today I'll show you how I built an AI agent that grows your LinkedIn audience by finding what is already viral in your niche handling all content types text images videos and documents rewriting it in your style and publishing after your quick approval and even turning your own ideas into ready posts so first I'll show you the system in action then I'll break down how I built it in NA10 and finally I'll give you my template to set it up for free so let's dive in with a demo of the system in action so let's go to NAN here is the whole workflow it has three main components the competitor scraping then content agent and finally autopublishing and then let's go to the air table here I have these five competitors that were scraping basically these are five influencers from the marketing niche if your blog for instance is about business you can plug in here Alex Herosi or someone else and then here we have all of the posts from these people that the system has scraped as you can see we have the date posted the username the content inside and as you can see the majority of LinkedIn posts that are not about only text they have also images documents videos and all that stuff and this is partially the reason why the system is so big and so complicated because we needed to transcribe all these types of data and this really takes a lot of work to set it up properly and then we have the URL and the repurposing text this is the stuff that we will use for our own posts right now let me show you it on example here we have the post from Neil Patel it's a marketing guy and let's open the post real quick as you can see it has some sort of infographics some sort of text and it is about the ad spent right let's get back to the system here is the repurposing text then you're just taking this text filling it here editing it a little bit and here is your own text that will be published on your page once you are satisfied with the quality of this text then you will just optionally add an image if you want to make your post more engaging in my case I've added a very simple image that is kind of in style of a tweet you know many influencers are using this sort of content to grab attention because it's very easy to produce but it is actually increasing the quality of the final LinkedIn posts and then when you're ready so this is the content for our own post we are switching the status bar from review to ready and the system will post it automatically and let me show it to you how it works so here I'm going to this part to the autopublishing i'm just pressing the test workflow and right now it will be executed so let me show you my own LinkedIn here's my page as you can see my last post was 5 months ago and let's just update the page and as you can see the new post was just created right now we have the whole text we have the image and uh it was done fully automatically as you can see this workflow has been executed and it went towards the branch with the image which is correct so this is the first and the main part of the workflow just see how much leverage we're getting here it is scraping all the stuff repurposing it for us and then we need only to add some small tweaks maybe add images which can be done using your assistant and then it will be posted for instance you can spend only maybe two hours every Sunday reviewing that editing a little bit of stuff and then it will post every single day during the week how much leverage is in here and now I'll show you the second part of the workflow which is content agent it is your Telegram assistant which you can use to create posts based on your own ideas let's activate it and now I will record a voice message to it create a post about the marketing trends in 2025 in Tik Tok as you can see the message has been received it has been transcribed and then the agent is using his AI capabilities and a tool to write the final post and then send me a message back so let's go here this is the tool to write the post here we're using Perplexity AI to do the web research and find relevant facts to make the post even more interesting then the post is being written and then it's being saved to the air table now let's get back to the main workflow as you can see I've gotten my message back then let's open the air table and here we can see a new post has been created the row idea is uh LinkedIn post about marketing trends in Tik Tok in 2025 and then we have the text itself so it works and now let's go to the NA10 here is the whole workflow and I'll explain to you everything but first if you like the video so far please subscribe to my channel because your support will help me so much in creating quality content like that in the future and now let's begin i'll explain to you every single node in detail so the first part of the workflow is competitor scraping and it all begins with the schedule trigger so we set it to run every single week on Monday we want to get a bunch of fresh ideas regularly right then we're getting our creators from the air table database five creators from the marketing niche in this case and then we are using loop over items to go through every single creator and then scrape their recent videos here these two steps oh sorry these two steps are needed to get the existing posts and then we will filter out what is already existing and what is not yet added to the database and we will work only on the new items so here I'm getting the existing posts then I'm setting the source name so I want to mark this data that this is the existing base this we already have because when we are scraping posts we might as well get the posts that are already added to the database and we don't want to create duplicates right so for that reason here we're using the API to scrape the LinkedIn posts here we need the token at the end of the video I'll show you how to set it up in the detail so right now don't worry about this process and here we're passing the username of the profile that we want to scrape then once we get all of those profiles we're setting the source once Again we want to mark that this data has new source which is called newbase which is different from the existing posts right and here we're doing the merge just combining all these data together and then we're using a very simple script in JavaScript code to understand what are the new posts that have not been yet added to the database and we will work only with them so here is the filtering and then if we have at least some new posts because if we have zero then this workflow for this creator is kind of finished already but if we have some then we're going to the switch and this is the interesting part we will understand what is the type of content inside this post and then we will have four different routes depending on what is