Transcription
Hey automators, welcome back to Let's Automate It. Automation and AI for non-techies.
And a couple of weeks ago, I had this request come in from a client. "Rob, please help me with the chore of posting to LinkedIn every day. I need a system that handles posting for me, but lets me control what is posted. I don't want it to scream AI generated, but at the same time, I want to I don't have time to post or approve every post. So each post will probably have to be AI generated. What's possible?"
So it was the request that came in and, uh, we built a system that's working pretty well, and I wanted to duplicate it and show you the process we went through in order to build this post. So stick with me here. I'm going to show you exactly how we did it. Let's build together. And if you stick around to the end, I'm also going to show you how we tackled the "I don't want it to scream AI generated" part, which is a little bit unique. So, if you're looking for a little bit of a a different approach to these things, uh, stick around to the end. We'll reveal how we tackle that particular problem. And I think it's working pretty well.
So, as always, we like to start with planning. Don't just dive right in. So, um, we put together a basic plan. And this is essentially what we were thinking about. So if we kind of start with the end in mind and work backwards, the, uh, at the end of the day, we want to be able to post to LinkedIn. So we need some mechanism to get to LinkedIn. Now, for that, Blotato is definitely the best tool. We've spoken about Blotato on our channel on a couple of times. Let me quickly jump to our channel. And if we, uh, scroll down, we will have, uh, let's jump into videos. We will have, here we go. Here's a video right here on creating viral social posts, and I think we've got a couple of others as well on, uh, here's another one right, uh, right here, social media automated. So we dive into Blotato. What is Blotato? Blotato is a fantastic service that lets you create, generate social media posts and post them to, uh, onto your various platforms. In this case, all we're actually going to use it for is actually to do the final posting. We will build the post here inside N8N. And the great news is Blotato actually has an N8N community node. So let's go and quickly have a look for BL. You may need to install it. If you haven't installed it yet, you may need to install it, but we're going to open this up and we are going to create a post. So that is the node we're probably going to need at some stage. So I'm just drag that on the canvas so it's there. And, uh, let's go ahead and hook up our credential as well. So, um, let's, uh, we're going to connect this to LinkedIn. So the platform here is going to be LinkedIn, and the list, we will just post this to my own personal LinkedIn for now. Let's leave it at that. Now, you're going to have to go over to Blotato, create an API key, and get that done. Our other videos cover how to do that, so I'm not going to dive into that.
All right. What else are we going to need? We're going to need the ability to create that prompt. Now, that's going to be an AI. It's going to be an AI agent that lets us create a prompt. So, let's go ahead and get hold of an AI agent. So, we're going to go and just grab an agent here. And, uh, we're going to have to put that agent before Blotato. I won't connect anything up here now. I'm just kind of pulling everything together. And, uh, we're going to need to put that together. So, there's our AI agent. We'll come back and fill out the details in a moment.
All right. Then we're going to need a database of posts or topics to talk about. Now, N8N has released data tables. So we're going to be introducing and working with data tables in this video. So I won't dive into it right now, but if you are on the very, very latest version of, uh, N8N, you should have access to data tables. It is still in beta mode. Um, but it's working pretty well. So, um, let's go ahead and grab. We are probably going to have to, uh, get rows. What am I thinking here? So, what am I thinking is we're going to need some kind of database where we are going to have some means of feeding our AI agent with topics or data to post about. We are creating a LinkedIn poster. So, we're going to have to have some kind of information, some kind of topic or information to dive into. So, we're just, today's connected up here. I'm just kind of pulling the the elements onto the page here as we go. Now, you could whiteboard this out. In the interest of this video, because I have actually built this before, I'm accelerating through it, but you'd probably start with this on a whiteboard. Don't get stuck into N8N, but I'm kind of using N8N as my whiteboard for the sake of expediency here on our video.
