Google. I will connect my Google account that I already have. You can do by yourself at the same. There is no extra setup needed. I'll select all, continue. Now we have successfully connected our GML to this make AI. You can see this is one. Now, what happens here, you can see where we need to send this email. Whether you can select let AI Agent decide or we can give this address. You can add any email. For example, in your company, there are a lot of teams have their own Gmail account like sales at the rate, company.com or support at the rate accompany.com, like that. So I will give the simple Gmail to understand that let's see. No, I will give the simple Gmail, which is in my account which sends a support. Okay. Let's save this one. Now, before that, you can see here, you can add a recipient. You can add so many respons if you want add recipient, other Gmail ID as well, and you can give the subject. You can select Let AI Agent decide. For the body, you can select two types. That is raw, HTML or collections of contact text images. If you are sending the message live for the different contact with images, text, and all those things, you can go with this collection of contacts. In my case, it is only raw HTML Image. I will send only the text, right? For the content. For the first time we are adding this tool, so let AI Agent decide to send the content, right? Now, let's say this one. You can see the advanced settings as well if you want. You can check the from, you can write you can select this one as well. You can do all those things. I'll just give the air. Now we have successfully added the tool that is Human acceleration, that is GML. Now we can click He Settings button, you can change the details, disclose it and you can check this. We can calcel this one. We will test from the chat. For example, let's take this is the support GML, which handles by the Human. Now, what I will do, I will send a message. Before that, we need to write a Agent instructions. Okay, let's jump into this one. Now, I will write this agent as instructions. Whenever User ask query out of your knowledge. Whenever user ask query out of your knowledge. Use tool. Let's give use tool. You can give the name that is who made scanation too. That I copy this one. Use email escalation tool, you send an email through team that is simple and save this one and it will update according to preferences. Now let's try from the chat. I will just write like I need. I'll just copy the previous one. We'll send this first. Now, let's check. Now, let's check this AI Agent we'll able to send an email to our team or not. But this query, I need 90% discount on all products right now. Let's check. It is taking time. Let's wait for a few seconds. You can see thanks. I can help with that. I have accelerated this request to a team to confirm whether a 90% discod is permitted and to get any aproveicount options or not. Let's check our email. Now you can see we have got the escalation email from AIH and you can see it has taken by itself that is customer requests 90% discount on all products unchained escalation. That is very good. You can see this is a message customer message. You can see it also retrieved my message. I need 90% discount on all products right now. You can see details, customer requests are immediate, so it is a summary of my usage. So you can see, now what is the main thing here. If you see here, it is simple it is like in paragraph. You can tell to EI to only give this type of information, only to send an email with the customer email name and the message only. In that instead of reading this whole message, I can only track the user message and the email and the phone number or name in which I can directly contact to that particular customer, which can save a lot of time. To do that, we need to write the simple instructions. We need to update instructions like before escalation, gather user name, email, and send gather user name, email, and user query, and send mail to team Pit user message or querF example, without name user name, email, how can I sales team can reach out to them? That is main important right. So when the email have the user name user email and message query, the sales team can directly reach out to that particular customer with the email or phone number or name in which they can easily track the user's query and problem solution, all those things, right? So to do that, I will save it. I will try it again. Whether this particular AI agent will ask the user name and email or not, and it will send it to our sales team or not. Okay. Let's check this again. I will copy this one. I'll write this one. Again, let's send this one. Let's wait for a few seconds. No, you can see, I can help. Thanks for asking. 90% discount is unusually large and isn't covered in the stand in materials I have. I need to escalate the request. Now, before I escalate, please provide know this AI agent asking my name, my email address. Then it will send the email to sales team with my name and user name, user email address and user query in which they can track the user queries. I will give the name name and email exam I'll take. You can see any additional details, order number or like that. Now for that additional details, I will just write no additional Details. Let's give this and how this AI Agent will work. Let's see. You can see also telling you if you prefer you can also contact customer service directly here this contract support. Now you can see, Hi safe thanks for sharing our quartet. How can you help you today? No, you can see. If you'd like me to escalate this to Team, I can send them an email just confirm you want escalation and type the message you want to include. Now, the major drawback of this particular AI agent is, it is not recognizing my previous chat. Why we didn't connect any data store. What we can do can see here, maximum number of agent runs in a thread history. Okay, up to ten messages, it can remember my previous chat, all those things here. So right now, we have just given the message out of ten times in which it is not recognizing my previous chat, and it is telling again, hi thanks. How can you help you today? Instead of escalating the mail to ss tap with my details, it is starting from scratch. Okay? To do that, I will just stop it here and I will run again with the new Message. Can you chat? I will send this message again. I need 90% discount on all products right now. I will send this is how you can learn the building a agents by trying out by understanding the process of all settings in the make.com. New to try by yourself, then you can get the knowledge of it. You can see thanks. I can help you with this quick note. I started an internal escalation so the team can preview explanation discount request. I still need a couple of details from you to complete that request. No, I need to provide my full name, email address or any details. Okay, then it will send these details to sales team. Now I will just give the name again. Et's skew the email as well. Are you asking for 90% discount on single product multiple specific products? The query is all products. Or are you thinking I am asking for all products. That is simple. Send this whether this agent is able to send an email with matails or not. Let's check it out. Now I have given my name, email, and simple message. According to the agent response here. Let's wait for a few seconds. We sending now. Now you can see, thanks, sir. I have got your request. Bilows Activa home appliances product catalog puller from our records, company summary, product catalog all those things. Now let's take our email, whether it will send it or not. You can see Now you can see we have got an explanation email from the AI agent that is customer request 90% discount on all products. Let's see. Now, you can see we have not got anon email. You can see customers requesting 90% discount on all products immediately. Please advise if such a promotion approval is possible and what steps are required. Customer details, name to be filled, to be filled. We have not got any message from name or like that. Now, to solve this issue, let's go with the scenario. You can click on our scenario that is simple chatbot we have created. Instead of data stores, we have another method in which we can send the user name email team. Just click on the edit, click on the Make here Agent. Quickly Advance setting here. Now, we can write advance additional system instructions in which we can send the message directly from the message here. Okay. Now what we can do. Send team when user. Ask QH talk. Your knowledge with user name. Okay, let's give shift Center. Sermil Ship center and query. You can see here. I'm telling to AIs. Send an email to T user ask a question out of your knowledge with user name, user email, and query. We have got the message from this one only. We can write the message like Analyze This message, let's drag and drop. And be option. Accordingly. Analyze and look for the I will automatically analyze this user message and it will look for user details and it will categorize accordingly here and it will send it to the sales team. Let's check whether it works or not. Let's save this one. I will save this one again. Let's dup into this one. Now, I will start from the scratch. Let's preview one. I will use the same exact query. I need 90% discount on all products right now. Let's wait for a few seconds. Thanks, I can help with that. I don't have authority to apply 90% discount on all. Can you please provide a name, email so I can escalate this request or team? This is the power of instructions. Then when I give the instruction, it will just following my instructions very strictly. You can see now what I will do, I will write my name. Email as You can write the query as. Or otherwise, you can leave empty, it will automatically take for better understanding, I'll just write again this query as well. No, let's send this message. Thanks. I have escalation this to the team and included your details. They will review the request and get back to you. Meanwhile, can I confirm your preferred phone number or additional details. No, it is the Agent is telling us, I have accelerated your query. Now, let's take whether we have got our main, we have got this one. Customer discount request escalation. You can see, no, we have got the user name, email as well, and query, I need 90% discount and products are not. C is also required in the current time. It is also suggesting me to please advise any discount policy, feasibility of 90% discount, all those things. Now we have simply connected the Hub systems to AIHt Whenever a user asks a question, which is out of the AI Agent knowledge, Agent will simply send an email to Women sales team. They can reach out even you can ask a phone number as well. Now in the next session, we will see how we can add the Google Sheets. It is simply adding the agent logs to our system in which we can track the user query and Agent response, whether the agent is working perfectly or not. Let's dive into that. 8. 2.4 Adding Agent Logs System Tool to Custom Chatbot Ai Agent: Now in this session, we are going to create how we can track the agent response and user query by simply just adding a Google Sheet to the agent as a tool whenever the message from the user gets to the EI agent and the agent will give the response, both EI agent response and user query are stored in my Google Sheet, then I can rectify or I can check whether the agent is responding very well or not. According to that, I can improve the EI agent efficiency by writing the additional instructions and prompt. Let's dive into meg.com. We are here in the EIgent. Now what we need to do, just click in the AI agent again. Now we need to click on the configuration again, just click here, click and add the tool again, add as a model, select a model or create a new model. Now, find the Google Sheets. Now you can find it here. Just click in the Google Sheet. For every interaction between EI Agent response and user query, we need to add a row. Just click here, add a row. Now we need to write this simple and creative connection. Let's sign in with Google. Now I have the GML code. I will just choose theta code, click the Catinu select all button, click on the Catinu button. Now we have successfully connected our Google Sheets. Now I will go to Google Sheet and I will create a spreadsheet to track our Agent response. Let's jump into Google Sheets here. I will just go with from here and let's set. Now, click the blank spreadsheet, create a new spreadsheet. I will give the name as gen Blocks. AI Agent logs. That is fine. Now, I will give the name as user message or user query here. Let's give the name as User query. In one case, you can keep ask for your requirements. Now let's give the name as Agent response. Let's do this one. Now we have successfully created the two columns, that is user query and Agent response. Come to the make.com and we will search this spreadsheet from here. You just click here search by path. Now my drive, choose spreadshet ID. Just click here to choose five. Now you can see we have got this AI Agent logs we have just earlier created here. That is AI Agent logs. I will select this AI Agent logs. No, automatically just give this sheet one because the sheet one is here. You can see in the bottom side that is sheet one. Now is for the user query, keep as it is, let AI Agent decide and let AI Agent decide. Let's click Ad Bottom. Now we have successfully connected another tool that is Google Sheets. Whenever a user and send a message and AI agent give the response, the user query and Agent response are stored in these two columns. Run this AI Agent. Let's see whether it works or not. I'll jump into here, I will save this. Now, let's jump into our chat UI and we will check. Let's start with high Nobody to enable workflow is running, that is live. Let's check. Let's wait. You can say hi, nice to meet you. How can I help you today? We have got the message from AIHt. Let's see whether our AI Agent successfully stored our conversation in the Google Sheet or not. Let's jump into Google Sheets. Let's refresh this. If you see her, we didn't get any user query agent response here. We need to write the instruction in the system prompt like stored response and user message in Google Sheets. Let's give the name here that is Google Sheets, add a Row tool. Let's copy this one. Come here. Tool. Let's save this one. We have to add an extra instruction to use this tool. Google Sheets add a row tool whenever a conversation between you and user occurs. Store your response and user message in Google Sheets using Google Sheet, add a row tool. This is a tool name here. You can change by yourself if you want. Save let's try with the high again. We have gone the response from AI Agent. Let's see our Google Sheets. Now we can see. Now we have got the user query har and the agent response. Let's try asking another question. Can you list products? Can you list company products? Let's give the question. Let's see the answer from AI Agent. We have got the product details from Agent and warranty, all those things from our company data. Let's check Google Sheet whether our Google Sheet have the data is stored or not. Let's check. Now you can see we have got the data. Can you list with our products, you can check this is our es or query, and this is the agent response. So we have got this one. By doing this, we can check the agent response whether the agent is giving answers perfectly or not according to the user query. Then if any issue in that, we can train our AI agent very well or by writing the system prompt or additional instructions to make our A agent work properly or to reduce the hallucinations or issues in the response. Not only as a tool, you can add from here Google Sheet directly to let's like this. You can select by path, not as a tool, you can also add the Google Sheet model here directly in order to save the user query and the AI Agent response as well. Or you can add as a tool. I will use as a tool only for better working property for better progress. Like we did all those things in the Enten also, we have just created the AI agent that is Agent logs. We have created the Google sheet to track our AI Agent response and user query from the chat section. We have also done in the NTN Astel. Up to now, we have successfully created the simple chat AI Agent, which give the as according to our provided data, which contains product product details, computer details and FAQs and support mail phone number like that. Whenever a user ask a question out of AI Agent knowledgebase, it will simply send that particular query to the sales team by gathering the user name and email with that and we can track the AI Agent response and user query by simply adding the Google Sheets to the AI Agent as a tool. Let's say we have another problem with this AI Agent. Let's come into this chat section. My name, let's send this message. I have just told to this AI agent that is my name is SF. Let's wait for a few seconds to get the response from AI Agent. Now I can see, nice to meet you SI. How can I help you today? This Agent is successfully recognized my name here when SIF. Now let's ask another question that is, what is your company name. Let's give this question to AI Agent. Let's fat. Now you can see our company name is AkmaHme ApplancesPrivate Limited. No, I will ask my name again. What is my name? Let's see whether it will recognize my name or not. Let's wait. Now you can see, I don't have your name in our records. Could you please tell me your full name and email address. Now when I told my name that is my name is Sive then I ask another question. Then again, I ask, what is my name? Then this AHT is not recognize my name. To solve this problem, we need to add a data store in our MEH. Okay. We need to add a data store in order to recognize my previous chats, previous name data that I have shared with this AI agent, then it will give the response very intelligently. It will learn from the previous chat, it will recognize my name, then it can reduce the errors. You can use any database like Superbas all those things or we can use the data store that is built by make.com to give the personalized answer. We will do all those things in the next session. Let's dive into that. 9. 2.5 Adding Agent Memory System to Ai Agent - Part 1: Session, we will see how we can add a memory to this AI agent in order to give the response more intelligently than before. Now, why we need to add why we need to add a memory for this AI agent? Let's check. I am in the chat UI. Let's send the message that is my name is S to AI Agent. Let's check you can see it is recognized my name. Hi, if. Nice to meet you. How can I help you today? That is well. Now I will ask another question not related to my name. Let's take where your company or where your products made. Let's check. Now you can see thanks all our products are designed in India, manufactured at Bangalore and Chennai facilities. This is the information which is retrained from my company documents, right? Now again, I will ask my name to this AI Agent, what is my name? Let's see. You can see the A agent simply refused to give my name. I don't have your name on file. Could you please tell me your name again? So it is not recognize my name, even if we have in the same chat, right? To avoid this mistake. So we need to add memory to this AI Agent in which it will save our previous chat, data, all those things in data store, and it will recognize my name, and it will give the response with my name, which improves the personalization to each customer, right? Like that, we can add a data store to the AI agent. Okay, now let's take how we can add the data store to this particular scenario in which our AI Agent can give the response more intelligently by recognized previous at, my name, all those things. Okay. Before adding a data store to this AI Agent, we need to understand what is a session ID, all those things here. Let's click here. Click and you can simply click here. You can see we have three things from webhook, we will get the three data sets like session ID, ID, and message. If you see here, the session ID is this one, right? Now what we can do. This session ID can be changed when we start a new chat again. Okay. Now for this chat, this is a session ID. This is some value, right? Now, you can see here, it is a 7a52 fe94. Now what I will do, I will just come here, I will stop it here. Now I will start run again. Now I will send a message like Hi. Let's send this one. Now let's come here, click Run once again. So we can see we have got the response from the UI chat. Let's see we are waiting for this again. Now I will send a message hi again. Now, let's check. No, it is run. Now, let's see the output here. Now, our session ID is simply changed, right? But it is not changed. Why? Why? Because this UI, this chat UI is available to different customers. Per customer, we will get this session ID unique, right? We will get this session ID as unique per user, not for all users. So in that we can track the user ID or user data in our data store because we are tracking the session ID per user. Okay, with that user name, all those things, we can store the user name or previous chat in this particular session ID. I hope you understand these points very well. So we need to use this session ID as a user chat ID in order to recognize the user name or previous chat to give the butalze response to that particular user. Come here. Now, after successfully giving response, we need to store that particular response of this AI agent and the user query, right? So to do that, you can come here. So we have the Agent response here. In the output, you can check. This is a response of AI Agent. Hello, I am here to help, right? So know what we can do. Come here, just click here this settings button. Just click at A model. We need to search for data store. So this is a data store. This is the built in action which is built by make itself to store our records as a data, all those things, okay. Even though you can add your own database like super Base, firebase like that, but it is very easy when compared to that because there is no extra setup in that. Okay? Now, let's tap into data store. Now what I will do. We need to add or replace record, right? Whenever a user give the message and the agent generate a response, the two IDs, that is user message and agent response are stored in this data store. Right? And this AI agent will recognize our chats, right? So we need to click here and create a data store. You can give the name Agent. Can give any name like Agent data. Let's take this Agent DP database. You need to create a data structure here again. Click created data structure. You can give the Agent data Agent data structure. That is for the specification, that is most important, click here add item. So we need to add two different items. That is to store our agent response and user ID or chat. To do that, you can give the name as for the Agent response, we can give this thread ID. So you can write the description as well, but we will keep it at empty. Now for the required, it is required right now. You just click here the required and just keep it as simple and we need to create another item. That is click on the add item again here. We need to add further chat ID like user chat ID. That is simple. Now keep it as required, and let's save this one. Now we have successfully created the data store. So you can keep it as a three MB or you can see it must be number 1-10. It is variable according to your account or plan you have chosen. But this pro account will get to 11 and ten MB. In between up to ten MB, you can store the user chats, all those things. Now, this is a drawba. But when you use your own database like SuperbaseFb Base, there is a lot more things you can manage. But I will just go with the five MB as data store. Storage. No further key and thread. Now, so to run this action, we need to require thread ID and user chat. For the thread ID, we will get from the EI agent. You can search for here. That is you can see here. This is a thread ID. Click here, it will automatically takes them. For the user chat, we need to get from the webhook that comes from this form, that is session ID. Okay, so we successfully store our agent response thread ID and user chat ID from the webhook. Let's save this one. No. Whenever a user send a query and this AI agent give the response for that q two IDs that is session ID and thread of AI Agent will be stored here. Okay, to run this one, we will run this again from the existing data you can select. Now, let's run once. Now you can see we have successfully run this webhook. Now it is AI Agent is running. Let's save this one. Now we have successfully run this action. To check the data store records, all those things you can come here, click on the Me button, just go with the data stores. I can see you have successfully created. Now you can check. We have got the thread ID of AI Agent and use a chat ID that is from the webhook, that is this chat UI. I hope you understand this points. Now what I will do, I will come here. I will stop this one. I will restart again for the new chat I will send a message that is high. Let's wait for agent response. You can see we have got the response, Let's check our database, refresh it again. Now, you can see it is created a new thread ID, but our session ID is same because it is per user, but thunder