inside as I said LinkedIn posts are no longer about texts only they have images videos documents and we want to process all that properly so the first type is document we're just basically checking that we have a specific field in our JSON file right we're checking this field and then we're downloading the document then extracting the text from the PDF because the documents are PDF only there and then we're setting fields basically we want to set all these fields that we will later add to our air table the content what is inside the scraped data and the text of the post then the ID which is the ID from LinkedIn then the date the URL username and blah blah blah so this is the process for the documents then the same process for images and here we're using open AI API to actually do the computer vision recognition and understand what is in this image what is the data the histograms whatever is potentially added to it then we are setting fields once again before we adding the data to the air table and then the whole process repeats itself to the video we have to download the video here it is a little bit tricky i kind of wasn't able to overcome it for a while when we were downloading the video from LinkedIn the format is not fully right and we have to manually set the right extension which is MP4 and only then we will be able to process the video using the OpenAI module and here as you can see we have two branches the success and the error and error here happens when the video has no audio because obviously person can edit a video to their post which is just like beautiful footages and nothing very uh meaningful no words no actual script in it and if we have no script then okay that's fine no audio just do nothing in this case but if we have the audio then we will transcribe it we will use the script to set the content data of this post and then it will go further and finally the easiest part is when the post is text only and of course we just have this field already scraped using the API right then we are once again merging all this data and then using the open AI to repurpose the content and this is where it gets really interesting here I have the system prompt i'm just asking it to act as a very professional like LinkedIn writer and then here is the prompt with the content itself here is the content that we've scraped and then I'm describing what is the format that I'm looking for how I want the final post to look like as you can see I'm asking for three different formats because it will be too boring to do only text posts they are low engagement we want to also add images to make it better quality better value right and for that reason I'm asking for three formats first text only post which is easy then text post plus image with a tweet or yeah something like a tweet so this is the example that I've shown at the beginning i want the system just to provide an idea of what to write in this tweet so it will be bold and hooking attention and finally text plus infographics or charts also another way to create a very simple image but provide more value more interesting data to the user and then I'm asking to structure the output as a JSON file and here is also an interesting thing i'm asking to provide a field called relevant because not all of the posts of these people that we are scraping are relevant to our own niche for instance if we're scraping Alex Hermosi he might has a video about the uh marketing but also he might have a video about something like morning routine and the second part we don't need it we don't want it it's not relevant to our marketing block right and we want it to be filtered out in this place so when the relevancy is not there then uh this field relevant will be just false and no other fields will be populated but if it is relevant then yes we will populate these fields and here I'm providing an example of the good post which has very easy to read structure and I'm asking chbdt to write in this particular style right and finally we have this toggle option enabled output content as JSON because we want it to be easily uh scraped and used in the future and if the content is not relevant if this field is false then we're just doing nothing we're not adding it to our database and otherwise of course we are adding it to error table and then uh it all gets back to the beginning of the cycle because at this point all of the videos all of the posts for this particular influencer are processed and then we can iterate and continue with the next influencer until all of the influencers are finally processed and all of this data is actually added to our database and now let's move on to the second part of the workflow which is the content agent it has a simple Telegram trigger so when a bot receives the message then the workflow will be activated and here the check sender note we want to check that the sender is us because if someone else is trying to use our system if someone else is trying to manipulate it of course we don't want to allow it for that reason we're checking the ID then when this is true we're going to the next stage where we are checking the message type it is either the voice message or the text message if it's just text then it's simple if it's voice we have to download the file and then we have to use the OpenAI API to transcribe it into text and then when we have this command to the AI agent we will merge it all together and go to the AI agent and it will understand what it needs to do right so we have the open AI chat model connected very simple GPT40 then we have the instruction to the agent you're a helpful assistant that answers my questions of course very simple because right now we have the tool created and uh this is kind of the only necessary thing here we can add multiple multiple tools so right now the only tool is that to write posts we can also add a tool for instance to check our statistics how many subscribers do we have how many incoming DMs do we have what are the latest engagements of our posts kind of to track the analytics kind of to track our growth and our progress and so on so on but for the sake of example I added only one tool in here because this is already enough this is already very useful right and this tool it is set up in the separate workflow very simple only five steps it is triggered by previous workflow obviously and here we are using the perplexity AI it's a simple HTTP request which goes to the perplexity API we have the token in here at the end of the video once again I'll show to you how to generate it so don't worry and here is the message we're basically trying to uh search for relevant information according to our topic then it will be used in our post to make it more compelling more interesting right and once we have it then we're just