Okay, so we've got a database, uh, we got a database tool. That's that. And now we actually need topics to post about. So if we come back over to our request from my client, one thing I did notice is he doesn't have time to post or approve posts. So my thinking here is if we generate topics ahead of time, he can bulk approve topics and then we can just let it run without having to prove, approve every single post that goes out. So with that in mind, I went over to ChatGPT and started doing some research. Now, my goal here with this particular workflow is to keep it as simple as possible. As few nodes in this workflow as possible. So, let's go and do some work in ChatGPT. By the way, you could use Claude, you could use Gemini, your favorite model, doesn't matter. And we're going to go and do some research to come up with, I think we are going to need a prompt for our AI agent. We're going to need, uh, and some data for our, uh, data store that we can feed into the agent. So, I did this work a little bit earlier on. So, I'm going to run you through my conversation I had with ChatGPT. Now, I'm not putting this out there as a gold standard. I just want to share what I did to create something that's working pretty well.
So, this was the pretty bad prompt I started off with. "Please do deep research. If you're on a paid plan of GPT, you can do deep research and find the best LinkedIn post structure that encourages engagement and positions the author as authority. I need a consistent structure that can be used for every post. Topics will cover AI automation and running an AI and automation business." I was using my own business as the research tool here. We won't talk about my customer in, in directly. Uh, so we'll just use AI and automation as our topic here. I do not need topics, but rather I need a style and structure of the post that is popular and proven. So basically I said to AI, "Go and find me the best performing LinkedIn post structure." And it went and thought for a while and came back with a lot of information, but let me highlight what it came back with. It says, basically, we need a powerful hook. We need short, skimmable paragraphs. We need educational value and insights. We need some good storytelling. We need some, uh, inspiring and positive framing. We need a clear call to action. And, uh, we need some formatting in here.
So I'm not going to dive into this in too much detail because this conversation did go on a bit, um, backwards and forwards. And as you can see, ChatGPT was being very verbose, giving me a lot of information. But what I want to highlight here is I actually didn't even read most of this. I skimmed right through it and I said this: "Please turn this into a system prompt for an AI agent to get it to craft a LinkedIn message based on this research that it had done." And it came back with a great prompt here. Here it is. This is the prompt right here.
So, that is where I'm going to start. I'm just going to grab hold of this. Scroll down and I am going to copy that. I'm going to head back over to my N8N. I'm going to open up my AI agent. I'm going to add in a system message and I'm going to paste in that system message. So now we've got our kind of basic prompt that we need, which is, uh, something that's going to be critical to get this done. Let's give this a brain. Uh, seeing as we are using OpenAI for everything, we will just be consistent and let's, yeah, 4.1 mini is fine. Let's use that and let's just connect it up to a credential. It's got some credits loaded. Okay.
So now we have a prompt. You see how I'm kind of pulling this together. Uh, I'm jump, not not jumping around. I'm following my plan. I'm kind of asking ChatGPT to help me formulate the prompt, uh, that's going to create a high-performing LinkedIn post with a set structure that is proven to work. So that's what we've done here.
Okay, so it came back. It gave me a couple more things. I'm going to scroll through here. But what I have learned about working with prompts is examples are like gold. Even if your prompt is not great, if you give it examples, if you give it the AI, the output that you're after, it does a really good job of trying to make that happen. So, I said, "Please give me three sample posts that I can use as examples for the AI." And my goal here is to inject this into the prompt. So, let's see what it came back with.
So, it's came back with, um, here's the, here's a post. It came back with a hook, context, value delivery, conclusion, call to action. So basically all those factors, the structural components of producing a great post, came back with those. And this is actually as an example. So it gave me another example and post three, another example, each time following the structure. Then it actually asked me before I got to it, asked me if it would like me to, if it would, if it should add those examples into the JSON trace training data. So basically, what I'd asked it to do here was, I said, "Please give me the JSON structure. Give me JSON structure of each example post in a format that I can work with that is structured." So I asked for JSON, and here it is. It basically took these examples, which are broken up into hook, context, value delivery, and conclusion, call to action, and it broke that into JSON. And so here we've got a hook, context, value, conclusion, call to action, and some hashtags thrown in here because we are posting to social media, LinkedIn in this case. So it gave me the prompt, which I've got at the moment. It's given me some examples, and it's given me structured data to work with.
So let's roll back and I'm actually going to go and grab these examples because I want to put these examples into my prompt. Just going to highlight all of this. There's my examples. I'm going to come back over to my agent. I'm going to just, uh, make this a little bigger. Scroll to the bottom here. "Here are three example posts." And I've just pasted all of that in here. So now I've got my three example posts in there. So I've now got a pretty powerful prompt. It knows what it needs to do. It knows what it's aiming at. It's been given some guidelines, and we're good to go.