is keep changing. So we need to clear that one, remove this one. This record is the latest one. Now, we will keep it as same. Let's jump into scenarios. Let's see this one. Edit. Let's go edit. No. Before sending an message, we need to add we need to check whether the user is stored in our database or not, right? So we need to add the data store again here. Click on this setting button, come into add a model, certify the data store again. So we need to search all the records because whenever a user start a new chat, our scenario will check whether this particular user is Already in our data store or not. If, yes, it will recognize my name, otherwise, it will just run the rest of other actions. I understand these points. Let's see the practical implementation. Let's select for the database. We have just created Agent DB here. And for the filter, we need to select user chat ID. For example, if user chat ID is equal to session ID. So then this particular user is in our data store already in which the AI agent will recognize our user name and it will give the response. Otherwise, it will run these other actions like new user. Okay, I understand these points. Now, when this workflow search the record, whether the user is already in our database or not, I, it will just give the key. Okay. So we need to add a model that is check the existence of a record here. So it will search for the records. If this condition will match that is user chat ID is equal to the session ID, then it will pass a key. We need to keep this key here, right? So you can select from this data store that is key. Just click here key. So we need to choose a data base we have created. Okay, we need to search for this key here. Let's say this one. Fun now. Now, what happens here? Okay. Now, let's save this one. Let's check whether it will works or not. I will send a message like my name is. Let's give the name. Okay, it is remembered my name. Hi, Sef. Nice to me, too. Could you please share your emails? I will ask another question not related to my name. Let's take what is your company name. Company name is also is given the answer. No, I will ask what is my name. Let's see whether it will recognize my name or not. What is my name? We need to add the threaded here. This is the most important part. We have just forgot, right? So for the thread ID, we need to use this record thread ID, right? So when this existing, for example, when the user gives my name, for example, when I give the name, the name is with the session ID, this search record action will check whether our ID is already in our data store or not. If it will pass a key, right? And when the true itself, the AI agent will get the thread ID from the stored database. You can see such records from the thread data, it will come. Now, after that, it will give the exact thread ID that the previous one it has used, right? It has used. That is simple. Okay, now it will add the record. When the agents give the response and user query are stored the data store again, it will pass the response here. Okay, it may be some confusion, but we will try by sending the message here. Okay, I will stop. I will start from the scratch that is preview. Now we have already said that so you can see this all the thread ID and user chat D. It is the user chat Eddy is same because it is per user chat UI, and this threat D is changing because we have not added the thread ID in the EI agent. It will create a new thread ID for every interaction. So to avoid this, we need to add the thread eddy previously we have done, right? So I will just so let's select all. I will just remove all the records. Cort. Now we have only one record. This is a per user. We already have the user chat it that is session ID, so it will pass the key. Let's come, we will check again. Let's see the edit. Now we will run this again. I will give the same message that is my name is Let's check whether it works or not. Now it is recognized in my name, that is thanks Si nice to meet you. No, I will ask another question like, what is your company name? Let's wait. No, we have got the response. I will ask my name again. Let's take whether this AI agent will recognize my name or not. Now you can see, your name is SIP. Now, you can see here this AI agent is remembered my name. You name is CIP. How we have solved this issue, you can see here. We have just added the threat ID of previously generated for the same user chat ID. I hope understand these points. So we have just added the threat ID from the data store that is search records. Okay, click here, you can get the thread ID from the record. When you check the data store, right? You can see we have got the exact thread ID. Now the thread ID and the user chat ID is same per user. Now the AI agents can remember my name, my previous chat, all those things to give the response intelligently. I hope you understand these points, right? Now, if you have any confusion, try by yourself, implement this whole system in your computer, then you will get the idea how it will works. Okay? We have another setting. So for example, if I clear all the data store data, let's take. I will delete all the previous data records. Let's select all. I will delete all. No, we don't have any previous chart. Now the new user can get the answer from AI HN by suppressing all these actions because it will work and it will send the message to AI Agent when the user is already in our database. If the user is new, how this AI agent can give the response. Okay? Let's check by practical. Now, you can see our name is not recognized because remove the previous data charts, all those things in the AI Agent data store. Now, I will send a message like hi. Let's see whether we will get the response or not, but you can see here it doesn't get any response. We have just got accepted. That means the webhook RL is working, but we not get AI agent response. Why? Because this action is only run when the user is already in our database because we have just add a filter here when the user chat, D is equal to the session ID. Okay, how we can avoid this situation. We will see in the next session. Let's dive into that. 10. 2.6 Adding Agent Memory System to Ai Agent - Part 2: In the previous session, we have seen how we can add a memory to this AI agent in order to recognize our previous chat, like user name, all those things here to give the response more intelligently, right? But this system only works when the user is already in our data store. Okay? If the user is new to interact with our AI agent, then how it can be done. Okay? Now, we have seen the example here. When I send a high message after clearing or deleting all the records in our data store, so it is simply given the accept, not an AI agent response. So to avoid this situation, we can add a router here. So click here. Before that, we can see the output from this one. Let's go with the existing data from this one, previous data. But it will stop here only because if you see the output here, the filter is called nothing, right? Because it is a new chat. The user is new. So this data store has a filter, you can see here. That means the new user is not stored in this particular ID. That's why it is simply written null. Okay? Red is not going forward for other actions. When the user has started conversation, we can add a router here, just click here add model. Now we can add a router here. You can search here like Router, that is rooter, just click in the router. Here we have two flows, that is one and this one. What happens here when the user is already in our database, it will return as a true. If the user is not in our database, it will return the false from this check the is of a record. Select here, you can add a filter. We will just take this. And we need to remove this one as a thread or let's skip this add this one. Now, we need to add a filter, set up a filter. When the condition, let's take when the exist is equal to true so what happens here? This router will just run these actions when the user is in our database. Then it will check if existence of record action will pass it through, then it will do these actions. The rest of this one just work as these conditions. Okay, I have understand these ones. Let's add a filter here as well. You can see check the existence of record from this one. If the user is exist is equal to false. Okay, this is a condition. Let's say this filter. When the user is in already our database, it will just run these other actions. If the user is new one, it will run these particular actions. To do that, we can duplicate this one lights, clone. Let's I will delete this module though it is connected automatically. Here, I will just clear this record the ID here from there. No, it is simple. No, it works for the new chat. Now, I will just simply add a webhook that is webhook response. Let's see the body from the agent response. Let's say. Now let's take whether it will works or not. Let's save this one. Now let's send a message from here. I will stop and I will start from the scratch. Let's take high message. Now, let's take our data store already have any data or not. Now, we don't have any data in this Agent debit database. Now, let's take our scenario. Let's take this one. Edit. No, this is a flow. No, I will send a message from scratch, that is, no. So why let's send a message here. Hi. Now so we have got other issue that is the agent response we are not getting the response from Agent. Why? We have one setting. We need to select this one data store. So go in the advanced settings. So we need to select this as Y. So continue the execution of the root even if the model returns no results, right? When the new eser comes, so this filter will not be run. Why? Because our database doesn't have this previous user ID, right? So in that case, we can use this as setting option in which if this action work or not, it will simply run other actions. Okay, I will suppress this action. These settings will suppress this action. Even this action will return no results. I hope I understand these points. Let's save this one. Now let's save this workflow again. Let's jump into this one. I will send a message like high again. We have scenario filled to complete. Let's check this again, now run once again. Sorry, let's come from the previous or Run once again. Now, let's. Now we have these two are working, but we have got an error from here. Let's check. Now, let's see. Now we have another issue here. Why? So when this particular action, when the user is already in our database, it will pass a key, right? If the user doesn't have and record, so how we can do this. So what we can do, we can use the simple functions here. Click here, write the user if the key is empty, written null. That is simple. Now what happens here, the AI will automatically create a function in which we can use it here, you can see here, you can add the apply, come here, just remove this one. We need to remove this null and we need to write a null here. I hope you understand these points. Let's save this one, save the scenario again, and let's take this one again. I will send a message high. Now let's see whether the Agent the response or not. Sticking time. Sticking time. Let's see here. Let's say this again. We will run from the once waiting for data. Now, you can see, we have got the agent response. Hi, welcome. I'm here to help. Do you have any question about an ECM product warranty set up all those things? No, it is working for the new user. Let's see how the workflow is working now. Now, I will just stop it. I will just run the previous one. Let's run once again. Now you can see it is working. Now it is going this as a new user. Right? You can check here. If the user is already in our data store, then it will take this flow, have just deleted all the previous records in our database. This workflow is taking as a new easer here. When you see here, filter, you can see is equal to false because this user is not in our database. Now, to store the data can also add the data as well from here Clone again here, add this one. So for the thread ID, select the thread ID from this AI agent session ID, keep it as it's safe. Now what happens here. When the new user pams give the message, all those things, this agent will response and it will save this new user to this same database in which it will recognize in the next time, this data store will check the existence of ARCor If the user is already in our database, it will pass this action. Now we have the filter, no it will run this action, this AI agent will recognize my name. Okay, I will go with the new name in which you can understand very well how it will works. Now what I will do. So let's save the Word flow again, the W is saved. No, I will start from the new like, my name is. Let's give as my name is. For example, we will take the name John. Let's wait for few seconds. You can say hi, John. Nice to meet you. How can I help you today? Could I get your email address so I can assist you today? So what I will do, I will just start from the scratch. Let's see whether it will recognize my name or not. Let's run this again. What is my name? I will ask you to AI Agent. Now, you can see your name is John. Now our details are saved in the data store in which the AI agent will recognize my name because this action is checking. This user is already in our database. I hope you understand this 11. 2.7 Adding Agent Memory System to Ai Agent - Part 3: Add another functionality to this particular workflow, particular user who is already in our data store. If the user wants to start a new chat. So if I told this AI agent, let's start in the new chat, start a new chat like that. So to do that. We can add a router here. Click here. Let's add a router. Now, add another router Just keep as a filter. Now, we need to add additional instructions, click in the advanced settings. In the additional system instructions, what you need to write? Like when us ask something like this. Ask like or something like this. Start a new chat. Then reply with new chat. Follow strictly. Basically what happens here. So when the user just tell start a new chat in the chat UI here, who is already in our database, what happens here, it will remove all the previous chats and it will go by scratch. How we can implement this one? You need to write the additional system instructions. You need to add additional system instructions in the AI agent by enabling advanced settings, you can get here, you can write this simple. What happens here when this AI agent run A Agent got any message like start a new chat, new chat like that, I will just simply reply with the new chat here. We need to set up a filter like when the response is equal to new chat. We need to write like new chat. At Zat the Agent response will give is equal to new chat. Then it will simply delete all the previous records. You can simply delete previous records. Otherwise, we can remove this one. When the response is not equal to not equal to new chat, then it will go as it is. If the response from AI is equal to new chat, we need to delete the records. We need to select delete all records. Agent we need to select this one. We need to add a setup filter. Come here if the agent response is equal to the new chat, Say this one. Now, let's add a response. Webbook we need to reply with start a new chat, send a simple message. Very simple. Now, how it works. Let's check whether it will works or not. When the user is already in our database, it will run this blow and it will give the answer as based upon our previous records. When the user want to start a new chat, what happens here, it will check the filter. Okay, the agent response will give the new chat as a data from AI Agent response because we have written in the additional system instructions, like you can see here, when the user asks something like this, start a new chat, then reply with new chat. Follow this strictly. That is simple. Now, it will give the new chat only. This new chat will see the filter. When the response is equal to new chat, it will remove the previous charts and it will start from the scratch. If the response is not equal to the new chart, then it will go as it is. We need to run this again, save this one. Let's see whether it will works or not. Let's ask another portion like what is your cramping? N. Wait for agent response. You can see our company name is home appliance headquarters, Bengalur India. Now, let's ask what is my name. You can see your name is John no it is recognized in my name because our previous chat and our data is stored in our database. Let's start a new chat. Let's see what happens here. I want to start a new chat. Let's say whether the A agent will get our point or not. We can see started a new chat, send a new message. Now what happens here. Basically, we can write the message from here like now, what is your company name or all these things? No, we can test what is my name. Now what we have done, we have just simply deleted all previous records because we have the delete record data store action here, like that. I will tell what is my name. Let's wait. You can say I don't have your name on file. Now, we have successfully deleted all my previous records by simply just sending a new chat. This is how we can implement this system. I will explain again how it works. For example, we have two users. These a agent have two users. One is newly one, one other one is previously already chatbot user. Now for the new one, when the user send a message, it now data store will check whether this particular user is in our already database or not. If no, it will return no results, but we have selected settings option that is continue the execution of the root, even if the model returns no results. No, it will simply pass no resuls but it will just go for the other action. Now what happens in the data store? This data store action, which check the existence of acorde. If the user chord is in our database, it will return a true. Otherwise, it will return false. When the user is new, it will simply run this function. If MT k is null, then it automatically go for the other action. Okay, now we are talking about the new user. Now we can come to this one. When the record is false, because the user is new, it will run this AI agent. This AI agent have that for every new user, it will create a new thread. Now in the message, we have got the user message from the webhook, that is simple, right? No, it will create the data for this particular new user with the generated AI agent thread ID and the session ID from the chat, from the webhook, that is save in the database. It will just give the response to B to this chat UI, right? Now if the user is already in our database, what happens here, user send a message who is already in our database, this action will run. Why? So this filter have one condition that is if the previous user in our database, it will check that is equal to, then it will pass a key. We have function. It will check the key. If the key is not empty, that means the user is already in our database. So what happens here, it will return the true option. When the truit comes, we have another filter in the router, that is if the exist of record is equal to true, it will run this flow. When when the user give the response, that you can find it here. But the AI Agent thread ID, we will get from the search records. We have just using here because in this search record action, we will check the user thread ID and Agent thread ID and user chat ID if our of previous session in which we can use the exact threadD in the AI agent in which this AI agent will recognize my previous data or the chat. In the message, we will just give our webhook message. Save this form. Now what happens here? If the already existed user will have the two options, they can go with the asking us. When the user start a new chat, when the user tell to AI agents, let's start a new chat, what happens here? Let's check. You can add additional instructions. You can in advanced settings, you can see the additional system instructions. You can say when user ask something like this, start a new chat or new chat, then simply reply with a new chat. Follow this. What happens here? This AI agent will generate a response new chat. Okay. Then we have the filters. We have set up additional filters here by using Router. If the agent response is equal to new chat is not equal to new chat, that means a user is not asked to start a new chat, right? That is simple. Then it will just as usually give the Response to the user query. Otherwise, if the user ask to start a new chat, then AI agent response will be the new chat. Our condition is new chat, then it will match the condition that is equal to. Then it will start work flow like this. Now what happens here in this data store action, it will simply delete all the previous chat records, even our name, all those things here, right? It will just give the response here. Okay, you can see started a new chat you can send a new message. Okay. This is how we can implement this chatbot or AI agent whole memory ecosystem in which we can store the previous chat of a user data, and even we can delete all the previous records when the user want to delete all the records or start new chat, all those things can delete all the previous records or a specific record from that particular chat. So the best option is you can simply delete record not all records. I understand these points. Let's take and let's add a model. What happens here. Previously, we have just added the delete all records action, in which it will not recognize my name, all the things. It will record all the records for that particular session ID. But what I'm looking to that, I need to only delete the record, which is simply just asking before it. Like, for example, I will explain. So previously, we have used this delete all records, but now I will just to go with the Delete record. Delete the record, which is previously not all the records, can select the key which is in the data store, search records. That is key, just click here key. Now we have successfully. We need to select this Agent DB. Now it is successfully add. So if you use delete all records action in the data store, it will delete all the records. But if you choose delete a record, specific record which started previously, not all the records, then you need to use the key. Okay, this key start from the search record. We will get from this here. Okay. Let's say. So if you have more confusion, try yourself, then you will understand how this will works. Okay, let's check it again. What I will do, I will just run this again from the scratch. Now, I will give my name. My name Sy okay? I'm a new user. Nice to meet you, S. Thanks for sharing your name. Now what happens here? Now I will ask another question not related to this topic, like what is your company name? Let's send this message. We can say our company name is sons. It is recognized my name side. I will ask what is my name. No, your name is Sif. Now we have successfully added the memo in which this AIGent will recognize my name all the previous chat. What happens here? I will tell to AI, start a new chat. Start a new chat. Started a new chat, send a new message. Now again, I will ask my question that is what is my name. Let's take whether this AI agent will recognize my name or not because we have used the delete a specific record, not all the records. Your name is side. That is how you can add the functionality. In the previous one, we have just using the delete all the records. In that, it also removed my name, all those things. I hope you understand these points very well. Now you can click here that is grid. Before that, you can rename your chatbot. AI Agent chatbot. That is simple. Save this one. Now, come to the grade system, which is available in the pro plans only. Now you can see it is how it looks like. You can check the service email Human escalation. This is a tool, this is a Agent chatbot AI Agent we have created, and this is the database, and this is the webhook. You can go with the select layer as well. We can check this data center. Up to now, we have successfully created the single channel AI HD, which give the response according to instructions and the company data provided as a context to AIGN. And we have also successfully added the omen escalation method when the user ask a question which is not in the knowledge of AIGen, it will simply send that particular user query to the Women sales team with their user name and email, and we also added the agent logs by using Google Sheets in which we can track the AI agent response and user query, and we also added the memory system to the AI agent in which the AI Agent give the response more intelligently by remembering our name or previous data. So we can start a new chat, delete cord or add all those things we have seen up to now, we have just created the single channel AI agent. Okay, in the next session, we will discuss how we can create the exact system for the Telegram, in which we can handle the user queries, customer queries directly in the Telegram. Okay, let's dive into that. 12. 