using the chat GBT to write the post itself and here once again I'm providing the whole example of the type of post that I want to create and then when it's finished we're basically saving the record to our error table database and then returning the output and the agent might do something with it as well so when the output is returned to the agent it will usually just send a message that okay the message is created the post is created inside the database and uh that here is the short snippet of it so here it is a simple telegram send message node it is using the same bot account so the message will get to the same chat and on that note the second part of the content of the whole workflow is done and now we're getting to the final part of our workflow by the way you can already be proud of yourself because you've done it so far and let's continue so here is autopublishing here we have the schedule trigger i set it up to run every single day because I want posts to be published every single day and I set the time 4 p.m because it's kind of the best time for my specific audience then we are getting our posts from the database specifically with the status ready which is already kind of has been approved by yourself and only then we want to be publishing them then we're picking only the first one because for every single day we need only one post then we're choosing if the post has an image inside or not if the post is only text then we're posting it straight away using this kind of LinkedIn note it already has an integration with any which is very very convenient here we're connecting our LinkedIn account and then choosing the person name or ID because in your case you might have both the person account and the company account connected to your LinkedIn of course you can use the system for your company but in my case I'm using it for myself so I'm choosing Ole Malikov and then on the other branch everything similar happens we're just downloading the image first because the post contains the image and then we're uploading everything to LinkedIn and finally we want to mark this post as posted and on this note the workflow is finished and now as promised I'll show you how you can copy and set up this system from my template for free and start using it for yourself but before we start I have a bonus for you i'm selecting five people for a free customized setup completely done for you all I ask for is your honest review as I'm building my portfolio so we will set everything up specifically for your use case we will analyze the influencers in your niche we will analyze the style that you want to have at the end and we will make it set up and running the only thing you need to do is go on the link down below there you will find this form just leave your email your LinkedIn profile and book a quick call with me where we will discuss all the details so if you are interested submit this form if you are not interested or if these spots are already finished when you are trying to check it out then I have the whole instruction how you can set it up yourself manually i think it will roughly take you 15 minutes depending on your technical level of course and let's begin the first step is obviously to create an NAN account if you still don't have one just go to the nitan.io KO here you will be able to create an account for free just press the get started the process will be very straightforward once you have an account then we are going to the next step you will need to download the JSON file with the template of the first and the main part of the workflow just go here download this JSON file and then we're getting back to the NAN and we will need to import it so here I will proceed to to this part these are your workflows probably you will see nothing if you have not created any workflow yet then you will press create workflow and here you will have the empty space don't worry it is okay here on the top right side we need to press these three dots and here we need to choose import from file and here we will use our file that we've just downloaded press open and bam we have the whole system in here so it is very convenient you don't need to create every single note manually right now we have everything and you'll need just to tweak some of the settings of some of the nodes so the API connections are set up correctly now let's continue this is our instruction it will be also linked down below check it out the next step is to set the error table fully we have this step then this and this step three very simple steps so let's copy this link you will need to copy the database let's go to the other uh other page paste the link open it up and here you can see on the top left copy base so you don't need to create it manually you will be able just to copy it press this thing and then create an account inside the air table very straightforward and then let's get back to the NA10 once we have it then we need to create an air table access token and grant necessary permissions and only then we'll copy the token right so let's get back to the air table here on the top right I'm pressing here then I'm pressing builder hub and here I have my tokens so you'll need to press create new token here create a name for instance NA10 token then choose the scopes you will need this one then you will need this one and then you will need this one three scopes and then you will need to give an access to your database once you did it then just press create token and you will see the token for only once and you have to copy it right now copy and press done and on that point the this token is created and we're going back to the NA10 and we need to go to the first air table note in here press it and here we'll need to press create new credential and paste our token in here and just press save and as you can see the credential is successfully tested which means it works absolutely fine we're connected to the air table now and then you will need to manually select these databases that you want to use so you're selecting LinkedIn posts even if you see that it is already selected don't be tricked by it select it manually many people reported this back because it has kind of to connect once again once you change the air table token in here and select creators in this node and then this node is fully set up and we need to go through every other air table node like this one and let me show you this one this one and this one set it up in the same manner and error table will be working fully at this point and let's continue getting back to our instruction the next part is to set up the API this is the software to scrape the LinkedIn data so just go to the api.com here you'll need to create an account probably go to the console and then once you create an