Okay. But what I'm actually after now is the source material. I need to kind of populate this knowledge base, this table. By the way, you could use Google Sheets for this. You could use Airtable for this. We're going to use data tables because they're brand new in N8N. And, uh, let's see what they're capable of. We're going to populate that with information. So, we kind of have that raw material already. If I come back here and look at the video at what our com, at our conversation we're having with ChatGPT here, it's given me this, this data structure that we've just discussed. So I'm going to go ahead and copy this code.
Okay. And I need to now inject this source material, these kind of sample posts, uh, into a data store. Now, it's giving me the examples at the moment, and I'm going to keep working with the examples. But what you would actually do in reality here is you would say, "Please give me, you know, 20, 25, 30, 50, as many kind of posts as you wanted to post to, uh, LinkedIn that would be covered by the duration you want to cover." Go ahead and post that. But we'll, we're going to dive into that a little bit deeper shortly. But I just wanted to show you how I got this set up to start with. I need to take this information here. I need to inject it into my data table. So I've asked it for JSON. It's given me JSON. There it is. I've copied it. I'm going to head back over to N8N. And now I need to inject it into a data table.
So, how am I going to do that? Well, straightforward. I'm going to grab a set node. Okay, let's just not connect it to anything. Uh, let's, uh, I tell you what, we're going to just put the set node up here. We'll get it out of the way. And we're going to go into the set node. I'm going to go here into the, hit the pencil icon. And I'm just going to paste all of that JSON that's just come in and save it. And now I have information in here. Now I'm going to go and get a data table node and I'm going to insert row. Okay. And we are now going to insert the data into this, um, row. But before we can do that, we actually need to create our data table. So here we can see we've got, uh, data coming in that's under hook, content, value, conclusion, call to action, and hashtags. So, I'm going to, uh, jump over into, uh, my same N8N. And as I say, if you've got the latest version, you should have this showing now, data tables.
So, I'm going to click into data tables. But actually, if you just hit this little dropdown here, you can go ahead and create a data table. So, we're going to create a new data table, and I'm going to call it YouTube, um, video, um, LinkedIn table. Not the best name, but that'll do. All right. So, we're going to create that table. Okay. So, here it is. It comes already with some, some, uh, rows already loaded, such as ID, created, add, updated, add. We're going to add, so not rows, columns. We're going to add our new column. So, hit column and we add hook, and we, and this is going to be type string. Now, one thing we've noted is once you create a column and you've chosen the type of data, that column, you can't change it later on. Uh, hopefully N8N will release some updates where we can change our data type, but for now, string is going to do, do the job. So, we'll add that. What else do we need here? We need context. So we'll come back here and we'll add a column called context. Also, that's going to be a string. We need a column called value. V value. Add column. We need a column called conclusion. We need a column called CTA, and finally one called hashtags. CTA, and we need one final one called hashtags. Don't remember, just remember to scroll across the page. Sometimes this, uh, add column little button, this little plus sign on the, on the, uh, right on the side here, sometimes that gets hidden off screen. So, we add column. And while I'm here, I'm going to add one more, which I'll explain in a moment. Um, but while I'm here, I'm just going to add it. And that is called posted. And this is just going to be used to keep track of whether we've used this data, whether we've posted about it or not. So, I'm just going to add that as well. And, uh, now we have our table. So that's the first thing you got to do. You got to go and create that data table.
Now we're going to fill it with data. So if I come down from this list, let's see, there is the table we've just created. And I'm mapping each column, V manually. And there is all the information coming in. So I'm just going to jump into schema mode here. And I'm just going to, uh, drag each one of these in. Context, uh, value. As you see, enough, it's come in as an array. So I'm actually just going to drag in the values here. And I'm just going to, uh, comma-separate them. Um, let's just get value one in here. Will it let me do that? Yes. And I'm going to comma-separate the next one and pull in value two. So that's just going to pull in our values that we can work with. And then we're going to get the conclusion in there. We're going to get the call to action in there. And again, we'll do exactly the same with hashtags. One, space hashtag. Next one. And the last one. Okay, done. Not going to worry about the posted column just yet. But what this is doing is now it's going to insert the row.