3.1 Building Telegram Bot Ai Agent - Part 1: Now in this session, we are going to see how we can create this exact agent system for our Telegram out in which we can handle the customer queries through Telegram. Okay. Let's see this. So we have already created the exact same Telegram bout in our previous Entertain mastery course. If you want, you can follow the neten mastery course as well for better understanding. So let's come to this so we can build the exact AI Agent system them for Telegram board and Telegram board will work similarly, but the input trigger and the output will change, right? In this scenario, this is custom chatbot, AI H, in which we will receive the user query through this chat externally and we will send a response to this webhook, and we will get the response here, right? So now we are going to build this exact AI agent from a Telegram board in which we will just add a Telegram action here instead of webhook, and we can receive the message from the trigger when we add a Telegram here and we can send the exact response from the AI agent to the Telegram action when we add here. Okay? It is very easy. We need to change some fields in the actions as well. Let's see how we can do this in this session. Save this one. And now let's come to another scenario. Let's create the new scenario. I will come to the folders that you have created and come to create a scenario again. Now we are in the new scenario builder. You can give the name like Telegram AI Agent. That is simple. Now let's add a trigger. What is the trigger? We need to watch the Telegram messages whenever a user send a message through Telegram board. Now we will find for the Telegram here, let's search for Telegram, you will get this Telegram board, select this action. Now we need to select the watch Telegram updates like this action. I see this one. This is the action we need to add as a trigger. Whenever a new update from our Telegram bod, it will watch the updates or messages, and it will send to further actions. That is simple. You can take the watch updates. Now we need to create a hook, I have already connected one, but I will just create an add again. Or let's give the name like I will just keep it as same, add it again. Now we need to add it to open here. You can add a number of bots, but I will show how you can create the new one. So I will just go to my Telegram account. Let's find this story grab. So I'm in the Telegram. I will search for the bot father. I need to go if you already have an account, just sign in with that and search the board father in search bar, you will find it this one. Come here again, and I will refresh this chat or I will delete the chart again. I will start from new. Now I will search for the bot father again. Now let's see we are here to create a new bot? Let's get started. Start. Now we need to choose this new bot here. Let's give this one new board. Please choose a name for your board. Let's give the name for our bots like AI support Agent one. Okay, let's give this name. When we have successfully named our board that is AI support Agent one, let's create the user name. You can see here we need to add a bought at the last. Well, let's give this one again and we will write Devote board. Let's see whether this username is available or not. Now, I will just find my board in this section by clicking Link here. Now, we are here in the bot, right? So before that, we need to copy that token here. This is the axis token we need to use to connect with our make automation. I remember one thing, don't share this token to anybody. Please keep it private. I will save this one and I will save this. Now let's tech. Let's give the name again here. A new connection. Let's wait. We have successfully connected this webhook. Let's save this one. Even you can see the webhook you on here, that is Telegram bot. Let's save this one. Now we will just save this. I will send a message. Now before that, I will run once. Now it is waiting for data. I will send a message here. I will go to my bot. Let's start this one. Now let's see it will works or not, I will cancel this board. That's come. So it is successful trigger is successful iron. You can see the output message here. Just click here plus. This is ID. This is a text. So we will get the number of fields from the Telegram. So in order to observe this particular text, we need to run this trigger once in order to see this is the exact text field we need to use to message AI agent. Okay? Now, we have successfully created a Telegram bot here as a trigger. Now we need to copy the previous custom chatbot AI Agent scenario and paste it here, and we need to remove the webhook. We need to connect this Telegram bot as a trigger instead of webhook. Okay, you understand these points, let's do that. I will save this scenario again. I will come back to my previous scenario that is AI Agent chatbot. Click on the red button, what I will do I will just copy the blueprint to clipboard here. Just click here. Now I will just go to my new Telegram bot automation workflow. This is I will come to this again. Click on the red button. No, I will paste the whole blueprint here. I paste. No, let's check. This is the whole workflow I have just copied and pasted here. What I will do I will just remove this model webhook model. This, I'll just take this. I will paste this again. Now let's click anywhere. I will remove this webhook. I will delete simply delete that, click Deletm model. We have successfully removed that webhook and we will connect this webhook to this. Now we have successfully connected Telegram as a trigger to this particular flue. Now, if you see here, we have some error. We need to remove this session ID. Remember one thing we need when we change the trigger as a input in the previous session, we have used the webhook webhook that webhook have the session ID. You followed my previous chatbot agent creation. The book have session ID. We are connecting this Telegram bod as a trigger. So now we need to remove the session ID, and we need to find the exact ID for our Telegram bot. Like you can find it from message and find for the chat and just click the ID here. So this is the message chat ID. You can see this when you see this particular ID in the film, you can find it here. Just go in the message bundle and just see the chat bundle in that you can find the ID. Okay, now, so and rest of all the settings, you can keep it as same. Even if you can add another database for this particular Telegram AI chat agent, if you want, you can add it from here. No, I will just keep it as same for the database, so there is no changes, so I will just save this one. So now the Iran is gone. Now, let's check this data store. When particular user is already over in database, then it will pass a key. So this key is exactly the same. You can get from the search records. You can get it from your keys. Now come, come here in the filter option. So if it exists, true, keep it as same, and this is also a correct because if it exists equals to the false. If the user is not in our database, then this particular user is a new When the condition is passed, then it will run as a new user. Now, this flow is for new user, and this flow is where the user is already in our database. There is no changes in actions, all those things, but we have to change the input and the output and fields according to the trigger and the output. That is simple. Rest of that function of the Telegram agent works similar to our previous customer support chat agent we have just built in the previous, right? So we will just change the fields according to our trigger. Okay? Now we can keep it as a trend ID as the same. Now we need to change the message. This is our most important part. So for the message, so we need to sell the chat from the telechrom bot. You can see you can find the exact message in the chat, you can get the text here. You can see that message from the Telegram that is start, we have sent this start in this Telegram bot. You can see the exact field from this Telegram bot fields. You can see this is the text. Just select this one. We have successfully attached this field. We can pass the user cody from Telegram to this a agent through this message text field. Now let's keep it safe. And for the thread ID of AI agen, you can keep it as same like before there is no change in these settings. Now, come for the new user. You can do for the thread ID, we will keep it as empty because when the new user will come, it will generate new thread ID. Now for the message, exactly, we need to remove and we need to take from the Telegram bout, you can select the text the same again. Save this one. Now let's come to this here. And for the filter, for the response, we will take from the make AI agent. So there is no change in filter settings. Let's save this as it is now come to the data store. No, for the data store, thread ID is keep it as same like we have done in the previous. But for the session ID, we need to change according to the Telegram bot trigger. So we need to select the exact message ID that is message, chat and ID. This is the one. Okay. Let's say this one again. For the delete a record, we need to save as it is, and come here in the data store, add code. Now, whenever a new user send a message, this A agent will create a new thread ID with the user chat ID. Okay. Then we can save that particular ID, which is a newly created ID a agent in our data base that is trend, it is correct. And in the user chat ID, we need to remove the session ID and we need to add a Telegram message chat ID. You can find the same thing here, ID. That is simply save it. Now we have successfully changed the fields according to the Telegram bot. Trigger. No, we need to add a output. I will delete the model and I will select a Telegram bot. Then we need to send a text message or a reply. Whenever user send a query, agent will send a reply to this action. No, I will for the chat ID, we need to use the Telegram chat ID like message message, chat, select this ID. That is simple. This is a chat ID for our Telegram board this ID will come from the same user the message has sent. Okay. And again, this Telegram bot will find this particular ID, and it will reply to this particular ID in which the user can get the response for their query. I understand this point? For the text, we need to select AI Agent response. That is simple. Now, let's say this one. But the webhook, we need to use this message because we will just sending the message whenever user will ask start a new chat. You already know this function how it will works. We have already discussed in the previous chatbot creation. Now I will just delete this model. Now I will select the Telegram bot and send a text a reply and I will select the message chat exact one we have done in previous in the earlier output Telegram bot that select for the message, charge ID. For the text, we need to use this one. Started a new chat, send a new message that is simple. I will save this one. Now for the new user to send a response to that particular new user, we need to remove this model. Before that, you can check the response, right? Dilative model, add Telegram bot, a replay message. Now for the chat d, we need to use the exact chat idea in which the user has sent a message. Set exact chat idea again, select that one. But the text, we need to use the AI Agent response. That is simple. Now we have successfully connected the Telegram to our support AI Agent in which we can connect all the triggers and actions with changing the fields according to the Telegram bot. Trigger. Now let's save this one and let's check whether our bot is working correctly or not. Now, I will just keep it as immediately as data arrives. Let's skew this one. Before that, I will clear all my data records in data stored before checking this one because it can cause some issues. Okay, sorry, we have simply deleted all the data store. So we need to connect this one again. Let's come with the suct Agent structure. Let's create this data store again. Okay, we have simply deleted all the data store. I will create a new data store. Don't worry about that. A Agent DB. Let's keep this 13. A ag DB. Now we have successfully created the third ID and use the chat ID. Let's come to the scenarios again. Just click on the Telegram AI Agent again, Adi. Just check our databases, that is correct. Select this one again. You need to all the records in that particular database. But unfortunately, I have just deleted my database. No, I have already created that one, you can follow the same thing here. Let's choose a database again. It is taking automatically the fields, that is good. This one. Database. Save this one. Now we have successfully added our newly created database. Now let's save this one. 13. 3.2 Building Telegram Bot Ai Agent - Part 2: Now, let's send a message from here. I will start from the scratch. I have deited all the previous conversation. Let's see that it's confirm. Now that is simple. Let's come to the scenarios again, select our Telegram AI Agent. Click here, edit. Now, let's see, we just make it this scenario active. I will just interact with this run once again. Now it is waiting for data. I will send a message as a new user. Hi. Now let's run once. Waiting for new data. I will send a message here like hi again. Let's sort out the issue. You're not getting the response from AI Agent for why. Let's clear this. It's taking time due to the Internet connection. Our new body is this one. Come to A agents again. Now it is working. Now we have got the response. Now it is we'll check out. Now we can see we have successfully got the response from H for the three times because we have sent high two times now it is working perfectly. Come here. What can do here? Just come to the data store again. Or come here data store, click your database, simply select all the records, delete all Confirm. Now come to the scenarios again, select our bot. AI Agent bot, click on the dit. Now we will make it live, so it is already in the live position. Come to the AI Agent again. I will just delete previous chart, all these things. Now we will search our bot in the bot further, click on that our bot. No let's start so we have sent a message. Now I will write my name is, let's send this one. Now you can see, hello, it is working perfectly now. So hi, welcome to the ACM homoplasss. I'm happy to help you. So it is giving the response for this chat. And for this chat, it simply given high Sci hi, nice to meet you. How can you help you today? I will ask a query to the AI agent, which is out of agent knowledge in which we will try and we will check whether this AI agent is capable of sending a message to our sales team or not. Okay? Now, I will send a simple I need 90% discount on all products. Let's send this. Let's wait for a few seconds to get the response from AI Agent. Now you can see, I understand you are requesting 90% discount. Please before I escalate, please provide this data. It is working perfectly. I will just give my name. Email. And equality, it will automatically takes now escalation. Let's send this message. So I have sent my details, name, email, no message. Now, let's check our email whether the AI Agent is simply forwarded the message or not. Now you can see the response from A Agent. Thanks. I have accelerated your 90% discount request. Okay, now let's check our mail here. This is the email now you can see we have got the email. That is AI prom from the company name. Come to this customer name IV, customer email, and the request customer is requesting 90% discount on all products. So you simply created the email with the details that the team member can directly reach out. Not only that when I tell to AI start a new chat. When I send this message, no this AI agent will send a message like started a new chat, send a new message. That we have seen exact working of function in our previous one. You can see started a new chat, send a new message. So we have successfully created the exact chatbot agent we have done in our previous custom UI chat, AI Agent can also check the Google Sheet whether this Telegram AI Agent is tracking the agent logs or not. You can check all those things here Google Sheets as well because we have connected the exact sheet to our previous chatbot Agent. Okay, now you can see, when I give the name, it has simply told accelerate repeat to learn with the subject, customer excel, 90% discount request. So it is simply stored my input, output as well, right for the cantigram bot. So we have issue that whenever I send a message, so we are getting the two times response from the AI. So to avoid this issue. We need to set up a filter to avoid that issue. Now, let's set up a filter first here. Come here after trigger, click on the set up a filter. For the label, I will take as ease user for the condition, just take from the Telegram bot, find the message, fail and from this ID. Okay. For the condition, just take as it does not exist and click on the save. What causing issue here that we are getting multiple a agent response from this workflow, even the user sending only one message. White happens? Basically, we are using the bot here for the messaging channels like Telegram bot here. When the user use our Telegram bot, the user will send a message this bot carrying the user message in this workflow, this user message will be pass to other actions and it will send to AI agent. Now this AI agent will give the response for the user query. Again, this AI agent will send to this watt according to our workflow, this at is carrying our AI agent response and it will show to the user. That is simple. Now for the next time, when the user got the AI agent response, again, this Telegram bout will trigger this workflow, right. This is a issue. Okay. Now, the user interaction is absence, but the Telegram bout interaction will be started, which causing run this workflow continuously in which it will burn out all credits and operations. So to avoid that issue, we have just added a filter, right. So you can see, I have just taken the label name is us. This is a condition we just use it from the Telegram bot. What is that condition? If the trigger is from user message, it will just pass it. Why this user doesn't have this bot ID, which does not exist, which satisfy this condition, and it will just run the other operations. When the next time the Telegram bot will trigger this input, what happens? Now, this Telegram bot will have the Telegram bot itself. You can see here. We have just got from the Telegram message field ID, which is exist, which is opposite to this condition. Now, this workflow will be stopped here because we had added a filter. I hope you understand these points very well. When you face this type of issue when connecting you messaging channels like WhatsApp, Slack, Telegram, or other, so you need to very clear that you need to add a filter. Okay, if you want any additional information, how you can add a filters to stop this issue, you can find the solution in the make.com community, or you can use the ChaGPTCloud or other EI MRs to get help, and you can search in the YouTube to solve this issue as well. Basically, you need to add a filter, search for the bot ID for this Telegram as usual, and you can also search Bot ID in the Slack or even in the WhatsApp. Just use the ID. Keep this condition, save this, and you can solve this issue very easily. So this is how you can create the multichannel customer support EI agents through different channels like Telegram, bot, custom UI, and other as well. But the Telegram not only that, you can add a WhatsApp and you can add a Slack or other messaging platforms by yourself. You just need to connect preferred messaging channel as a trigger, and you need to change the fields as well, actions, data store actions and the output actions as well. Try by yourself for more advanced features. Now we have successfully created the Telegram bot, try yourself, and in the next session, we will see how we can create the exact support agent for our Fb submission method. Okay, When the user submit a form in our website or Google forms like that, so we can send an email to that particular user in the next session. Let's dive into that. 14. 4. Building Form Submission Ai Agent: Session, we are going to see how we can build this exact AI agent for our customer support forms submission method. Okay? When the user submit the name, email, and the query, so the AI agent will automatically take that query and it will reply to the user's email which they have provided here. Okay. So by using that memory system, it can recall the name as well. I hope you understand these points. So by doing this, you can build the multichannel customer support AI agent. That is very easy. We are going to see how we can create that particular exact AI agent for our form submission method. So we have already created this type of form submission in Net. Let's come into the make in order to create the AI agen for our form submission method. Let's start that. We are going to implement the exact strategy like we have done for the Telegram board. Let's come into the scenarios. Now I will start, create a new scenario. Instead of that, you can clone this AI agent. Why? Because we are going to use the webhook coding. Like that, for example, let's see our previous custom chat UA Agent. So we can replicate this whole workflow blueprint. Why? Because for the on form submission method, we're going to add a webhook, right? Form submission is externally integrated with the website. Okay, we need to use the webhook URL to receive or to take the user name, email, query, and send it to the AI agent. So for that we are using the webhook to receive the data from external sources like this form submission UI. We will keep it webhook as it is, and we need to change the webhok URL for that. But the output of the form submission AI Agent is different, right? So when the user will fill the form and submit, this AI agent will give the response through email. For that, we need to chin the output. In this case, we are just copy this whole blueprint to clipboard, paste it in our form submission method, AI Agent. Let's compute the scenarios again. Let's create a scenario. New scenario, you can give the name. I will just give the name as form submission. Agent. That is simple. Now what I will do, I will just paste it here. Let's let's choose a model. I will take a webhook. Otherwise, you can paste it paste. Add the webhook. We will create a new webhook. Let's say this trigger is okay let's do one thing. I will delete model. Delete anyway. Let's take this one. Let's select the webhook again, custom webhook, create a webhook and you can give the name as form Agent, like that. You can the name as you want. Now, this is a URL. This is the most important part, just copy this URL, come to this code. I will provide this code in the document itself, you can follow exactly the same. I have just generated by using ChachiP I am providing webhook URL, please, can you create a UI form to send a user query name email to the make webhook. I will provide this code. You can use by yourself. I will just replace this webhook with the newly created one. Okay. Let's save this one. Let's go to the premium mod. Let's come to this one. When I send the details through this form, it will show the data is determined successfully. Let's see this one. I will send my name email for that, I will take this one for the user query, I'll take high. Let's send this one. Now you can see the form submitted successfully. Let's jump into this. You can see it is successfully data retm. That means our webhook is working correctly. There is no issue in that. Can let's save this form? Now, we will add these two. Okay. So as I said, we need to change the fields of data store, memory, AI agent, all those things because we have changed the trigger that is web home. Okay. So you can check this data store. As I said, you can create a different database for different AI agent methods. So for the form submission method, we do not require any memory system because we get the name from the form submission itself. Remove this one. We can directly use the AI Agent without memory system. Okay, but specifically form submission method because this is not a chat UI, but it is simple form submission. We will send email to the user, delete a model. We do not require any memory system to this particular AIHen We can also remove this router because we do not require all those things. I will remove this model as well. Let's keep this very simple AI chen. Let's delete this model this model, this To. This and this. Even if you want to create this memory system, but the form submission method, you can do that, but you need to integrate session ID. But most of the form submission methods, we do not require any session ID or like that because we get the name of the user from the form itself. For that, just we will use that webhook as a trigger and make AI agent for giving the response. We will keep the same AI agent we have created with the tools and Google Sheets, we will keep the same as it is, and we'll remove the thread ID. We do not require any thread ID and we also remove this additional submissions we do not require. But the message, we need to take from the webhook that is query. This is a user question we will send to the AI agent in order to get the response. Let's say this. Now our next step is to send email to the user. That is simple. Let's find out the email here. Or you can find the Gmail. Can find the Gmail. Nobody to send an email or you can go with the reply to an email. I will take the reply to an email. For the M Gmail connection, I have already connected so you can do add. You can sign in with the Cuckoo. Let's click here. Let's sign with the Go. I'll go with