00:00
Project Intro
Introduction to the client project: automating local SEO page generation.
00:34
Systems Thinking
Importance of systems thinking before diving into detailed tasks.
01:20
Make vs. n8n
Comparing n8n and make.com for the client's workflow.
01:47
GPT-5 Integration
Using GPT-5 to query and structure content for better results.
02:34
JSON Output
Why JSON is preferred for structuring data for large language models.
03:45
Workflow and Understanding
Importance of understanding AI workflows for effective implementation.
Summary
CopyDetected: English
Copy transcript
00:00
What's up, y'all?
00:01
So I'm just recording myself as I work on a client project so you can see how I'm tackling things and, yeah, you know, where things work, where they don't.
00:12
So the actual MVP for this client is essentially to create a automation workflow using AI, to be able to spit out local SEO or GEO ready, optimized pages at scale for an agency.
00:34
And the options have been using make.com or naten.
00:38
But what I realized as I was going about doing this is that if you are someone who's wanting to understand the systems thinking behind some of this stuff, before you just get into doing all the nitty gritty detailed stuff, it's important to actually kind of take yourself through a workflow that you might have already used yourself somewhere else that's similar.
01:02
Right?
01:02
And this is the whole thing around systems thinking because there's many ways that you can approach a problem.
01:08
And for this, as I was working on the client, initial setup, I was thinking to myself, is there a much easier way to do it do this at least for an MVP?
01:20
And so I thought to take the original make, actually, the original n eight n, workflow setup, which you see here.
01:30
I then started to try and apply some of it into make.com, and I realized actually there there are some parts of make that make this better for the client from their experience.
01:40
If I was to do this all in n a n a 10, I think it'd be overly complicated and an unnecessarily complicated for them.
01:47
So in make as I started doing this and what you're seeing here is actually just, I kind of refreshed everything because I took it over to g p t five to query some things.
01:56
So what I actually did was I put in, I put in a lot of the work that I did in n a 10, and then I just kinda stripped everything and went into g p t five.
02:10
So if you see here, I created a folder.
02:13
Right?
02:13
And I uploaded the files.
02:15
Okay?
02:16
Now these were specific guidelines that I was given, and this would be feeding into the automated workflow that I'm currently setting up.
02:23
But to make things simpler, if you actually just put the files into this folder and then give instructions, right, here, you can actually get something really good.
02:34
And the output format output format, you can see, is this in JSON.
02:39
Okay?
02:39
Now when you actually give JSON to LLMs, especially GPT five, it tends to produce way better structured aspects.
02:45
Right?
02:50
Other LLMs actually will may prefer things like XML or, similar sort of structures.
02:56
I prefer XML, but, you know, JSON is the way to go generally for more of the western models.
03:03
So if we go back in here, you can see I just said create this.
03:08
Create a geo targeted service page for us.
03:11
Just that.
03:12
Just this.
03:12
Okay?
03:13
Then it outputted the JSON.
03:15
Okay?
03:16
And the content is all in here.
03:19
But, of course, I wanna see what this would look like as a blog post.
03:22
And so I then said, spit this out a normal copy to copy and paste into a WordPress site.
03:26
And then it spat it out.
03:28
Now the next part of this process would be to actually take this and then use, a tool like page optimization pro, or pop to then rate it to see whether or not it hits a 95% grading or above.
03:45
K?
03:45
But going through this process, you can start to see how if I can understand how it works in one way I've been doing it, you know, in generally, then I can take the thinking over into tools like make or any attempt, any of these other automation tools.
04:03
Because this is the thing that I think a lot of people will forget is that a lot of these things you can get AI to do, but I think it's still important for you to actually understand what's going on.
04:14
And one of the best ways to do this, for for myself at least, is to kinda reverse engineer things or try solutions in a way that I already know how and then map it into something that might be more user friendly for the client, the prospect, whoever it is you're working with.
04:28
Okay?
04:29
This is a valuable part of the process, when when doing these sort of things.
04:33
Alright?
04:34
So there you go.
04:35
Thought I'd just share this with you guys.
04:37
If you have any questions, let me know and speak soon.
Made with Bullet