So if I run this step now, uh, we just need a little trigger node here. So why don't I just grab a little, uh, trigger? Let's just close all this out so we can see what we're doing. And now we're just going to grab, and there's the manual trigger. Just connect that up. It's all it was grumpy about. And now if we execute this. All right. And we go back and look at the data table we've just created. There's data tables. Let's come back into the one we've just created. And there we can see our information injected. Very, very straightforward. We've got our hook, context, value, conclusion, and call to action.
So, this actually serves a couple of, uh, roles for us. So, if we come back onto here and we scroll up to what our request was that was coming in, is this person doesn't want to, doesn't have time to post every day or approve every post. So, each post will probably have to be AI generated. What this is going to let them do, it's not a super simple user interface, but it is a, not particularly complicated, and it keeps things simple because now what my customer can do, and actually what my customer is doing in his own N8N, is coming into here and actually reviewing all of this. The hook, context, value, conclusion. If you just piece these together, they actually create a complete LinkedIn post. So this way they can just come in here, read it, are they happy with it? And they can make some edits as well at the same time. So that's working very nicely and, um, gives them that ability to pre-approve so they don't have to get stuck into human in the loop approving every single day or every single week, however often they want to post. So that is working really nicely.
So let's come back over here, have a look at our plan. So we now have the ability to post to LinkedIn with Blotato. We have our prompt for our AI agent, and we have a database of post topics, source material, and, uh, and topics to post about, and we've injected them into our table. So, if we come down here to the workflow we're busy building, I can now open this one up. I go and find this table. I'm going to be getting rows, and it must match any condition. So, let me open this and show you what I'm talking about here. I want to go and get posted, and this must be, um, is empty. And most importantly here, I want to return just one post. So what's this saying here? What have I got here? I'm saying, "Go to this data store that we've just put three entries into and go and get me all the rows, but only the rows where the is there is no value in the posted column." Remember we added that posted column in, and just bring me back one row. So it's going to bring back the next one. So if I run this now, there we go. We've got our hook, context, value, conclusion, call to action. We've got all our information coming in.
Now, we could just assemble this and turn this into a LinkedIn post, but we want to cover the eventuality that maybe the, uh, customer has gone in and messed with these and they don't read so well together anymore. So, let's feed them to an AI in order to make sure they're polished, consistent, and on brand. So, we're going to do that. So, I'm just going to connect this up to my agent. We already have our prompt in here. So now I'm going to want to feed in this information. So I'm going to want to feed in the hook. So let's do that. And let's feed in our, uh, context. So you can see what I'm doing here. I've got a prompt that's already trained on this structure. And now I'm feeding it the material that it's going to use in order to create our, our post. Pretty straightforward. This is going to force it to cover this structure that our deep research within ChatGPT has shown us that it works. And if you can bear with me as I type through this, hashtags done, and that's it. That's all we need to get in here. And now this agent should go and create a great post for us.
Let's go and run it and see what happens because we're feeding in this information. And let's just move this over to expression. We'll just pop this out and we'll have a quick look here what we've got. "You're an AI agent given a particular job. Here's your post requirements. You're going to be given a hook, which you've given. You're going to be given context, what we've given it. Uh, you're going to give it value delivery and so on and so forth." And it's all done. So, uh, you can screen grab this if you want to go and grab this workflow, this, uh, prompt. So just screen grab at any point here. Go and feed this to your large language model. Ask it to pull out the text, and you will have this prompt ready to go. So let's go and execute this. And ChatGPT 1 mini, jump into table mode, has given us a pro, a post. "Automation isn't about the techn, isn't failing because of technology, it's failing because of the way we implement it. It's, etc." Okay, looking good. I don't mind that.
Now I can feed this into my Blotato and I can just drop this in here, and it, that will literally just go and post to my LinkedIn as simple as that, once you've got your Blotato credential set up. It'll go and post. Now, I'm not going to post because this is posting to my own personal LinkedIn at the moment, and, uh, I'm creating that LinkedIn feed. So, I'm not going to go ahead and post. But literally, if I just ran this, it would create a text-on post and it would put this data in, formatted with the correct spacing. These backslash ends mean either single or double space, and I have a beautifully formatted LinkedIn post.