this one. Continue. Continue. Successfully connected my Gmail account, we need to use the send an email method instead of reply to email because our trigger is the book not a email. Let's delete a module again. Let's create the Gmail, send an email. That is most important part. You need to select your account you have connected and for the two, we need to add a recipet. How we can get the recipient, you can get the recipient from the webhook, that is from the form submission here from the user email. We can select from the webhook email that is simple. Now, you can add so many recipients if you want, can add your support team or AI Agent developing team email in which you can check the AI Agent response, but the subject, what you can do is just write reply to query. But the query you can take from here. That is simple. You have the two fills that is raw historial collision of contents. If your body contains the signature signature content attachments, like text, image, file data, you can go with that, so it is a simple reply of AI agent. For that, I will just take with the raw HTML. For the content, you need to select the make AI agent response. That is simple. Let's save this one. We have successfully created the form submission EIgen. Let's check. Save this again. Now, let's live this particular scenario immediately as data arrive. Save this one. Now I will send some my query. I will give the name as email further go with the Okay, I will use this email. And for the query, I will ask a question out of AI Agent knowledge like we did in our previous Telegram chatbot and custom UI chatbot. Okay. So I will ask a question which is not in the knowledge of AI Agent. I need 90% discount. All products. Let's send this one. Let's see the working of EIH. You can see the form submitted successfully. Now I will check my Gmail that I have provided here whether I got a mail or not from EIH. This is the email I have provided. I will refresh this one again. You can see we have got the mail from the AI Agent that is reply to I need 90% of discount on all products. Thank I can help with request. I don't have any information about documents. I will explain request to a team before I send explanation a new full name, email address. It is asking full name and email address. Again, I need to send my email address name, but it is not recognized because we have just removed the memo system. But how we can do that. In that case, what we can do, come to the mag.com. You can write the additional instructions, let's see, click on the advanced settings. And for the additional system instructions, when user ask question or question which is out of your knowledge. And so before that we need to gather when user ask question which is out of your knowledge, just send an email team which user details. Now you can take these details from the form webhook.This is the power of additional instructions in the map. So I will take as a user name as we can dragon drop from the webhook, Let's dragon drop here. But the email dragon drop the email. And quitting. Simple. Okay. Save this one. Not only that you can write like this, when a user ask a question which is out of your knowledge, just send an email. You can write just craft and send an email to team with user details. User name, name, email, these are the data we have taken from the webhook. Save, I will check again, submitting the form. Come here. I will send a message again. I will take the exact Gmail ID we have used. But the query I will write, I need 90% discount. Send this one. Now form submitted successfully. Let's check our Gmail whether we got the email from Agent response or not. We have got the mail from reply, I need 90% discode. You can see I have sent your request to our team for review. A member of our team will get back to you with information about discods and any available promotions. Can it confirm your full name and the best email or phone number to reach out? In this case, need more details. So it is asking to confirm your full name and best email because it is not asking email and name because it has got my name and mail from the additional instructions of AI gen. We have used the webhook details like name email query, right? Now, let's check our support team mail, whether the AI Agent send an email to our team or not. Now you can see we have got the email as well. Customer name SIFT is my email that I have submitted in the form, and the query is, I need 90% discount, please advise whether such discount is possible and any applicable promotional policies or not. So I can reach out this email or even if you add a phone number here, so even you can get the phone number of a user as well in which you can directly reach out phone number. This is how you can mail, right? We have successfully created the form submission. So before we use the additional instructions, we have got the AI agent is asking me to submit my full name and email address. You can see here please provide full name, email address. Now when I use the additional instructions of AI agen, when a user asks a question, which is out of your knowledge, just craft and send an email to team with user details because the user is already submitted the details, user name, email. So I will just dragon drop from the webhook. So this is email, query, and email. Then the agent will automatically take that email and all those things it will send to the team like we have seen in this email. Okay? This is how we can build the form submission method, AI Agent. So not only that you can gather so much things like different data from user in the form, submission method. So just you need to use the webhook you are created in the make.com. That is simple. Not only that you can add memory system we have did in the Telegram board or in the custom UI hat board, you can add it by Uself for more advanced features. I understand these points. We have successfully created the AI agent for our form submission method. In the next session, we will create the exact A agent for email handling. Like when the user submit a query through email, then this AI agent will handle that particular email also. Let's see in the next session how we can build that particular AI agent. Let's dive into it. 15. 5. Building Email Handling Ai Agent: This session, we are going to see how we can create the exact AI agent for our email handling customer support. That's in this session. This is Osmsion Agent. We have already created the exact AI agent in the Innet and also. You can see we have just added the Gmail trigger to AI Agent. This AI agent will give the response to that particular email from which this AI agent got the message. Okay, I hope understand this point. So these are the rest of that are the same, but we need to change the input trigger and the output. That is simple. Okay, we can create the exact Agent workflow in them.com as well. Okay, let's create in this session. We can follow the two methods, create with the memory system. Otherwise, you can go with this form submission as well. I will just show how you can create the exact email handling support agent by using this particular method. Even if you want to add a memory system to your particular email handling agent, you can follow the Telegram bot or custom UI chat agent. You can just copy that, paste it here, change the fields according to the trigger and output, you will ready to go. Now in this session, we are going to simple email handling AIHen. Now for that, what I will do I will copy this model only. Let's jump into scenarios again, save these changes. I will create a new scenario. Before that, when you see the folder multichannel customer support agent, it is showing only two workflows. We need to select then we need to create a scenario in order to get in this folder. But we have three scenarios, right, to dragon drop to this particular folder. When you click here, so we have three agents in that folder. We need to select folder, then we need to create a scenario in order to get that particular scenario in that particular folder. Now we are in the new scenario builder, so I will give the name as Email AI Agent. So you can give scenario name according to your requirements. I will just give the email AI agent get started. So I will just add a email trigger because when user send a message to your Gmail or company email, let's take an example, I'm using the Gmail. Okay. I will select the Gmail. Now let's take the Gmail. No, we need to use the watch emails. So we need to add a watch emails as a trigger. You can see how it will works. It works. Whenever a new email received in our inbox, it will watch that email, I will give the response to that. That is simple. So we need to add this watch email action, we need to choose our account. We have already created. That is, this is my Gmail account. So what is? This is a company Gmail account. Okay, this email is shared with the user in order to send their queries to that particular email. Okay, I understand these points. Let's choose an account. Okay, continue. Let's sign in with the company mail. So it can be the support mail of your company or like that. So let's save this one. Now we have successfully connected our Gmail. So for the filter type, just keep it as simple filter system folders in box label we will see in the next session and for the criteria, all messages mark Emil messages read when fetched, so just keep it as. But the limit you can change up to your requirements, and we just take up to f or like that. So it will give the response up to five. Immediately. You can see you can put the limit here, so you can go with the five and according to your requirement, just go with the five. Let's save this one. Now choose where to start from now on. Save this one. So what happens here you can see here every 15 minutes. So it will watch the emails every 15 minutes. When the new message received, then the agent will work and give the response to that particular user email. That is simple. That is simple. Now, let's paste the AI agent. This is the AI Agent we have just copied from our form submission method AI Agent channel because the AI Agent is same, but we are here to create the multichannel customer support agent. The instructions are same. Tools connection like we have added the Google Sheet for tracking the AI Agent response and user query and the tool that is when the user asks a question, which is out of knowledge of AI Agent, it will send email to the human sales. Okay, it is rest upp that same, but we need to change the additional instructions, according to the trigger. That is simple. So we will remove these additional instructions. We will check. We will update later. Let's focus on giving response to that particular email. Let's add a output action to this particular A agent in which the agent can give the response. Okay? So we need to add a email again, Gmail, reply to an email. Just click Reply to email or send an email. You can use the send new email as well. I will just go with the reply to email. No further thread ID, we need to select from the GMA. We will get this thread ID because it will reply to that particular user email from where the query has received. But that we need to choose the thread ID of that particular user mail. So we can find it from here. GM thread ID, that is simple. I have to understand these points very well. So, you can follow the exact thing here. For the reply to all original sender, body time, just keep it as raw HTML, and for the content, we need to select the response from AIHN. Just save this one. No. We have just changed the trigger and output action. Now, save this, something unusual is there. Okay. We need to remove the previous query and we need to add from the Gmails. Find the text here. You can simply send a mock message. Run once, ignore warnings. Let's check again. For the query, I will remove this. I'll save this again, run once. So let's see the output operation. So it is taken one credit. Now, let's just click on the AI Agent for the message to send user query to the AI Agent. We need to select a message here from the user itself. You can find out in the from subject, reply, receive. You can check in the message folder or you can select the full text body or HTML body. I'll just take the full text body. Now, let's save this. For example, if you want to subject, you can add that as well if you want. Let's save this one. If for additional instructions, you can ask like the user ask a question which is out of you want knowledge just craft a sunny we do not require this. Let's save this one. Just turn off the one settings. Let's save for now and for the Gm, it is the same the day which comes from the user mail. Body type content that is simple. Let's save this one, save the scenario again. Now keep it as live for 15 minutes, you can change for the 5 minutes or 4 minutes like that. Let's save this one. Now I will send a simple message to this particular company mail. Let's see whether we got the response from AI Agent or not to my customer mail. Let's assume I am customer. So I'm sending a mail to the company subject and query. For the recipient, I will use company mail. So this is a company mail. You can see this one. For the subject, I will take discount, need it, I will ask exact question which is out of AI Agent knowledge. I will regs, I need 90% discount on all products. Right. Right, I will send the mail. So you can see the message sent successfully. Now, let's check our company mail for clarification whether we got the mail or not. This is a simple company mail. We have got the mail from customer. You can see I need 90% discount on all products right now. This query is handled by AIGD. Let's check my mail, whether the AGR response is given the response or not. This is my mail. This is customer mail. I will refresh this. Now let's refresh. So the message is not arrive. Let's check this again in the agent. We need to wait for 5 minutes in order to get the response. What we can do to test it out, we can run once again. Now the query has passed it to the agent because we have sent the mail right now to that particular company mail. That's page the agent is taking time to give the response. We can see the message sent. Let's take our mail whether we got or not. So you can see here we have got the Agent response to my customer mail. This is a customer mail. We can see tanks I can help. We don't have enough information in our materials, about 90% discount on all codecs. If you would like to accelerate this request to our team review, I will need you a full name and email. I need to include my mail, all those things here, right? This is not unlike the chatbot or Form submission method because the form submission method itself have the user name, email field in which we can send it. Again, I need to send a mail to this particular AI agent with my name and email address to talk with the team. To avoid this, it's simply come here, click on the AI Agent, click on the Advance setting. You can write the additional system instruction like when user asks questions out of your knowledge, send an email to team with user details like name, user name. So for the user name, you can drag and drop from here. You can see the from name. You can take this field from name, and for the user email, so you can select this from email as well. It is very easy. For query, you can take full body text. So that is simple. So we have done like this in the form submission, right? So we need to follow the exact method in order to retrieve the user information directly. Save this again. Now come instead of sending a new email, I will just go with it the previous data. Previous action run once again. No it is taking that previous query that is I need a 90% discount right now. So it is working now. Okay, it is sent. Let's check our customer mail again. I will just refresh it. You can see we have got the message from here. So thanks, I have sent your request to our team for review and record it. When I write the system instructions with Gmail fields, user name, all those things, the response is different. Like thanks, I have sent your request to our team because now the A agent has access to my email address in the additional instructions here, which is taken from the customer main trigger. I understand these points. Now let's check our company supportive mail. So this is the supportive mail. The EI agent sent urgent discode request explanation from the user. You can see user name your products, right? The mail is customer mail is this one. I need 90% discount on all products right now, please review and advise a feasibility or provide authorized response offer. Thank you. When the user asks a question, which is out of the agent knowledge, it simply send an email to the humane sales team, you can see here. So this is our customer mail. It has simply given the response to me as well, and it has sent an escalation mail to the sales team that is here, and we have got the mail in our inbox company mail inbox as well, right? We have done the three things here. This is how you can build the different A agents for your different preferences. Even though you can check our Google Sheets for tracking purpose. So we have just connected the exact tame. Okay. Why? Because we only created this particular A agent once and we have added the tools that is Google Sheet and Gmail. Only once, then we are using the exact AI agent in different channels like we form submission, Telegram bot and email agent. The A agent is same. The tools are the same, but according to our channels, we are changing the triggers and the output. We need to change the details or fields of each and every action according to the trigger and the output action. That is simple. This is how you can build a multi channel support agent. I hope understand these points. Check the Google sheet as well. We have just created in the previous chatbot, AI Agent logs. You can check here. So this whole thing. So we have just you can see I have just a simple I need 90% discount on all products right now. This is the Agent log. So you can check all the logs, and you can try to improve your agent instructions, all those things. Not only the Gmail, you can add different apps like Zendex or email support apps, automation tools like Active campaign, Bravo, there are a lot of apps you can just find it here, add it there, credentials. You can connect each and everything by yourself if you want to add more advanced actions in this particular AI agent. You can add data stores like we did in the Telegram chatbot and Custom I A Agent chatbot. You can add it here. We have successfully created the email handling AI agent, try by yourself. In the nest session, we will try to add prioritize email method. That means you can label the user's customer queries high, low, medium according to the user query. By using sentiment analysis in this make.com and we will try to create the more advanced email handling AI agent in the net session. Let's dive into that. 16. 6.1 Building Advanced Email Handling Ai Agent - Part 1: In this session, we are going to see how we can create the advance email query priority system in which we can solve the customer query quickly according to the user message category, it is a high priority or low priority medium priority, like we did in Enten mastery course. You can see we have created this Gmail handling AI agent that is advance priority system we have added in here. This is we have created in the United mastery course. If you are already watch this particular EI agent system, so it is very easy to understand in the mag.com as well. If you're not familiar with that, there is no problem you can follow in the MG When send a message through our mail. Now, according to that mail, this text classifier will just simply see the user query, and it will divide that particular query into the three different categories like higher priority, medium and low priority. Okay. So according to that, we can add a label JBL, it is a high priority low priority email or medium priority email in which the AI agent will just give the answer or the team can quickly use that particular email, which is high and medium first and after that like that. We can build the exact system in make.com here. Okay, let's see how we can build that in the M. Nowhere in the previous email AI agent scenario. So we will just integrate the advanced priority system in this AI agent here in this workflow, right? Now, for that, we have two methods. Either you can use a simple AI chatbot like OpenAI model. So let's click in this icon, add a model. Now, you can set for the OpenAI. This is OpenAI, generate a completion or like that. You can select any of the action. Otherwise, you can use the inbuilt make AI agents Make AI tool in which it will directly curve we can find that action like classifier AI agent. You can find this make AI toolkit in which we can use these amazing basic AI actions like ask anything extra information from text, or you can use this categorized test, translate test, summarize analyse sentiment. Okay, you can also use this one. Now you can use this analyse sentiment for the feedback collection, whether the review is positive, neutral, or negative. But we are here to create a simple advanced priority system, in which this workflow will simply categorize the emails according to their priority, which is a high priority, medium priority or low priority in which the team can easily reach out to the user according to the priority of user query. Okay, for that you can use the main make AI toolkit categorize text, but to make things simple, we can use the OpenAI model itself. Otherwise, you can use the M AI agent, but we'll start integrating the OpenAI model. But to search for the OpenAI, show more actions, now you need to generate a completion here. Create generates a chat completion or like there. So let's see this generate a completion. So you need to create a connection. For that, you need to add your API key. So how you can get there just come to your account. So if you already have an account, go with that. So in order to use your API key in open E, you need to have some credits. You can buy that. Otherwise, you can add any chatbot model like Cloud or other as well. So I have already this platform.cop.com API, I will just copy that. I will create a new one. But this meg.com, I will just take with the test one, two, three, one, two, just create a secret key. Now I will copy this one. Done. Just come to the mag.com, paste API, OpenAI APK, save this one. Now we have successfully connected our OpenAI with this MC. You can select the model. So I will just go with GPT 40.1 mini. This is system, select this. You can use according to your requirements for the message, what you need to write is simple. User, you can go with the assistant as well or text content like that. I will just give the text content. For the text content, we need to use this HTML or full text body from the gym from where the query has received. I will just copy this one. Now, that is simple. Click on the advanced settings. You can change this sequence, all those things according to your requirements. And now, you can write the prompt here. What is a prompt? Just write this one. Analyze, analyze this text. Just go with the quotation marks, simply reply. You can write like this. Analyze this text. We we need to write a text here. Analyze this text and categorize into high, medium or low. Now, you can write. Now, this model will check this text is a high or medium or low, how we can categorize it. So we can give the extra instructions for high main subjective words like product damage and urgency. So for the medium, you can take like order tracking, order cancellation like and for the low priority, we can write simple customer FAQ or consider other not specify above above category parameters, and just give which type of category it is like high or other. That is simple. Unlike this text, this is full text body from the user male query. Okay? Now, when the user will send a query to the company male, this OpenAI, it will get that particular user email body and it will analyze. So it will works or not let's check it. And for the image input we do not require, let's save this one again. So when the user send a mail to the company email, it will watch email, then that particular email of the body will send it to this action. OpenAI will analyze the user query and it will category into the high, medium or low. If it is a high priority, it will only give the reply like high. Because we have written in the instructions, like, you need to give reply as high, low, or medium. That's it. You can write this follow strictly like that in order to get what we required only. Okay, can write like follow strictly. That is simple. Let's save this one. Now, so come to Gmail in which you use support email for your company. As I said, we have this email. You can check it out here that is info AI proms. Let's jump into this particular mail. This is the email of my company. Let's take. You can see this mail here, infot A PmsmtT is the one. So we need to create a labels here. You can see here the labels just click on the plus button. You can do the name as a high high priority like that. We do not require this one, create. Now you can give the color for this label color that you can select as a red because it is a high priority. Just click on the plus again. You can write a medium priority or medium. Click on the create again, select a color as orange. Click on