account you need to copy your API token go here then when it is downloaded by pressing settings and here API and integrations and here you will copy your own token press here once you copy it just go back to the NAM and in this note A5 scrape LinkedIn posts you will need to substitute your own token and for many many executions the software is completely for free so you will be able to scrape LinkedIn for free which is awesome right on this point API is fully set up then we need to set up the OpenAI API account just go towards this link you will need to create an account here also very straightforward process once you have an account then you press this here you will find your API keys at the beginning you will have no keys you'll need to press create a new secret key then you create it then you copy it and also go back to the NIN and use it as a credential in here in the first Open EI note just go here press create new credential and then use your API key i will not do that right now so just close and this node will be activated but before you're able to do the successful requests to the OpenAI API you will need to add a little bit of balance i think so so you will need to go to the this link and here you can see that for instance in my case I have the balance of two bucks you can also add two bucks and this will be enough for many many executions right uh I know this is not very convenient but still this kind of pays off because if you're building different systems you still need the AI engine to run the complicated functions so of course open is absolutely worth it so once we created that once we added a little bit of balance then the openi is fully set up and then we need to finally set something small in the database inside uh air table just go to your database in here i'll press back here and as you can see right now it is populated with my posts that I've scraped you can delete all of them just select them delete them make it absolutely empty and then go to the creators part and here you can paste the usernames of the accounts that you want to scrape from LinkedIn the username basically is the part of the URL of the LinkedIn profile just open the URL find this part and substitute it in here right and on that point the system I guess will be fully set up right the last thing we have to just activate the workflow so as you can see uh on this top right just press from inactive to active and this thing will be able to run fully at least the first part of the workflow which is competitor scraping these two I will explain in a bit so let's get back to our instruction the next part is the content agent which is optional you don't necessarily need it to run the whole system because the main part of course is scraping the competitors but if you want of course go do that first of all you need to go to the telegram to the bot father here I have open the chat with this bot father and you'll need to create a new bot just press new bot then you'll need to write a name I don't know like bot bot bot and we need to create a username for this bot for instance bot 239 bot uh bot like that okay This is the name it is already taken okay let's just add more random digits right now we have created the bot and now we'll need to copy this access token we'll use it to connect our N10 to the Telegram right so let's get back then we'll need to copy this token we did it then we need to use it in our N10 node so at this part just go here press here and you'll need to create an account create new credential paste your token and use it and also use this thing in this node as well and on this point the Telegram stuff is created then we'll need to set up this check sender node which is quite simple you'll need to test the workflow you'll need to press test workflow right or test this node only then your bot is listening for your messages in this case I have this bot which is my LinkedIn strategist let's send something just hey bro we send this message and as you can see the message has been received and in this case uh our ID is this so this is ID of my telegram account i'm just copying it and I'm just going here and pasting it in here so in your case it will be a different ID we want that to check that our system can be used only through your account not anyone else's right so the next step once we have created this if okay this is fine then we need to download the template of the second workflow which will be used as a tool of our agent right so once again just go towards this link download the JSON file and uh once again we will need to press uh we'll need to go here we'll need to create a new workflow press create new workflow then once again press these three dots import from the file and here we are selecting another workflow which is LinkedIn post writer just press open and here it goes and here we'll need to set up the perplexity openi is already set up just select the right credential in here the right account and also select the right credential in the air table account once you have it you need to set up only the perplexity and the instruction to do that is right here very easy go to this link here you will need to create an account as usually then you'll need to create an API key for instance this is my API key i also have some little bit of balance almost three bucks also top up your balance copy this API key then get back to the NA10 and just use this uh token inside this node right here beer and then goes your token once you did that the whole system is set up I guess let me just check the instruction yeah so at this point the second part is set up and the final part which is very straightforward only one step it is the auto posting let me open it once again here this setup this setup this setup of course the only two things we need to set up is these two LinkedIn notes just open it and here we need to connect our LinkedIn account just press create new credential and here you will need to select standard right and then press connect your account and then you will have a new window popped up just connect it with your LinkedIn and then you will be good to go it will work and uh select the same credential in this node as well and you'll also need to choose the person name or their ID in my case I have only one ID connected to this LinkedIn account so it is my name and on that note the whole system is set up and running just press from inactive to active and it will be executing every single week to scrape the stuff but also the content agent will be receiving the messages whenever you are ready to send them and then the autopublishing will be publishing new posts every single day if you have them in your database and if you like this video check out another one where I built an AI system that helps create viral Instagram