So, we're actually, in a way, we're done. We've done the heavy lifting, but we do want to do something else. We want to be able to tell this that the post has been done, that we've used this particular data point in our data table, and we've updated it, and we've posted. So I'm going to grab data table. We are going to go, uh, I think we can go with, uh, update row. We are going to connect with our new table, and we are going to, we don't need to update the hook. So I'm just going to delete these out. Value, conclusion, call to action, hashtags, but I'm going to say posted, yes, I'm going to put some data into that. And now it's posted to that. Now I have not run the Blotato node, so I cannot use that because it's not passing any information through the Blotato node, and that's just because I don't want to post. In reality, you would definitely do the setup. So I'm just, uh, actually, don't let me not delete that. Let me rather just connect these two together. Okay. So there, there it is. It's all connected up now. It should know exactly which, which column it's going with. So we're going to go and update that. Oh, at least one condition is required. Uh, let's have a look if the, oh, we'll need to know, uh, let's have, we'll be able to need to match up. So does the column hook string equal? And if we go back here, we can equal to that hook. So it's going to just match on the exact string within that particular row. Does the data inside the, what in, on all the rows underneath the hook column, does it match up with that properly? Let's just see if that'll work. There we go. Okay.
So if we go back over to our data table now, and let's just refresh it, and we scroll across here. There we go. We can see now that posted has been updated with yes. Now, if you're a data table person, you're probably cringing the fact that we use such a crude method to identify the row. You could definitely be using the, uh, ID of the column, something like that as well. Uh, we're keeping it simple, and that definitely works because each of these hooks is going to be completely unique. It won't be exactly the same. So, we could definitely use that as a means of finding the correct rows in all these rows in order to know which row to update.
So, now we have it. So, that this row is now updated. So when we run this one again, remember we're going, if it's empty, this was ID 1. Let's run this again. And you can see now it's brought down ID 2. In other words, row two within the database because row one has now got a value under the posted column. And we're actually only want to return those that are empty. So now we have a basic working workflow. The only thing I'm just leaving that out for now.
So the next thing we need to do is actually schedule this. So we need to add in a schedule. So the way to do that is we are going to add in a cron job. Uh, it's called scheduling a trigger here. And, uh, we're just going to connect this up. Okay. And I want to post this at 9:00 every day, Monday to Friday. So, there's a particular cron pattern that lets you do that. And if I'm not mistaken, I think, um, you can do this trigger interval. Oh, yeah. I want custom trigger interval. And, um, and then we're just going to put the custom cron job in here. So, I actually have just offscreen, I have actually built the cron. So, I'm going to go grab that pattern quickly, but you can just go to YouTube. In fact, go to, let's do that. Let's do this live. Let's go cron, uh, cronix. There we go. I had one already set up. Cron examples. In fact, even better, better, better, better. Let's just go to, let's go into here. Give me, me a cron expression for 9:00 a.m. every day, Monday to Friday. And let's just see what we've got here. There we go. So, right in Google, we've got this expression coming back. So, I'm just going to copy that expression. Come back over to my N8N workflow. Open my cron. Paste that expression in there. And that expression is going to now trigger this workflow 9:00 every day, Monday to Friday. That's it. Weekends it won't be triggered. So, that's fantastic. We now have a working workflow that is going to do the job. Just pop this one back in here.
And folks, there we have it. We have this sorted out. But I promised you, we're going to take this to the next level because this person did say to us, they don't want to scream AI generated. And what I'm foreseeing happening here is we've got very similar source data coming in. We're using exactly the same prompt every week, every day in order to post. And if you looked at these posts back to back to back to back, Monday versus Tuesday versus Wednesday versus Thursday, they'll look very, very similar. We'll be talking about different topics, but the result is going to be quite similar. So, what if we could have a different agent, a different prompt for each day of the week? And I think that is how we're going to solve having a rather unique post that's still going to follow a similar structure, but it's going to be uniquely different for each day of the week.