the plus again. Let's go with the low priority. Even you can give you low as well. Create again. Sorry, you cannot create label name low brat is a result system label. So please try another. Now let's remove this one, just keep it as blue. Now further this one, I can take AJ yellow. Okay, now we have created a high priority medium priority and low priority. Come to our workflow email AI agent. Now, according to this output, we can label our email according to the high priority medium prity or low priority. When the user will send email OpenAI model will divide, it will analyze first, and then it will just give the reply like high or medium or low. Then we need to add a simple model here like router. So what is the use of router is? So we can use the filter options here in order to label our Gmail, in which we can organize the emails of customers according to their priority. So how we can do that. The output of this particular AI model is high, medium or low. So for the rooter output, I will unlink this one. Now click here, unlink unlink here. Now, I will take three different output actions here. So for the first one, I will use the Gmail. You can search for update email labels. Just to click on the updated email. You can find it in your connection. And for the message ID, let's take Gmail message ID thread ID, click on the message ID. Now when you clear that, Labels Add, you can add a high priority. That is simple. Now what happens here? When the user send a message with the product damage or urgency required type of queries, then this OpenAI model will categorize and it will just give the reply as a high ten set up a filter. Now, we need to write this as a condition when the output of this particular OpenAI model. You can take the result. The result of the particular model is equals to high. When this high, it will run. Otherwise, it will run this, otherwise it will run this. Now let's check this one. So message ID, you need to select this message ID from where it will come. After that, you need to add a label to that particular email. So you need to select the high priority. That is simple, save this one again. No come here. Otherwise, you can click here again, add a Gmail, update a email labels. Add a message ID exactly the same we have done for the previous one, select a medium priority. Okay? We need to set up a filter again here, click on this one, set up a filter, write the condition. The result of the OpenAI is equal to the medium. If it is medium, then it will just run this flow. Now, do for the low prity as well. Click on the Gmail, update email labels. For the message take from the trigger. Select for the low, save this one again. Now we have successfully added this 31. So you need to add a set up a filter. When the condition is result of this OpenEH AGBT is equals to the loop, then it will simply run this flow. We need to write the instructions very clearly in order to get the exact result. Okay, just to give type of categories like high right like medium or no. Just remove this other follow strictly. This is simple, let's save this one. Now we have successfully connected this system. This system works like this NA ten if you see this one here. So this text classifier is there in the NA ten, but we can use this particular simple AI model, help understand these points. Okay, you can add this one. So you can write, you can just add the agent for this one. So this agent is a similar one that we have used for all type of methods for the Telegram bot, custom UI hat, form submission, all those things. Okay. Now, you can just clone this one again. You can add this one as well. Again, you can add for each email. So by creating this particular system, we can easily organize the customer queries into three different parts like high priority, medium priority, and low priority in which the team or AI agent can give the answers in which we can organize to save our time and the customer retention. I hope understand this one clone this type of output as well, just click here to add this one, Clone just click. Now for this one, we need to add a text from the AI Agent response, select this, we'll keep it as same. Now, with the response we will take from this AI agent. Now we have successfully created this system. 17. 6.2 Building Advanced Email Handling Ai Agent - Part 2: Let's run this one. Let's check. You can click here Auto Align. It will automatically align that process. Now I will save this one. Now this scenario is saved successfully. I will send a message to this particular email. Let's check whether it will works according to our instructions or not. Let's send a message. This is my customer mail. I will try to send this. So let's write a mail to info. This is a company mail subject Agent N. This is a subject line. No, I will write for the high priority query like my product got damage during order dispatching. This is simple mail that is urgent need. My product got damaged during order dispatching. This is the hiperti because my words have product damage. I will send this. Let's check whether it works or not. I will send this one. Now you can see the message has sent successfully. Now let's run once again here. Now, let's check our company website whether we got the query from customer or not. Let's refresh this again. Now we have got this particular message from the user, urgent need, my product got damaged. Now we run this whole workflow, run once again. Now it is running, you can see it is taken as a low priority. For what let's check is running. You see how he's running two times, it has taken for the high priority and the low priority, again, for what purpose we'll check this whether it is low or not, let's check the output of this one. For the operation one, the GPT one, top V, let's check this output for the first one, let's take your automation to kits ready. No, this is our bundle two, right? So this is a message we have sent. My product got damaged during order dispatching. So for this bundle two, you can see you can check the operation to. It is simply categorized into the high. You can check here. You can see the bonder that is high. No further high, it is take this one, the bundle high. High. No this workflow is run. Again, let's check our company male whether we just labeled high priority or not. You can see we have got this label, high priority here as well, high priority. Click on the high priority, you can see this particular message from user that is gen my product got damaged during auto dispatching. This is a response from AI. For the low priority can check, this is the message we have got from the other company. That is not important. This is the one. Now, come here. We have successfully created the high priority system. Now, let's check for the medium priority in which we will send the message compose again, I will send a message to info damage like this one. Now we can also check our support team email J, whether we got the AI agent mail to Women sales team or not to manage that particular Query, you can see this is a support email team which handled by the humans. You can see we have got that email from agent that is A damaged product during dispatch, customer explanation. So I have got my name as well, email and my product got damaged during auto dispatching. This is the mail you can see here. And this is a customer mail. I will send to the A agent again, which is come under the medium for the order tracking. I will just write order status. I read the query like I am looking for my product tracking status. Let's send these messages to mail. The message has sent. Let's check our company mail, whether we got or not. Refresh it. So this is a new email which is order status. Now let's run once again here. That is running. Now you can see, now this second workflow is running because we have sent a email which comes under the medium reality. That is order status because we have just written the instructions in the OpenAI model, you can check here. But the medium priority order tracking order cancellation like that. You can check the company made in order to label this into medium priority, refresh it again. You can see this message has successfully added a medium priority, you can check here as well. We have got this order status. I'm looking for tracking status. This is the EI agent response. Now, let's check our Women's sales team. Again, whether we got the escalation email from AI Agent or not. You can see we have got that as well because the status, product damage or all those things are not mentioned in my company document. Then this is out of the AI Agent knowledge. Now, that particular AI agent will send email to the Women's sales team. You can see this is our sales team email. So we have got this user name, exact name, email, query, I'm looking for my product tracking status. So we have got that medium priority email as well. Send a low priority query to AI Agent. No, this is again. So we have got the response from AI Agent as well. Thanks. I have accultd this to our support team Lower your request. Again, ask a question, What are your products which comes under the low priority, right? Okay, I will cope with this. Paste it here as a message for the subject. That's simple FAQ like this one. Let's send this. Now the message has done successful, let's check our company mail again, refresh it inbox. You see we have got the new mail that is simple Q, what are your products. Now run the email agent again. Let's check whether it takes in the low category or not. This is the third flow. It is waiting. Now, you can see, we have got in the third workflow. Now it is categorized into low priority, email. You can check this one it is running now. Now, this will give the response instead of sending the man escalation team because it is a simple FAQ we have already mentioned in our company documents as contacts we have already provided to our AI agent. Let's wait to get the response. So you can use this router with the filters for the multiple flows. According to that condition, it will run that particular flow. Let's check our email box, refresh it. You can see it is simply categorized into low, so you can check here in the labels as well, you can get this one. Now you can say thanks. Here are the A home amlance main products. You can check the response from the AAgent customer support that is here. Simple FAQ here the company products details. Now, let's check. So now you can see we didn't get any escalation email from the Agent because that is a simple FAQ quien because it will handled by the AI agent for the simple FAQ quotiens that are already mentioned in the document. So this is how we can organize the customer emails according to their priority in which we can improve the customer engagement, trust, all those things. You can add the Google Sheet whether this AI Agent has sent a mail or not, you can track the status of that. As you can try other models like Cloud perplexity.ai or you can add the other email systems like email Make platform, Microsoft Outlook, you can add according to your client needs or your business needs like that. Okay. You can turn on this every 5 minutes, it will run this workload. Now I'm giving an assignment to you. Add a simple email system to the Human sales team. You can add to this, for the customer support team, which is handled by the humans, not by AI Agent, you can add the labels here as well for the humans sales team, or human customer support team in which they can also get into the three different categories that is high, medium, low. So what happens here when the user send the email company email invox also categorized into three different categories and when the user ask a question out of the AI agent knowledgebase, then this AI agent will send a email to the women's sales team. So the Women's sales team also have their unique Gmail or business email. Okay, now you can also add a label to that particular Women's sales team email in which they can also reach out the user according to their party system. You need to add simple come here and just for every AIA chain, you can add a tool here, add a tool, add a model. Okay? Just select a model or create a new one, creative model. Search for the Gmail again. You can select the update email labels again. Now, the Gmail connection should be your customer support. You need to click Add it button here. You need to sign in with the customer support Gmail, which is handled by the Humil sales team. Then we can add the labels to that particular system. You can try by yourself. This is an assignment for you. 18. 6.3 Testing Advanced Email Handling Ai Agent : Have another issue in this particular AI agent, that is, if you see here the customer mail response, we have got in the paragraph format from the AI agent, which looks like not professional, right? This format of response is best for chatbot like Telegram chatbot, custom UI chatbot, it should for them. But for the email, it looks like clangy or not professionally, even if you can see giving in the response like paragraph. So this is not a professional, right? So we need to change the simple prompt AI agent message here that is symbol. So how we can do that, click on the AI Agent. You need to change the message one. You need to add a message to, like we have did in this AI agent. Okay. Now you can see this is the custom AI prompt that is message one. We will send to the AI agent with the customer details detail like customer email message ID, customer email subject, customer email body, customer email address. This all comes from this one here. This is trigger. Okay, Gmail trigger, will just retract from. You need to write this customer email message ID. After that, you need to dragon drop from here, like message ID here. For the subject, just dragon drop here. For the customer email body, you can take from the full text body. For the customer email address, just dragon drop this from email. That is simple. Now for the message too, just click on ad item, you will get the message too. You need to use this simple prompt that is format the body content message to send to the customer in HTML, keep it simple this is the format. This is a simple format if you are not understand where you can use the Char JPT to get knowledge about this format. Okay. Now I will just use this as it is in all AI agents to work properly. Okay, I will just save this one. You need to add this exact format in additional instructions as well to send to the human sales team when the user ask a question out EI knowledgebase, it will send the email with this format here. Okay, we need to write the aiginal instruction as well in this box. Just save this. Now we will run. I will send this one again. Run once. Let's check. It is taking. Now it is taking as a high priority email. Before that, we will analyze. This is the previous response from EI agent. Rick was 90% discount. Hello, thank you for contacting us. This response is looking like a paragraph, but not as a professional email body from company. Have revived the AI Agent prompt message, additional instructions with the email format, it has successfully run. Let's check our customer mail whether we have got the perfect email or not. Now, you can see we have got the email. Now you can see just now we have got the best email. With the good format, professional email, you can see, hello, this is my name, hello, your products. Thank you for contacting us. I understand you are requesting 90% discount on all products. What we did, I have explained your request to our team for review. This is a concise and very easy to read, and it looks like professional. You can see what to expect next. Our team will review the request to respond within 40 to 72 hours. S is a Mail, thank you for your patience. This is a real actual professional looking EI Agent response. Okay, in previous one we have just got in the paragraph, but in the next one, while we revive the prompt message by giving additional details, user details and the format of email, we have got better email. Now, let's check our human sales team email. So we have got from AI Agent when the user asks a question not in the AI Agent knowledgebase. You can see. So we have got in the good format that is user name, user email query, you can see please advice and eligibility. If you see in the previous, we have got like just in this paragraph after reviewing the additional instructions in the AI Agent with providing format technique, you can see format the body content, we have just got the message in the clear format. This also looks like a professional. You can follow the same. I will provide all these templates in the document, you can get it from there. Now let's copy and we will do for each every agent in this workflow. We'll just copy and I will save in each agent we have did in the high priority email. Save this one again, come to the low priority as well. Come here. Say this one. Now let's do for the prompt as well. I will copy this whole message. I will do for the medium priority email, AI g that is simple, come to the low priority mail AI ag, paste it here. Save this one. Now for the format of email, copy this message to content, come to the medium priority AI agent, add item here, message two, paste it here that you have earlier copied from high priority mail AI Agent. Just do similarly for the low priority email AI agent as well. Save this. You have successfully created a simple AI agent which handles a query and it will organize the emails according to the high medium priority and low priority in which the human sales team can reach out to that particular user according to the priority of that query. Help understand these points. You can do for each and every agent up to we have created. For example, if you look here, just come to the scenarios again. Now, if you check AI Agent chatbot simple AI Agent chatbot here, we can edit button. So I'm giving an assignment for you to try by yourself. You can use the exact additional system instructions with the format, this prompt, and this prompt and this prompt message, you can use for each channel A agent that we earlier created. Whenever you use a email model or action in our work, you can use this particular exact message in the Telegram bot or in the form submission AI agent organize your AI agent response into the format of email. So for example, you can use in the form submission AI Agent. When the user submit a form in website, this AI Agent will give the response to this mail. But you can see here, we have got an email in the paragraph. How we can solve this, you can use exact this format message in order to get the professional email from AI Agent as well. Just add item message to. You can write the format, the body content message to send to the customer in HTML. So you can write this one. When you use this particular email as a output or a tool to send to anyone from the AI Agent, you need to use this particular format prompt technique, right? This is a simple prompt. Even you can use so many things we have just done in the email AIgen, you can add the user ID, user name, drag and drop here, user email drag and drop, query drag and drop. That is simple we have done in the previous session. Try by yourself, then you will create the professional advance AI Agent. In the next session, we will see how we can create the feedback collection AI agent by doing sentiment analysis, whether the review is positive, negative or neutral. According to that sentiment analysis, we can send a personalized email to that particular user like that. Okay, let's dive into that. 19. 7.1 Building Feedback Collection Ai Agent: This session, we are going to see how we can build the feedback collection AI Agent in which we can collect a user feedback, and we can categorize into positive or negative or neutral by using sentiment analysis model in the AI Agent that we have done in the Int and marshy course, building multichannel customer support AI Agent. In that course, we have just learned how we can create the feedback collection AI agent by using sentiment analysis. So basically, what we have done in that previous course is we will collect the feedback user name, email, and the feedback from the form submission, and we will just pass to the sentiment analysis. This is the sentiment analysis node, and it will use the OpenAI model brain to analyze the feedback of user from the form submission, and it will just give whether the review is positive or negative. If it is positive, we will send it to the this AI agen and this AI agent will raft the email by saying thanks to that particular user for the positive review. It will encourage that particular user to subscribe our newsletter or to enable the notification for future product updates like that. Okay. In that, we can increase the sales as well. If the feedback is negative, then this sentiment analysis node will send to this AI agen. This AI agent will say, sorry to that particular user. And it will offer the solution, or it will offer the discount to increase the sales as well, like that. We can add so many possibilities to satisfy the customer. We can build the exact system in make.com. Let's dive into that. So I am just go to scenarios again. I will just go in my folder that we have created multichannel customer support, IGN, Al Creek on the create scenario. Physically, there are a lot more ways you can collect the user feedback. Okay? Like we have done in the tin, we have just used the inbuilt node by the NTN. Okay? In them.com, we don't have the separate form model, but we can use with different apps like Google Forms, Microsoft Forms, or custom form that we have done in the form submission customer support channel. You can do that. By using Google Form, you can collect so much data from the user like user name, email, phone number, feedback, or other data. Basically, you can collect all the data through Google Forms. It is a best recommended way. I will just add a Google Form model. Let's search for the Google Form. Basically, we are just collecting the user feedback through Google Forms. I will search for the Google form. You can select yes, this is a Google form. Just click here. No need to use this watch responses. Okay. When a new response is received, this will pass details of that particular user that we will take from the Google form. Okay, I have to understand these points. And before that, we will connect this account. Just click on the advert just sign in with Google, and you need to add a new Google form. Okay? For that, just come to the Google forms, sign into your account, create new one. Now you can give the name as feedback form collection like that. I will just give name. I will write form name as feedback, feedback, collection, form. We can name according to requirements. Now I will just write, please submit your feedback do. For the question, I will write. Let's take it to the name holder. Now for the short answer, let's keep the required one and add another field that is email. Short answer, add another fill. Feedback. Paragraph long required ones. You can add so many details if you want to gather from user user photo or user phone number like that. You can add. I will just go with this here. I will just send to the publish, click on the publish. So you can give the name feedback collection name here. This is very most important. Copy this whole name in order to get here. Now, click on the Google forms, search it. Document all this feedback collection form. Now, it will automatically search that particular ID. Now you can see this is ID of our form. For the limit, you can see the maximum number of results to be worked with during one execution cycle. You can put a two, three, four. Basically, I will recommend up to five. That is simple. Save this one from now on, save. Now we have successfully just what I will do. I will run this form whether it will works or not. We will check by sending here. In the Google, just copy. Open the Google form that we have created in the new tab. No, this is a form we have created. Now, I will submit my name, email, and feedback, and we will run this form whether we got a details from user or not. Okay. Click on the run once. It is running. Now, let's send the details. Name, let's check. This is my name. But the email, I will use the customer mail. That is, this is one. This is a customer mail. Come here. This customer mail. I will just send a feedback that is you are a product is good. I will submit this one. Now, this is a positive review. Let's come to the feedback. We can check. We have got the response in the Google form as well. This is a three your product is good. Now let's check to this form. I will run once again. Now you can see we have got the response from the Google form. This is a bundle one. You can click. This is a response ID, created time. In the answers, you can click here, we have got the name, text answer, answer, one. You can see value, this is ive. This is the value. We need to dry and drop this particular field value from the answer bundle, answer field to pass this field to the other action. Okay, let's see here. You can see for the email, you can check in the text answers answers. One, this is a value. For the feedback, you can follow the same text answer answer, one, where you put is good. This is a basical feedback. This is the user mail, and this is the user name. Let's add another model that is AI agent. Before that, we need to categorize that particular feedback from the user is a positive or negative or a neutral, like we did in the Nitin muscle codes like this. Before giving that particular feedback to AI agents, we have just used this sentiment analysis node in the NTN. We can do the same in the make AI Agent by using Make AI toolkit. You can find this tool kit by searching like Make AI tool kit. You can select that model. In that Make AI tool kit, you need to search for the sentiment analysis, like you can see, Analyze sentiment, click here. So this is make AI tool kit. So you can use the model that you have chosen. Otherwise, you can change if you connected your open a cloud like that, you can change it from here. You can select these models as well. I will just go with the Make hen. I will just take log GPD five mini, and you need to simply take that user feedback text to analyze. Okay. So where you can find it, we have just seen that where we can get the feedback, see these answers, feedback, text answer, answer. One, this is a value. So we can find out the exact thing from the gold forms. Click here, need to go to answers, feedback, text answer, answer, Value. We need to click here. That is simple. Now, just you can see, for example, a customer review or response on social media. The model will return whether the text is positive. So this is the result from this particular make AI tool kit model. So we will get exact this text. If the review is positive, we will get this lowercase sensitive word that is positive. If the review is negative, we will get this text. Remember this one, these are the three results. According to our text, we will provide to this make AI tool kit. Now we need to select this return and description of the sentiment, clip on S. Okay, save this one. No, we have successfully added the sentiment analysis to this particular feedback, Google form. No, we need to add a router here. Okay, take Router. So what happens here? When the user feedback is positive, it will give the result like positive. Okay? That positive comes here, so we need to add a filter. To run that particular flow for the positive, set up a filter. For the condition, we need to use result, result from the make AI tool kit. Just click here the sentiment. Whether the sentiment is positive or negative, you can sell it from here. Sentiment and if the sentiment condition is equal to positive, So you need to use exact this word in the lower case sensitive characters because the output from this make AI toolkit is in the small low case sensitive characters only. Let's save this button. Now we have added a filter. Now I will just add a filter for the negative review. Condition is exactly the same. If the sentiment of this make AI toolkit is equal to negative. Okay, say this one. Now, I will add another router for the neutral one. Add this setup filter again. If the condition sentiment is equal to the neutral, this rooter simply run this flow. Now, we have successfully added the sentiment analysis to the Google form with the router. We need to add a AI agent. According to the review of user, it will send a personalizer email. How we can add that, click on this one, use a Make AI Agent again. Run an AI agent. Now choose your AI agent that is exactly the similar chat AI agent we have user for all channels. So just add a message here. Craft an email. So before that, you can add the user details in the message and the sentiment analysis of this one, AI Agent can learn more about user. Then AI Agent will thanks user. We are sending an email to the user, right? In that case, we need to send a professional email, not like a paragraph email. So for that, you can provide this user feedback. So you can dragon drop from the Gool form, that's text answer. Again, answer value. Hit Inter, user name. We can take from the name text answer, answer, value. User email. So we can take from email field, text answer value. So we have simply just given the details of user. We will just tell to AI, craft, thankful email to user according to user feedback and encourage user to subscribe or enable notifications for a future product update. Save this one. Now I will just clone it. I will add it here again, delete this model. I will clone this again, add this one. No, basically, we have added the AI agent for neutral, negative and positive. Now, click here. Now, write a email to user according to the user feedback and offer user and offer discounts to user or solutions to user to make user happy with our product. Save this one. Now for the neutral one, we can write write thankfully email to user according to the user feedback and encourage. So we will keep it as same. Okay. Like neutral, positive or the same, but it will different for the negative one. Added the filter, you can give the name for that positive feedback like that. Okay. Save this one again. You can see we have labeled this one as a positive feedback flow. You can do it for the rest of dates flows. I will just set up a filter again. We have just not saved in the previous one. If the sentiment is equal to the negative, save this one, I will write as negative feedback. Save this one. So we have got this negative feedback here. Now I will write set up a filter again. But the label I will take as a neutral feedback. If the condition sentiment is equals to the neutral, then it will run this flow. So we have successfully added this three different flows according to the review of user. Now, we will just add a email here. Or when a user submit a feedback through Google forms, this A agent will send a personalized email. So we need to send an email to user right. So for that, click here to add another model, search for the Gmail again. You need to use this send an email action. Okay? So for the recipient email address, just click here. You can use the Good form field that is email text answer, answer, value. Okay? This is the Gmail of user. But the subject, you can write thanks for your feedback like that. Thanks for your feedback. Now for the body collection, you can go with the raw HTML. But the content, we need to select the response from AI agent. Just click here, AI agent. Now, just save this one. Now, we need to clone this one. Let's clone and we will add to rest of that flows. Clone, add it here, Clone, add it here. Just add a response. From this AI agent. Now, we need to add a email format. In order to send a professional email to the user, we need to add a format technique. Save this one. Now, I'll copy the exact format prompt from previous email AI agent. I'll just go to the email AI Agent. I'll click on the Edit button. I'll select this agent. I will copy this format, body content from here. Now I will go to my This is. We need to give the name as feedback collection agent here. This click here. I will add a new item in the message that is format, the body content message to send to the customer in HTML. Use this for every AI agent in order to send a professional email format. Let's say this one again. Add the message two. So we have done for the three agents. Now I will give the name here. Feedback collection agent. Simple, right? We can add another feature here. Like we can also save this whole user name, email user feedback, and the EI agent response in our Google Sheet in order to track this AI Agent workflow, whether these make AI agents are working correctly or not, according to instructions. So to do that, you can just click on the add model here, click here, select the Google Sheets, create the new Google Sheets. Let's jump into Google Sheets. Now I will create a new spreadsheet. Now I will give the name as a feedback collection. That is simple. No, I will give the name for the A column that is user them. Email. User feedback. For the D, I will just take sentiment analysis. I will just take with this one. No further E, I will just take EI agent response. Okay, these are the pills we have created. So when the user submit the user name, email feedback, and sentiment analysis of AI agent and AIGenRsponse. We will store all the details here to track our workflow efficiency, whether this workflow working perfectly or not, like that. Okay. Add a Google sheet here. In the Google Sheet, add a click on that model, just to go with your connection. Such a bypath myrive click head to choose a file ID. You can see this is the feedback collection we have just earlier created. Now for the sheet name, this is basically Sheet one. Now for the user name, we will need to take from the Google forms, name, go with exactly text answer, answer value. But the email, exactly do for the email field, email text answer answer, value. But the feedback, do for the feedback field, feedback, text answer, answer, value. For the sentiment analysis, you need to take from the make AI tool kit. This is the Make a toolkit. Click the sentiment. But the AI agent response, you need to take from this response. Now let's save this one. Now I will clone this one and I will add the rest of these flows. Clone it, just click and add it. Now click here again. Now we need to select the response from that particular flow AI agent. Save this one again. Now, click here this alignment. It will automatically align your workflow. Now we have successfully created this whole feedback collection AI agent, and let's try with the three different reviews like positive, negative, and neutral. 20. 7.2 Testing Feedback Collection Ai Agent: See whether it works or not. I will save this workflow. Now the scenario is successfully saved. I will send a new form, then I will run once. Now you can go with the on demand. Let's save this one. Activate scenario. Now this scenario is activated. I will send a positive feedback, right? I will give them my name for the email address. I will take the customer mail. Let's say this one. This is the email. Feedback, I will write your product. Looks amazing. So this is a simple feedback I will submit. Now, let's check our customer mail whether we have got the email or not from EI agent. Let's let's refresh it. Okay, we need to run once again here. Come to the AI Agent. Just click here run once. Now it is running. Let's take it is taking as a positive feedback this flow. Now this Agent simply correcting an email to send a personalized email to that particular user. Let's wait for a few seconds. Notice send a message. Nobody successfully added the Google Sheet as well. Now, let's check whether we got the email or not. No, you can see we have got the email from the AI Agent. That is thanks for your feedback. You can see. I have prepared that SM Email, your feedback to Acords below is SML body format as requested. You can copy paste this directly into your email client. This is the email here. Dear Si, thank you so much for your lovely feedback. We are thrilled to hear that you think our product looks amazing. Your support means a lot to hold this company name. You can see Mont More updates. It is encouraging to subscribe senable notifications here. You can see this is warm regards from this one. This is a professional email, right? We have got this one here from the response from AI gen. This is unusual one. So to avoid this particular issue, you can write the AI agent here, like Always always reply with email only email content only, not other words. So by using this one, what happens, we will not get this type of message from AI Agent. That is, I have prepared the HTML email and save your feedback. This is basically from AI hen. You should not look in the user email. I have just written the prompting that is strictly follow. Always reply with email content only not other words, strictly follow this one. Let's save this one. We will add the exact prompt for each and every AI agent. Save this one, go with the MCI Agent. Follow and save this one. Now, let's check our Google Sheet whether we have got the formation or not. You can see this is my name, email, user feedback, your product looks amazing. The sentiment is positive, and this is the agent response. Like you can see here. It is basically showing this, this is a email format here. You can check this Agent response and you can track your AI agent whether the A agent is working perfectly or not. Now let's submit another I will just go with my name. I will use a customer mail. Feedback is I got damaged product from your side. I'm not happy with your products. This is a negative to you, I will submit this one. Now, let's run once again. Save this one, run once again. Let's check whether it will take either negative or not. Now you can see it has taken as a negative feedback. You can check. This is a flow for the negative feedback, right? Now, let's wait. Let's check. Our customer may, we have got an email or not. Now you can see, thanks for your feedback. Now we have got the professional email from the AI Agent. That is DSI, we are sorry you received damage product. That's not the experience we want for our customers. Thank you for letting us know. We would like to make this right immediately. Please choose one of the following options, replacement free and refund. 20% of your next purchase, extended warranty, contact name. So this is the exact email from the AI Agent. You can see how we have got the professional and easy to read. Email with the replacement refund and adding a discount or offering solutions like that. So you can see here for the negative review. In this way, we can make customers not to leave our platform or our product and services. Like that, you can build this particular amazing AI agent automation. Okay, you can check this how well it is written this email. Can check the sheet as well. This is one. This is for the negative one, let's submit a form for the neutral review. I'll give the name as usual email Z one. So now I will just take products or well, looks like normal, normal and simple. This is, I think, neutral feedback, submit. Now, come to this one, click on the Run one's button. Now, let's check whether it will take as a neutral feedback or not. Now you can see it is taking as a neutral feedback, right? Now let's wait how it will craft the email. Okay. Now let's check customer mail again. Now we have got the email from the EIHN Now, you can see this is for the neutral thanks for sharing your hones feedback. We truly appreciate you taking the time to tell us that our product feels normal and simple. You can see we are always working to make our products smarter, more useful, more distinctive, so we can subscribe for our future updates. If you have any questions, this is the time and all those things. So this is the neutral EIHent email. Okay? Now we have successfully created this whole amazing feedback collection Agent to check whether the given feedback is a positive or negative, and we can send a personalized email to that particular user for further actions. Okay. So now you can see our Google Sheets, as well. It is simply log all the things. Okay, for the neutral as well. You can check for the negative view I got for the negative view, we need to get the excelation email, you can see here to the man sales team. So if you see here, this is a simple paragraph response because we have not used the format prompt technique in that particular AI agent. So just click here AI Agent. You need to write a additional instructions, click on advanced settings. In the additional instructions, you need to write when the user submit a feedback your knowledge like that, you can use exactly the same we have done in the previous AI agents to escalate the particular email to the man sales team. Try by yourself. You can change the name of this particular agent that is feedback collection AI Agent. Save this. Now we have successfully created this feedback collection AI agent in which it will store the user feedback with the sentiment analysis and user details, and it will send a personalizer email to that particular user for further actions and can use this Gmail and Google Sheet as a tool in the AI agent as well here. You can add it by yourself. If you want to test the AI agents in different scenarios, different ways, add them Google Sheet as a tools to this particular AI agent and check the output whether we will get the expected output or not by yourself. In the next session, we will discuss how we can add a specific EI agent which checks our main EI agents performance, whether they are working according to our instructions or not. Okay? And we will discuss some effective methods, how we can add it, and there are some practices which help us to keep this AI automation safe. We will discuss each and everything in the next session. Let's dive into that. 21. 8.1 Adding Ai Agent Auditor - Part 1: In this session, we are going to see how we can add a specific AI agent editor in which we can track the AI agent performance, whether it is working under the conditions of our instructions or not, like to avoid unwanted response or which is not lited to our product and services. We will discuss some effect two strategies, method and best practices in this session. Let's dive into that. Now, I will jump into our first scenario that is EI custom chatbot no, I will send a message to this particular AI agent like, Hi, no let's send. It is working. We ask some question which is not related to product and services, like what is an AI. You see here. So it is giving the irrelevant to our product and services. You can see, I have saved our conversation how like the explanation. So it is not working properly. Okay. When any customer comes to your chatbot, ask any question related to our product and services, there is a good chance you will get the expected answer. When some users ask the question not related to your product and services, if they put the question that is what is an EI, when they use a general chatbot like Char GPT, our EI agent should give the response not related to that. It should refuse to giving that particular response to that particular question. Like, I am not trained on this, so I am only the EI agent for this so and so company which solves user queries like that, it should give that particular response, but here is giving the different response. Why? So there are some issues the main problem is EI agents model. Okay? Now, we have just taken this chat AI from the mix provider. So one of the best effective parties, you can change this particular AI to the most advanced OpenAI cloud or Gemini models for more advanced reasoning come to agent settings. You can change the model as well to test it out. Now, in the connection come here, you can use open E AI agent models in order to make the A agents work properly. Otherwise, you can write different instructions. According to it, you can click here Improve button. I will improve all those things you can write here, so it should work. Instead of that, you can write by Uself as well, Cole write here only do only give response based on above instructions and simply refuse you response which is out of er and rules instructions like you can give the context. You can examples, which helps the agent to understand our instructions very well, and it will simply refuse to give that particular answer. Like, you can give the example as if user ask query like what is an AI or other not related to company or product services, that's simple. Now, let's save this and we'll check whether it works or not according to our instructions. I will save this one again. No, let's send the question again. I will write as what is an AI, you can see, I'm sorry. I can't answer general questions about AI, and for our support policy, I need to escalate this our team. So again, you can see here, it is asking my name and email to send the details to team because it is out of the knowledgebase. So here's the problem. Now, let's try another way. I will jump into our AI agent again. That is chat AI, click on the configuration. So we already here. Now, you can see before escalation, gather user name, can write. Before sending mail, check query is not out of context, it is a general query out of company and product. If it is simply refuse to give answer. You can see here. Before sending mail, check query whether it is the out of context, general query, which you can write here, I'll copy this one. You can write exactly here. Before sending mail check query, whether it is not out of context like it is general query which is out of company or products information, if it is, simply if you answer. Okay, let's say this check again, whether it works or not. So our main goal is not to ask this question as well. Please provide your full name and email so I can send your question and details to the team. Okay. Now what I will will just reload again, ask a question that is what is an AI this say, I'm sorry, I can't answer general questions about A that are outside the scope of this documentation. Our support policy, I need to escalation this to team. Again, it is asking, know what I will do. So I will just come to our AI H en. Let's copy this whole system of instructions. I will jump to ChagPT for writing or I'll jump to Cloud for writing system instructions. Which is even better than my writing prompt. Okay? Now, I will improve this system prompt, work properly. Now I will paste my instructions. Let's give this prompt to the Cloud. I prefer Cloud for writing the system prompts for the AI ages. You can use this is a system prom, customer service AI Agent, core identify machine, knowledge boundaries, what you cannot answer. This is simple refusal template. Tool usage guidelines. This is very long prompt. You can use this. Let's copy this one. We will check whether it works or not. I will copy this one. Come to the agenda again. I will hold this prompt here. Now, let's check from starting. This is our system prompt. I will remove this. Now, this is a core identify knowledgebase rules and boundaries, what you call refrigerator plate. This is a company name we need to change to our company name that is ACMAH appliances. That is simple tool usage guidelines do not escalation, escalation process, email. We will just remove this one because according to our instructions it will automatically take because we added the tool with the E Agent instructions. That is simple. Now, response framework. So this is you can see step by step process. Examples, we have seen examples as well. So instead of company name, I will write this with your product and services example too. Is writing them, well, you can see, no, let's save this one, re improvements made. I will remove this one because this is a simple we do not require. I will save this one. No, I will send a question again. I will reload is what is AR. You can see I appreciate your quotien, but I am specifically designed to assist with ACM home applies, product and services. But general information, things I would recommend searching online or concerting other resources. Is there anything related to our company that I can help you today? So this is the response I am looking from the EIgen. So this is the power of writing the clear AI Agent system prompt, which helps our AI agent to understand our process and all those things. So we have successfully added a system prompt, which is effective one by using Cloud. So I will recommend you to use the Cloud to write or to improve your existing system prompt. So if you are looking to learn how you can use a different AI tools, you can check the complete prompt engineering master class. You will get the benefit in that. You can find it in my profile. So this is one of best practices. Now, come to another, come to our EI Agent chat boards. We had already added a simple Agent log system in this AI Agent chatbot. So in this AI agent will simply enter the logs, that is user question as well as and the E agent response. AI Agent manager can track whether our E agents were working properly or not. So that is also a manual work. We need to hire AI Agent editor. This is also manual work. We can add another AI agent here before sending an output to the user. So you can add a simple AI agent here, right? 22. 