So, let's dive in and have a look how that works. Let's head back over to ChatGPT and I'm going to walk you through exactly how I said about reducing the AI effect as a, as a process to see how we could make this look a lot less AI, a lot more original. So let's jump in. So I went over to ChatGPT and I jumped into a conversation. Now, this is obviously a conversation I've had before. I had it a while back. Let me just scroll all the way back up to the top here. Uh, am I in the right conversation? Let me just find out quickly. No, this. Let's jump into this one. Okay. So, what I did here is I opened up a new conversation and I went and grabbed the deep research I had in the previous conversation that we've just looked at. And you can see here I said, "Here's some research on posting to LinkedIn that you did for me in a different chat. I wanted to continue the conversation." And I just dropped in the research result that it gave me. So, I'm going to scroll down through all of that and I'll show you exactly what I did with this in a moment. It's worked out really well. It's pretty darn powerful. And this is what I said. I said, "What I want is five different prompts." Now, we're posting Monday, Tuesday, Wednesday, Thursday, Friday at 9:00 in the morning. I want five different prompts, one for each post, Monday to Friday, that will yield different prompt posts, given the illusion that each post is not being created by AI using the same prompt each day. I want posts to loosely follow the structured research, um, that we've been shown how to, that has been shown to work, but I need a variation on day-to-day basics.
So this is what it's come back with. "Perfect. We're going to design a 5-day LinkedIn post engine." And I love what it came back with. It says, "On Monday, we're going to do a momentum story. On Tuesday, a framework drop. On Wednesday, a contrarian take. Thursday, we're going to do, have a look at some trends. And on Friday, we're going to do some future forward gazing and some reflection." So, it's giving me a theme for each day of the week and a post to go along with it, or at least a prompt. So, I was very happy with this, and this was my reply. "I like these. I'm going to give you a master prompt and I want you to take the key elements from the master prompt and ensure that they're covered in each, uh, prompt variation." So, I went and grabbed the master prompt that we've already stuck into our AI agent on our N8N canvas. I pasted it in here. Here it is, word for word, um, all the way through, complete with the examples. You can see the examples posted in here, and I said, "Go for it." And off it went. Now, I did have some backwards and forwards with ChatGPT here, uh, because it kept kind of leaving out the examples. So, I'm going to run through to the end, but essentially, what I wanted it from here, and I'm going to show you right here at the end as I scroll through very quickly. A lot of this is just conversation to get exactly what I wanted, but I'm going to show you the end in mind. So, you'll see what I got.
So, right at the end here, what I wanted was JSON code so that I can inject it into a data table, an N8N data table, and I wanted, I wanted a hook, I wanted context, value, conclusion, call to action, hashtags, and a prompt. So, what I said to it was, "Create me 25 posts. Give me 25 posts ideas for 25 days. Each post must have a hook, context, value, conclusion, call to action, and I want a, um, all the prompts to be different for each one, but not each one. I want it to be five in a row. So I want a Monday prompt, a Tuesday prompt, a Wednesday prompt, and so on." So, actually, let me see if I can't find exactly where I told it that because that does get a little confusing. There it is. "Here is your final order, 25-row JSON data set, perfectly structured for import into your data table. Each row aligns with your 5-day posting cadence, Monday to Friday." So, it's the same prompt for every Monday, same prompt for every Tuesday, but it's different from Monday's prompt and so on and so forth. Please give me that setup. So, it scrolled through all the posts and gave it to me. And then I went and copied the code and then I mapped it into the data table in exactly the same method as we did here. I went and grabbed the JSON, put it in, and mapped it into a data table. And this one I've already got ready. So, I'm going to show you what that looks like.
I'm going to come in here. I put it into a demo table. And I'm just going to use this little jump out to show you what I created. So I created a brand new table. You can see I've got exactly the same headings here. Posted, hook, context, the value, conclusion, call to action, hashtags, but this time I injected an additional column called prompt. And, and, and I mapped the prompt to that. So you'll see here, first one is a Monday prompt, Tuesday prompt, Wednesday prompt, Thursday prompt, Friday prompt, then back to Monday, Tuesday, Wednesday, Thursday, repeats over and over again. And so Monday prompts are similar to each other, but they're not similar to Tuesday, which are not similar to Wednesday, and so on and so on. So there is my data table there.
So how have I made this work? Well, I've gone into here. All right, I've gone with D demo table. I'm going to give me one that's empty. So let's execute that. Okay, here we have it. And now we can actually see the prompt being pulled down as well. Okay. And then here in my AI agent, what I can do here now is delete this entire prompt and actually pull in the expression for the prompt. So now the prompt will change every day, even though I'm using the same AI agent. And as we can see here, I've just got some things that are broken here. I think I used different, uh, capitals. So let's, I think I'm just going to go the lowercase. Yeah. So I just, these values, these, uh, expressions just broke because I did not use capital letters. Uh, you could obviously also just drag these in from scratch again. I'm just typing them in, and hashtag. There we go. So now we're feeding it a brand new prompt for each day of the week, and we are feeding it brand new ideas as well. So it's, it's going to give us what will look like a completely different prompt every single time.
So let's look at a different result. So here's one. "I shipped an automation that made everything slower." So Monday, Monday, or Monday mistakes. So this is something that we did. We made a mistake. This actually did actually happen coincidentally. Um, and now we have this. So let's now go and update this. Uh, what happened here? We're updating the wrong table. Let's just go and update the demo table. What are we doing here? We're updating whether we have posted. So columns, I must, validation error with data store request data. Columns must not be empty. Oh, so I, I loaded up a new table here and I haven't mapped in what I'm after. So I don't want hook, context, value, conclusion, call to action. I do want to keep posted. And this posted is going to say yes. And now we'll run that. And now it's posted.
So if I run this again a second time around, let's go and set fetch the next one. Now, just bear in mind, let's quickly have a look at what this post looked like. "I shipped an automation that made everything slower. A client wanted a full workflow on day one. I built it. We ignored messy handoffs. Basically, made a bunch of mistakes. Don't do what I did. This is how we learned." Let's see what the next post looks like. So we go and execute this. There we got ID 2. It's got the next one. We have here a different prompt from before because we're pulling the prompt and uses expression. So a Tuesday post is all about teaching a clear framework or workflow. So let's see what it does. We'll run that. And here it, here's the framework. "Stop automating everything. Start automating decisions. Teams often get stuck chasing tasks, automating steps without a clear sense of why." Very, very important. That definitely is close to my heart, and we have a very different post coming through, which we now using can post to LinkedIn the following day. And if we go here and run this now, it's updated post number two, row number two, to say that we have posted it, and therefore when we run it again, it is going to pull down number three.
I hope you found this useful. This is a lovely way of tackling a very simple, what have we got here? 1, 2, 3, 4, 5 nodes. That's all we've done. We've given the, uh, the, uh, client the ability to go and edit the posts or pre-approve the posts by looking at data tables. If that you might, you might want to use an Airtable or you may want to use a Google Sheet, and that might be easier for the customer to, uh, work with, and they can go and edit those, pre-approve, approve those, meaning we don't need to have, kind of, post approval after the AI agent has made the post. That takes out the chore of having to be involved with daily posting. It still gets the job done, but we've still got quality control. We have this wonderful tool called Blotato that is actually going to post to LinkedIn for us, and we have this action at the end that updates our data table to say, "Yes, it's been posted." And then every single day, this will just run. Now, in this case, I loaded up 25 posts. So that's 25 days, 5 days a week. So we're looking about 5 weeks of posting here. So, and in 5 weeks' time, I'd have to load up some more or just go ahead and put out, you know, 300 odd posts, which will cover you for an entire year. Let it run. You've got a super simple, super powerful, uh, LinkedIn automation process here that, um, does not scream AI, in that it is unique in every single day.
Do you want to get this post? Well, straightforward. Just copy along with this. Otherwise, come and join us in our community. And, uh, we are a small but very powerful, very connected, very networked community. And, uh, underneath our learning section, you will find all our YouTube videos. And if you come into N8N automations in here, you can run through and find every single workflow for every single automation we have done on YouTube and more. Come and grab this, download it, import it into your JSON, complete with the posts, the prompts, everything that you possibly need to get up and running is right there. Details in the video description. Thank you for spending time with me. I sincerely hope this is useful to you. And until next time, good day, good morning, good night, wherever you are in the world. We'll catch you next time. Cheers.