8.2 Adding Ai Agent Auditor - Part 2: Create another A agent, which works like it. So for that, I will use Cloud to write a system prom. Okay, now let's jump into our cloud. Now I will write a system prompt per AI agent, editor, which checks AI agent performance, whether it is working under our instructions or not like above. Now, let's send this to the Cloud AI. Now it should write the system prong that we can use in our new AI agent, which checks the AI Agent performance. You can say this is a system AI Agent performance editor, core identifying mission, you are agent, editor for evaluating customer service agent performance. Your role is to analyze conversation logs, see. So this is a performance evaluation responsibility, compan checklist violation to flag and edit process. So this is a based upon the previous details. By this, we can copy this whole prompt that is co identify mission from here. I will copy this one. So it is very long. Now we can change this prompt according to our needs and requirements as well. I will tell AI. I am looking like this when AI agent is not given response according to user query, this AI agent editor should you response positive or negative? Only text. Let's give this part. I will just copy this whole prompt from here. I will just copy this one. Now, come to AIHt, come to the AI agents again. I will create a new AI agent, oe AI agent, Make AI Agent for the agent name, I will write AI agent. I will remove this particular unwanted prompt here like this one. Let's say this one. I have just copied from the cloud. This is one. We need to add a context. By using this contexts, then the Agent will check our company details. I will give this as usual, company docs, all those things, click on the upload. We are not here any tools here because the A Agent editor work is to check the output of the AI agent. I will say this. Now, come to our scenarios. We need to add this particular AI agent editor for every AI agent. Then we will get exact responses. Not only that, you can add a simple, more advanced instructions like the output should be free from errors, free from hallucinations, free from irrelevant words. Now, let's come to our Agent chatbot. Let's click on the Edit button. Now let's add that particular AI agent in the mid, I will add, add a model, search for the make AI agents, run an AI Agent, choose our AI Agent dit. Now for the thread ID, record thread ID that is here. Now for the message. So we'll just take this response from the AI agent here. Save this. This AI agent will check this response. Okay? In our instructions, we have specified each and everything. If the response matches our user query, then it will send a positive this next model or action. We are adding a simple router again here in order to run the specific flow when it is positive and negative. Let's set up a filter. Now for the label, you can write like dit result response is equal to. Let's give the positive. Then only it will run the next flow. If this agent will pass a positive, then only it will run this one. It is negative, it will stop here so we do not get the response from AI agent. You can add additional system instructions like if this response is negative from AI Agent editor, what we need to do next? If the if this response is negative, simply give the reply like this. Please try again. Let's save this one. So we're not required to add a filter here, right? So I will just remove this one. We need to add another router, save this one. Again, I will add a router here. Okay. Click here, add a router, no. Let's take this one. Basically you are adding another AI agent which specifically check the Agent response. We'll add a filter that is addit is positive. If it is positive, it will just run this whole flow. Now, let's give the response is equal to positive. Okay. Let's say this one. Now we'll add another flow. Now for this one, I will add a filter like if the dit is negative. But the condition I will take from the response from AI additive is equal to negative. Save this one. Now, we will send please try again. Now, I will just write it please let's say this one. We do not require additional instructions because we have added another flow, so I will just remove this one. Now let's try whether it will work, can add the same A agent here to work properly. Now I will just clone this one, a clona come here. I will add in between that. Add a router again, set up a filter if it is positive, y. Let's take addit positive. Condition should match this particular response from AI is equal to positive. Save this. Now we need to add another flow that is here. A simple webhook again. Just to give the response, please try again here. Okay. Now let's add a filter as well here. Set up a fillna. I add it negative, this is a name. If the condition of this particular AI agent is equal to the negative, Save this one. Now we have successfully added this A agent. Let's configure this particular A agent as well here, this is the same. For the new user, it will create a new thread by itself. We need to take from this here. Let's try whether it works or not. Now we will ask a question again. Now let's ask a question that is what is an AI. Let's wait for the response from AI Agent, whether it will check the performance or not. We'll get response as usual, but we'll check in the workflow, whether it will providing the positive or negative. We'll run once again here. Waiting for data. So what is AI? Let's send this one. Let's come to our chatbot. Now it is running as well. Here it is stopping. Why we need to check this one? The response is positive, well. Here Bundle one, bundle Bundle three. It is running, we'll stop this one. There is an issue in that. We'll check why it is happening type hook. This is a problem. We need to delete all the records. Then we need to try because it will run as a new user. Now I will save this one. Before that, I will jump to data stores again. Open this data. I will select all Confirm. No, we do not have any data. Let's comp into workflow scenario. Come here. Now, let's try again, run once. It is waiting for data. Now I will send the new message from chatbot. I'll write What is AI. Let's send this one. Let's come to our AI agent. It is working. It's taking time. Now it is working perfectly because the simple it is No it is not going in forward. Why? Because we have added this filter here. We need to remove this filter. Unlink again, we have added. We have added this know here. Stow router order. It is working. We have got the two operations that we have got the positive in the bundle one. The output is positive, so it is working for pit now. If you see here it is coming with the higher case, positive. Let's check our filter we have given in the correct or not, positive, positive. Positive if the response is, so we have given the wrong. Let's correct this positive word correctly. Now let's save this one. We also check this filter as well. This is a problem. We need to write the exact word that will return this agent. Then this router matches that particular word and it will run the whole flow. Now I will run once again, devious data, run once. Now let's check whether it takes this flow or not. Now, it will give a positive. You can see it is working perfectly. Now, let's check. We have got the response or not. Accepted. It is working now. Let's come to data store again. We need to remove this datas as well. We need to start from the scratch. D all records, you have to confirm. A H Chatbot. But if you see, we do not required to add a thread ID because it is causing the issue here. We are adding this agent to check whether the AI agent response is correct or not, but we are not saving this thread into our database. We do not require to add thread ID to this AI Agent, save this one, run once again. Now let's stop it. I'll just go with the run. This here. Let's check our response. I will start from the scratch again, stop, come to data stores. We need to try different things. Then only we will get a response from AI Agent. Now I will run once again. I will send a message high. Let's wait for the AI Agent response. It is working now. It is taking the existing user. Speak on response. You can see we have got the response. How can you assist today? That is working correctly. Now what I will do? I will stop this. Now, I will ask a question which is now as a positive, it is run this workflow, we have got the response from EI agent. That is, how can I assist you with this ACM home appliances today? It is working perfectly now. I will run this module only here. I will run this model. For the response, what I will do, I will take the agent response as a negative. Let's check whether it works or not. The response, I will write EI is an artificial intelligence like that. Let's run this model only. I should give the negative response, right? You can see we have got the negative from this AI Agent because this A agent is checking the performance. Now, the response is negative. It run this one, okay? The response from this AI agent, this negative. If this negative and this negative matches, then this webhook send a response that is, please try again. You can write this. Please try again. I will give response to that queries which related to our product and services. So I'm giving an assignment to you. So just add these exact AI agents in different AI agents that we have earlier discussed like email EI agent, feedback collection AI agent, Telegram AI Agent. Okay 23. 8.3 Adding Ai Agent Tool Auditor: This specific editor AI Agent for this specific output models or actions in this workflow. What if if your output actions are connected to AI Agent as tools. Okay, so this is the AI Agent we have created for our customer support. So we have added this human escalation and Google Sheet add tools as a models to this particular AI agen. So what happens here is whenever a user ask questions out of the knowledge of AIGen it will send email to the team. Okay, this is one of the one tool, another tool is, so it will just create a new rule with entering the user query and AI Agent response in here. These are the two which is handled by the company. There is no end of the user here. Sometimes, according to our business, we're required to add a output tool here. Instead of the actions, we have just created here a webhook these particular output models, we can add as a tool here, in which we can simplify the E agent working how we can add that. For example, this system, AI agent editor will best work for this particular workflow because it is for the end user. When user submit a question in the chat booard, this AI agent give the response in that particular chatboard which is a end user. There is no problem in that, the customer queries get the solutions from AI Agent. There is no problem. If we get the response, even which is wrong response in these particular tools like human escalation, Google Sheets. These two are handled by the company employees. We do not get any problem with this particular workflow, so we can rectify it. What if you are not able to add this A Agent editor in this particular workflow for the end of user, there is a problem in it, the customer can get the irrelevant rest. For that, come to these particular scenarios again, click cut scenario. You need to come to model tools, click here. So these are the tools we have connected to each AI Agent, that is chat AI. We can see how it works behind the scenes. AI Agent, how it works, what are these one. In the AI Agent, we have added only the Google Sheets, Google Sheet and Women exclon for the email. But what are these? We actually these two works and we will try to add a AI dit agent in this particular models, and we'll try to avoid irrelevant response in the tools as well. Click here, Come here. No need to unlock it. If you see here, this tool is so unlock convert and edit as a scenario. So from AI agents to settings if either one Ag. This is our model. We have just added to the AI agent that is GML, but it is called as a scenario. In order to unlock this model, so we need to convert this module to scenario. So you can click here Unlock. Now it is convert to the scenario. It is not come under the model. Now, let's click on the edit. You can see this is a scenario input and this is the scenario output. You can check here. Scenario inputs and outputs, you can click here. These are the scenario inputs. These are the scenario output. In the previous AI chat a agent, the chat A agent will send the data to this particular scenario inputs. So this scenario inputs come from the agent. This is a major advantage of this particular scenario inputs and scenario outputs. This is a scenario inputs, you can check here, click here. You can see subjec text decide automatically by itself, content, you can click here content. You can use HTML code text that is simple. This is a scenario input and what are the scenario output, you can check here as well. This is a content in a scenario output tool output. What is the tool output? This is a Gmail. The AI agent will send the email. When the user asking any questions which is out of the company knowledge, the AI agent will craft email and it will send this scenario input. This scenario input have the data of email subject, email body content, all those things that we have created in the previous AI agents, and it will send this email to the company support team, and this is output. That is simple. This is how the actual tools works behind the scene of each and every agent we have created. Before sending email to our support team, we can add a module here. That is AI Agent. Let's add this one. Add a module. Search for Make agents. Run an agent. Click here. Choose your Agent that AI Agent is AI Agent editor. Click here, you can keep it thread ID as a empty because it will create a new thread ID by itself. So for the message, you can add the scenario input. So you can add the subject as well. Okay, you can write the email subject. And for the email body, we can dragon drop this content. Now, you can write any additional instructions like analyze this email. And do your work according to instructions. Let's say this one. Now, we have added a Me AI Agent. Previous customer support Agent will give the email. If the email of that particular agent and which is not explicit to our rules and regulations data, no this E agent will pass a positive then it will send a mail to our support team. As usual, it will work normal. If this particular dit AI agent pass a negative result, then we need to add a router here as well. Like that you can add this particular AI agent editor for tools as well. Try by yourself. I'm giving an assignment to you. You need to add a simple this particular AI dit agent. You need to add for the email AI agent, feedback collection AI Agent, form submission AI Agent and Telegram AI Agent by yourself, then try whether it works according to our expectations or not. Now this human escalation tool is converted into scenario. Now come to the model tools, we have only the one tool that is Google Sheets. So this is not converted into scenario, we can do that just click here again. Unlock, now it is converted into scenario. That's it. You can edit this one. Add a simple AI agent edit here as well. Before entering the user query and Agent response, it will check. If it is positive, then it will enter all the locks of the E agent and user query. Try yourself, you will get the idea how you can do that. Now we have successfully created the AI agent edit system and we also discuss the different best practices to avoid irrelevant response from AI agent. We have discussed the three different best practices that is you can change the instructions of AIgent and you can change the model according to reasoning like OpenAI cloud like that, and you can add additional AI agent editor here for the output of this workflow or for the tools like we have did in this session. There are other practices as well. Once you learn how this business works, how the worklos works, how we can add the AI Agent instructions, all those things. And you will discover so much more best practices by yourself once you try so many times with the different applications and by checking the output of AI Agent performance as well. Okay. So in the nest session, we will add these multi channel A agents like AI Agent chatbot, email AI Agent, and form submission E agent, and feedback AI Agent for ecommerce website for our company product and services. So we will discuss each and everything in the NS session. Let's dive into. 24. 9.1 Connecting Multi Channel Ai Agents to Ecommerce Website - Part 1: Session, we are going to discuss how we can connect this multichannel AI agent to our ecommerce website. Let's tap into that. I will jump into my ecommerce website. I have created this website. You can see here at my Appliances. So this is a ecommerce website. I have just created for sample purpose by using lovable dot. What basically I have done. So I have just taken this chatbot code from the Cha JBT and this form submission code as well. I have just provided it to this lovable.ai created ecommerce website. So it has simply generated all those things here. I have provided my company docs in order to create the ecommerce website, according to my product and services. If you already have a website or clients website live in the browser, you can edit the code of that particular ecommerce website by adding this chatbot code and this form submission code according to their preferences. E Commerce website, you can see in the product section, we have got the different products. And in the contact section, you can see here this is a form. Okay, this is simple, but in lovable dot, attractive UI. Okay, then this one, you can use other AI web coding tool like boll dot NU or other AI tools if you want to test this type of AI agents, practically invert any website. You can see we have just created the form submission, AI agent as well here. You can see here this is a bottom, right side, bubble, this chatbot. This is a code chatbot. Basically, the UI color, different because but behind the scenes, working is similar, right? So we have simply connected three different type of a agents here. You can see this is the AI Agent chatbot custom chatbot. Okay, this is a form submission a agent, and this is the email AI agent, a simple company name, right? So what is the process of this automation here? When user use this chatbot. Okay, so when user ask any question, this chatbot will run this particular scenario in our Make platform. This is a AI Agent chatbot. When come to form submission, this particular AI agent, let's jump into our AI Agent scenarios that is form submission agent. This form submission agent works because we have just added this type of webhook URL in the form submission. Now, right. So let's come to this one. For this email, when user use this email to send a query, this type of agent comes to the scenario again. So email AI agent. EI agent will work. Okay, now, let's have another EI agent. Let's come to our Telegram AI Agent. This is Telegram AI Agent. So for this type of Telegram AI Agent, you can add this Telegram bot directly here as well, or even you can add here of this custom UA chatbot, you can add a Telegram or WhatsApp or other communication channels if you want. Okay, you can choose any of that according to your requirements. The process is same. So this Telegram bot, you can add a WTSP, a Slack. According to this trigger, you need to change the output. You can learn from YouTube how to connect the WhatsApp to make.com, how to connect Slack to the mag.com, like that. You will get the information. Now, we have another a agent is feedback collection AI Agent. When the user purchase any product from your company's website, you can send them after one week off product purchase. Google forms link feedback link to that particular user to their email or to the Wat sub number like that. So they will provide feedback in the Google form. So we have used the Google Forms to gather the user feedbacks, details, all those things. Now, it will automatically analyze and it will send a personalized email to that particular user according to the nature of the user feedback. If it is positive, it will send a thank you message with offering products like that. We have already discussed each and everything in the previous session in the feedback collection AI Agent creation, right? How you can automate this that you need to copy these Google forms link and come to these AI agents. For example, let's check in this website. When user use chatbot, like asking any question and E Agent give the response to that particular question. After solving that user, this AI agent will simply send a link to the user. It will show the link here as well in the chatbot. This AI agent will ask a user to provide their customer support feedback. So it will show the link of the Google Form link that we have created here in the chatbot itself. When the user click on that link, that user will provide feedback. If the user use this form submission, when the user submitted with their name, email address, and their question, the AI agent will give the reply to that particular user email. After solving that user query, the AI agent will send a feedback linked to that particular email of that user submit a feedback. If the user use this particular email for query, what happens? The Agent will give the reply to this particular email of that particular user. After solving that user query, the A agent will send a feedback link to that particular email to submit the feedback. Okay, like that. So we have just automate all the things when user use any different channels to talk with our AIHAN for solving questions like that. So we can send a feedback link to that particular user, how we can do that. So let's see the example here. Now into my Google forms. Let's come to this one. No, this is a feedback form we have created. Now, I will copy this one. No, let's come to feedback AI Agent. So we need to edit the chat AI Agent, come to configuration. So this is the A Agent system prom, right? Then I will just write the system instructions like ask for feedback. You can pat this particular link. Use this feedback link to send user after providing solution to the user. Okay, Save this one. Let's try it whether it works or not. So let's jump into our company website. Now before that, we will make this live. Let's come into scenarios again. Let's come to this AIIH. So if you want to learn how you can create a different web apps and the different landing pages, websites without writing any coal or without any experience in the writing prompts, you can check it my white coding mastery codes in my profile. It will help you in a lot for creating amazing websites, web apps, and much more. Let's come to this one. I will use this chatbot. So I will ask this question, which is out of a agents knowledge. Okay, so what is the physically happens here? So when the A agent will get the question, which is not in the knowledge of that particular agent, it will simply send that particular query to the sales team with my name, email, all those things. I should ask my personal mail name, all those things here. Okay, let's check whether it locks or not. I need an identy percent of discount on all points. Let's sell this one. Let's wait for it gets here. Thanks for asking. I understand wanting the best possible price. I cannot approve this blanket 90% discount. So current promotion, I can check if any items, bulk prossing. So you can see if you want to tell me which products or models you are interested. So when I provide this name email, then it will send to the sales team. Let's give the name. Email, and you can see, would you like me to check the current or which products are models your interest in the quantities. So I will just take as all products. All products, each one. Okay, let's send this. If you see here, would you like me to check the current promotion specific items, excuse to our sales team? Also, we have finished with appreciate feedback. Here's a short feedback form. Basically, it is not working as we have expected, right? So in that case, we need to use as a tool to send a link to that particular user in the chat board. Okay? You can see here this is simply given this shot feedback form here. So what you can do it is also they come out of this particular chat UI. So what you can do is you can solve this particular issue by chatting with any bcding tool to solve this particular UI. Now, let's come to this one. Thanks. I've noted your request and details you provided. This is a name, so I'll just confirm it. Confirm. Escalate. Let's see asking me to confirm my details, I have just, please try we'll give the response to the device only which related to our product. Okay, it is asking. Let's jump into our main whether we will get the response from EI agent or not. We'll come to this AI agent again. So we need to remove the data stores we have just created in the previous one because we have used the data stores for different type of EI agents which cause some issues in that. Okay, what I will do, I will just copy this whole. Records. I will delete all those things. Again, call form. Not is basically clean now. Let's jump into this one. I will remove this one. No. Let's before that, what we will do, we will just rewrite the agents instructions, come to chat AI again. You can see you can use tiny URL. Let's come to this tiny URL. Let's use this tiny URL shorter. Basically, what happens here, the Google Form URL is very long. You can short in it. Okay. Otherwise, you can short in the Google forms as well directly, but I am just using this tiny URL. Short in this link. You can give AI's name as well. Verify domains. Okay, shorten link. So, we need to remove this column. Sorry, shorten link. You can give the name as here areas like customer support, like that. Copy URL, come to the AI Agent again. Okay, we will replace with this shortened link. This is the shortened link we will send to the user for feedback. Ask for feedback when user. What satisfy. W Agent response. And you can do this one. Use this feedback link to send to the user. We will remove this one because sometimes overwriting the system prompt can cause some issues because the Agent cannot understand very properly, right? You need to change the different models to check whether our Agent is working perfectly or not. So I will recommend you to use the OpenAI models advanced model instead of the MEI models. Okay? Because they are the reasoning models, you can use that. Okay? Say this let's chart from the scratch, we can write this one, start a a new chat because we have added the start a new chat here as well. Let's come to the scenarios again. AI and chatbot. You can see when the user asks for a new chat, it will start the new chat by sending the new message. Chat start send a new message like that. We'll send this one. Let's check if it works or not. Instead of that, you can use t coding tools to create a new chatbot. Button here when the user clicks on that particular new chat or clear chat the all previous chat will remove from this UI. You can add that particular button as well by asking to the lovable daughter or others. You can see new chat started. How can I help you today? So it is asking. So I will just ask you a question. I need 90% discount on all products. Let's send this one. You see it is working perfectly. So in the previous one, it is simply just given the feedback link. Okay, now it is not Austin because we have changed the system prompt light you can check here. Ask for the feedback when user got satisfied with the agent response. Okay, use this feedback link. That is simple. Okay. So that is the power of writing the system prompt. So you need to check so many times in order to write the best prompt. You can see it is asking me to provide my name, all those things. You can see escalate to our sales team to review special bulb. If you want escalation, please wide a full name and email so I can submit the request. I will submit my name all those things here. Let's now this EA agent should send a message to my sales team. You can see please Sagan and give response to that users queries only which relate it to our product. The problem is we have not enable that particular scenario. In the previous one, we have just converted that tools to scenarios. So we need to enable these scenarios in order to work that particular tools we have added to AI agents. Now I will just come to this human escalation team. Scenario, I will just edit this. I will remove this particular AI agent. That is specifically we have added in the previous session, delete this model. I will save this one. This is a problem we are getting. Question I will ask again I will use this exact response again. So you can say, understand, let's pleasure this escalate it. I will send this name, email this escalate it. Now this AI Agent will send mail to my support steam. It should work because we have published the scenarios, which are the tools to send an email. Cast again showing this. Let's come to scenarios again. Come to make AI agents. Now, let's check our make AI Agent. Instructions because we have just improved that particular system instructions by using loud, so there is a problem in it. Let's configure that and tools usage, human escalation tool, customer issue requires management authority, technical problems. So this is not according to our requirements. That is why you need to use your own knowledge. When user ask query. At knowledge. Send the mail steam with user details. Save this. So when you useNIsstain models like Cha GIP Cloud for writing system proms or for anything, you need to use your own knowledge as well. Let's check again. I will ask a question this copy, send this one. I will copy this. See our data records are working correctly or not. It is working perfectly. Now you can see here it is working now. I'll stop this. It is basically taking bite itself. While testing, you need to use you need to check so many times before submitting the project to that particular client. Okay? So because the issue is to our data stores, let's come to the data stores. Again, selector, Delator confirm, let's jump. So this is going to be very long because we need to check, right? Then we'll get the response. Now what I will do. I'll refresh this website again. So sometimes it's cautious issue, so it is created. Now, I'll start from this I need, discount on all products. Let's send this is working right now. So I will provide my name, all those things. You can see it is asking your name, email, if you want to screen Stip. So let's provide my name again. Let's check. In time. Let's check whether it will send a message or not. You can see we have got. You can see pricing escalation, 90% discount, customer name. This is our name I have provided. You can see product details. A customer requests 90% discount on all products, details provided by customer name annually. Now it is working perfectly, right? And you can improve this particular response from AI Agent as. 25. 9.2 Connecting Multi Channel Ai Agents to Ecommerce Website - Part 2: Let's jump into form submission. Let's come to scenarios, jump into our form submission AI Agent. This is one. I'll make this active. Okay. Now, let's jump into our form sale. Let's give the nail. Quickly. This is my customer mail. Let's give normal question like I need all products names. Let's send this one. So now we are successful set. He's working now. You can see her webhs calling. The ma agent is working. Now let's check our Gmail whether we get the response from A Agent or not. Okay, so it is sent. Now, this is a customer mail. You can see here. This is one I have used 949. Let's refresh this. You can see we have got reply I need all product names. Thanks for reaching out. Los as a professional mail. You can see thanks for reaching out to have us complete list of products, names we have received requests, and our team is comping the catalog. See here it is asking to me provide feedback. Okay, when I click here, it will open Google Forms. You can automate all those things here. So it is all about instructions. Okay. You need to change the AI agent instructions to work properly. Try by yourself. The issue with this main.com is not the instructions, but the model. So I recommend you to use the OpenAI models or Cloud AI models Gemini models to test in all these particular AI agents. Now, let's jump into our email AI Agent. I will just copy this email company mail. Now, I will send from this customer This is company mail urgent needed. Okay? Let's give my dt Got why we are sending this. If you check here, let's jump into our email chain. No, I will just send a high prity email. That is my product got damaged. Now, we will send from the customer mail. I will send this. Before that, we need to save. Okay, now we have sent the message right. Let's jump into our company mail whether we have got or not. Come to inbox. Let's refresh. We have got Agent Ned, so it is not labeled because we have not run the email AI. Now what I will do I will just click on the Run One button. So we have issue. Let's check open Lock. Okay, it is running now. Let's jump into our email. Let's refresh. Basically it is running. Let's jump into. Let's remove these tabs. So let's check our Gmail as well. So you can see we have got this Agent solution. It is running. Let's jump into. So you can see below, Okay, dear customer, we are sorry to hear that your product damage. We need to provide order number, product model, date of purchase, brief, so we need to send a mail to this particular agent. Okay, you can see this is simply asking me to provide the product details. Okay, Agent need it, right? No, let's jump into a company mail. I should be labeled. You can see we have got in the high priority, were not got it. You can see we have got in the high priority. That is Agent solution required. My product got damaged, this is from user. So this is the response of AIH, which is also shown here. So you can track them from here. You can check our sales team as well. You can see customer damage product escalation. You can see it has simply created my name, email, and query, my product got damaged. So basically, we have automate all those things here. Right? If you observe here, we haven't got any feedback link. Why? Because the user are not satisfied with my product and services. I have to understand these points. When I provide email, okay, my solution has solve and my product got repaid, then this AI agent will provide a feedback link in the email in the chat like that. When this AI Agent or sales team solve that particular problem, this AI agent will provide a link to the user to provide feedback. Okay? We have added custom AI chatbot AI Agent, email AI Agent. And the form submission E agent, and we have automated the feedback agent by providing tell you so much things. Try by yourself for better optimization, better automation. Basically, the issue is in mind this whole systems, the make EI Agent provider. Okay? So please use a different AI models from other providers like OpenAI for better results and change the system instructions. Always change the another best tip is before writing the proms in the EI agents, you can use SAGPT or cloud to test out that particular you can test out that particular whole system prompt in that cloud before you try system prompt in any I automation platforms. You can try on that particular system prong here in the free cloudIhatbards, or HAIPIt all those things are a experiencer or expert customer support agent. You are going. So now what you can do, you can provide your own extensions that you are going to put in the automation AIgent platforms. You can put here, you can add a links or you can add your company data that we have done in theming.com. You can add it here. You can test it out this whole chats like AI Agent. Okay? You can do all those things. If you want to learn how you can write the advanced prompts for different types of applications, how you can master the writing prompts. You can check out my complete AI prompt engineering mastery course in my profile. That is, try by yourself. You will get so many things by learning implementing all those things here. So this is how we have created the different multichannel AI agents by using main.com. And remember one thing before submitting all the project to your clients or yourself. You need to come to the web hoouse come here. So check each and everything properly. So to make our webhook very click on the edit. So you need to add APIKey, click on the air APIKey, create a kitchen. You can add EPAKey value you can generate from the HGPT or Chat, but otherwise, you can by yourself not no longer te well. Save this particular AP or you can use the HGV to create a new API for them make.com, or even you can search in the YouTube for better understanding how we can create the APAKey form.com as well. You can use these whole instructions in JGB to create a new APK. After that, you need remove that chat in the ha GPT or like that because that is EI, there is a chance to show you are EPQ to others. But what you can do, create a new API in the hagiBt or other EI models, copy that and paste here and update that particular API with your new key. Let's suppose I have created this APAKy in the ha JBT. I have pasted it here. I will remove this random one. No, I will add I will remove this one. I will add with that. Click on the Create button. Now, you can call it. Let's say you need to add the API for every time you use the webhook as a trigger. I understand these points very well. You can change the name of these particular models. So we have just forget it to edit the name of these particular model names that is right click up. You can just click on the rename. You can give any name like input data or webhooks. Webhook are just to keep it as webhook. You can change the data to name. You can keep it as it is for AI agents, right click on this, you can rename it. Okay? So you can give the customer AI Agent or support AI agent like that. So you can give any name for this particular A agent. The names of each AI agent we have created in the previous session Bselfo, that helps to check the A agents. You can rename this particular AI agent like AI Agent editor. You can do that. So we have just forgotten to rename all the models. Try by yourself, you can rename all the models for every AI agent or every scenario you build in main.com. You can change the name that looks better when any other person will check this workflow. They can easily check that particular AI agent directly. I hope you understand this points. In the net session, we will see what are the different ways of making money online by using these building agents in main.com or other platforms. Okay, let's dive into that. 26. 10. Discussion of Online Money Making Ways Using this Skill: Previous session, we have discussed how we can connect all these different channel EI agents to one particular e commerce website. We have discussed each and everything in the previous session. Now in this session, we are going to see how very ways we have to make money online with this particular AI agent building skill. So you can take any AI agent. You need to create one sample AI agent first. You can take any customer support AIHN for the ecommerce businesses. So basically what you need to do main thing about selling AI agents, you need to explain what they will get. That is the most important part because any businesses founder or manager, they didn't know about what is AI Agent, what is make.com, they don't know, but they only need how you can help them to save their time, save their money, and increase profits. So you need to show that particular results and show off the process. First step is create a sample EI agent automation. After that, try to create a sample website. This is one. Okay? Like, for example, if you're targeting particular businesses who are selling ice creams online. What you can do. So try to create the exact ice creams selling E Commerce website. You can use any free wipe cutting tool like able dot, Weblate Agent three. Okay, Boldt new. Even you can use a HGBTCloud as well to create a sample website. According to that particular businesses, after that, build a sample AI Agent. You can use this exact AI agents. I will provide you all these AI agents blueprint that you can directly download it from in the dashboard and you can easily import here show. I will just take this AI Agent chatbot. Okay, I will provide all this AI Agent chatbot. That is Blueprint. I will provide this come here, three dots. I will click on the Export Blueprint. You can see it is downloaded. I will come to scenarios again. In your account, click on the creative scenario again. So I will provide all the scenarios file. In the document, you can get that. So come here, just click on the three dots. Again, you put blueprint, click on that, choose a file. So this is the file I have just downloaded from this particular scenario, open this one. Save this. Basically, you have got all the workflow. I understand these points. You need to change all those things according to your requirements or your client requirements. Now, after that, just create a sample website which looks like similar to that particular businesses. For example, if that business is selling ice cream online, you need to create a Ice Cream ecommerce website and show them how you can create a AIGent chatbot. Create a sample, Okay, which looks like the name ice cream bought like that. So create them and try to explain them. I can build songs with this particular AI agent for you, which tomato process of customer support can increase the ROI and it can save a lot of time instead of hiring the humans for the customer spot that you can create and you can explain them. When they satisfied with your idea, they can invest money in building AI agents. From there, you can start building A agents for them. This is how you can build the real AI agents with the real businesses. So that is approach. You can also join the freelancing websites like five. Upwork freelancer.com or freelancer dot. That is go.com pepper, however, these are the freelancing websites. You can just come here, create a free account. List O services like I will bull a customer support AI agent for so and so businesses. So just to go for the specific one of broad one because there is a lot of competition and just start building a profile. Before that, you need to have the sample AI agents for specific businesses. Okay, the client need only the results, not your process. Explain the results instead of process. Okay, I hope you understand the points. It will help you to close the clients very easily and very fast. After that, when you start getting the clients, sell to them and get the feedback from them, after that, create your own agency. You can create your own portfolio website. You can come here, use the lovable replate Agent three to create a simple portfolio website. You can give your name. You are all skills datasets. You can provide the in document. You can create the amazing and very good portfil website, which contains your name, your contact information, your skills, and your past projects and real testimonials from the real clients and the businesses. So this is how you can build agency or you can build the service businesses. Okay, you understand these points. This is one way selling agents to the clients. Another way is if you have your own businesses like you are selling the product, if you are looking to sell a product in the online, you can also integrate this type of AI agents into your website as well. The way is you can sell these templates. Okay, come here, click on the Export Blueprint. When you create AIchen for the specific people, you can download this blueprint. Okay. Before exporting that blueprint, you need to remove if you use any APIs in model, you need to remove that sting check any AI agents or any models if there is any personal data or not. You can sell as a template as another way of building income stream by using AIGT. You can build any app. Build a app, mobile app or the app. For example, let's take you can build the app like HAGPT. For the HAT, what you can do? Come here, just add a webhook and add a OpenAI model here and just add another webhook. That is simple. You have built the HGPT app. Okay, so for the UI integration for payment integration, what you can do, you come just here, come here lovable dot Agent three, it will help you to create all those things. For example, without mag.com, even this agent, replegenlovable dot, they can create a exact apps web apps or mobile apps like Cha JB do, even without makege.com. So why we need to use.com is you have any complex idea. In that case, you can use this main.com. You can add this mag.com with your app. Okay? Even it is a mobile app or a web app. I hope understand this why this is a third way also we have another way to build a income stream you can build a chatbot or AI agent which do the online work for you. For example, you can build a simple chatbot like Cha JEPIDO or other AI chatbot, you can use AI coding tools like lovable Agent three to create a chat UI, and you can connect to that particular chat UI with this make.com MCP server. What is actually that, you can easily connect this whole scenarios directly to chatbot in which according to your user requirements, you can build the different scenarios for different purpose, you can build the web research agent or keywords extractor online. You can do so much things, create different scenarios for the specific application and you can connect this make.com Ct server to your chatbard by calling API keys. You can come to the dashboard, and you need to come to this profile. Click profile. And you can come to this API access. Click here. You can click Add A tokens. You are connecting to Make MCP server. You can click select all. You can the name. You can give, for example, custom app, so let's add this. You can connect this particular NCP server to any chatbard by creating UI like Cha JBT or Cloud. Even though you can directly tested out this MCP server in that you OpenAI agent builder by creating OpenAI agent, you can directly connect that particular A agent to the CP server scenarios. So you can automate all those things. There are a lot of ways you can add this MCP server to the external chat services like that. You need to have simple idea how this works, how these ecosystems work, how these platforms works. You can add so many things. If you want more details about this, how to connect the MCP server to different AI chatbards like Cha JBT Cloud or even you can create your own chatbot with this MCP server to create a web like that, you can search in the online for more data in which you can try it out from that. Otherwise, you can follow my account. I'm looking to create a web or simple chat app using this NCP server. Please follow my account if you want. This type of content for you. You can do all those things. No, come to the scenarios. We have discussed some different ways of building income straints. You can create so many different EI agents or workflows for different types of businesses. Life personal assistant EI agent or lead qualification EI agent or market research AI agent like that create number of AI agents in this mag.com. You need to have how this platform works, how to connect the apps. You need to understand how the automation works. Then you can create any type of AI agents in this particular make.com. You can use this grade. Let's say I will check this email AI agent. You can click here. You can click this grid which is available in the apro or paid plans only. This is the layout. This, you can see how your workflow is working. You can check all those things here. You can select a layer as well. This is explore credits, explore what are the apps you are connected, how the data is transferring between the models or actions, operations, you can check all those things from here. If I click on the operations here. Okay, you can check this is the operations done by this particular Telegram AI Agent, A Agent chatbot. This is our main chatbot, email AI agent. So this is another feedback collection, AIgen and this is our AI Agent blogs. You can check all the things from here. So you can check any scenario in the grid. We have different ways of making money online by selling this type of AI Agent building skills, like we have seen freelancing websites. You can sell these EIgent to your clients by reaching out to your local businesses or even with the freelancing websites like five Upwork. You can build a profile on that and even you can reach out in the LinkedIn as well through outreach message to particular business owners, target a specific niche. Then you can target the specific clients. The clients also will believe in your work because you have in past projects with the clients for the specific one. They clients can trust you can give the work for you. You can build your own webps by adding a webhook as well by adding a MCP server to that particular chatbot by using my putting tools if you want. Otherwise, you can create your own AI Agent for you product and services like that. There are a lot of ways you can do. I hope you understand these points very well. These are some moneymaking ways how you can use this particular E Agent building skill to monetize your work in online. 27. What's Next?: Congratulations. You have successfully learned how to create multichannel customer support EI agent like AI Agent chatbot, Telegram AI Agent, Pom submission AI Agent, advanced email handling AI Agent, and feedback collection AI agent. So we have already discussed how to create these five different types of AI agents, and we also discuss how to create any ecommerce website and how to connect all these five AI agents to this particular ecommerce website. And we also discuss what are the different ways of making money online by using this EI agent building skill inmkdt Make com. So I will also provide this EI agent templates you can directly access from this document. I will provide all the five different AI agents template built in make.com. You can see I have just provided that is a custom AI Agent chatbot, Telegram AI Agent, form submission, email handling AI Agent, and the feedback collection agent as well. I also provided with some useful des, like we have created a custom chatbot in the harGPT and the FmUI. I have provided here. You can directly click and you can directly download it from the browser you can use, you can integrate in your website. Not only that, I provided some troubleshooting steps that you can follow in order to solve the AI agent issues like we have seen infinite AI Agent running issue, right? I also provided all the troubleshooting steps can solve that particular common issue, which helps you to build a efficient and time saving agents in them.com. After that, you need to complete a project that is build a support agents for WhatsApp and Slack channel. We already discussed how to create a Telegram AI Agent. It is particularly for the Telegram, but you can build that exact Telegram AI Agent for the WhatsApp and Slack. Okay? So I have also provided each step how you can build that particular WhatsApp and Slack. So you can follow the exact steps in order to create the WhatsApp AI support agent and the Slack AI support agent. You will get all the steps and the requirements in this project. So don't stop here. You can build any type of AI agent in the mag.com. Not only the customer support, you can build for a personal assistant AI agent or lead collection AI Agent or social media AI Agent, which handles all the social media. So you can build any type of AI agents in the mag.com. So try by yourself, that's it for this course. So if you want them more about AI apps or how you can create the Chrome extensions, web apps, Android apps, or even more advanced AI agents in this AI era without any coding experience, you can follow my profile for more advanced courses in future, which helps you to achieve your career goals in this era. Please make sure you follow my account for more advanced classes. If you think this class is provided a value in your life, please provide your feedback or review, which helps me to improve my content. I hope you understand these points very well. I will meet you in the next class till then. Good luck, happy building and happy learning. Thanks for joining this class.