Transcription
Is learning Natan or Make.com still worth your time? The biggest leap in AI just happened. And thanks to Gemini 3.0 and Claude Code, it is now possible to build fully automated [music] AI systems in minutes instead of hours.
For those of you who don't know, I'm Jack Roberts. I built and sold a top 100 UK startup and got that to 60,000 customers. And now I run a 7-figure AI automation business. And in this video, I'm going to show you exactly what's new, which tools really matter going into 2026, and how you can use them to save time and make more profit regardless of your skill level, so you're not wasting [music] any of your time going into 2026.
So, after building AI successfully for years with Make.com, Natan, Lovable, Bot, how we're building these AI systems is fundamentally changing right now. There is a huge shift. Now you can now automate essentially with words. Let's take a live example. Let's say that you wanted to scrape the number one AI newsletter by subscribers in the world, which at the moment is this one called The AI Rundown. How do you most easily accomplish that? Maybe you want ideas for your social media content. Maybe you just want to get the information. We used to just build workflows. We load it up. We grab our coffee, which by the way, if you haven't had, grab that coffee 'cause we're going to lock in and kick some butt right now. We'd normally build workflows on Make.com on Natan and then labor through it and make it happen or use a template. And this requires skill and time, but now you can essentially just type and build, not just front ends, but full AI systems that just build and do the thing that you want. And you can build these amazing AI systems. This right here is an exact example of what I just said. And literally, guys, I built this yesterday. I'm going to show you in this video how you do things like this. This is that scraper. I click on newsletters. I click on refresh newsletters. And this just pulls down and bear in mind I didn't touch any coding platform for this. I just had a conversation with the model. It will refresh it. 32 articles processing and then I'm going to get full access. Look at this. Look how crazy this is. And guys, this has the intelligence to go into the actual article itself and extract everything from it. Like it really is insane. And I've got all this stuff. I can save this. I can come back to it later and do anything that I want to.
You see, these platforms are amazing and I've spent years with them and I freaking love them. They are integrators. They basically connect things to other things. That one is a writer downer. That one is almost prophetic in its poetry in a way in its prose. But then you have Claude and Gemini, right? Which are executors at the code level. Said another way, these connect things and these do things. And that is very cool. But the thing is these platforms here are not dead. They actually become more powerful. You just need to know how to build AI systems and when to actually use them. And I'm going to show you in this video how you can actually save hours of time and build profit using those systems. Regardless of your experience level, I'm going to show you how you can deploy these AI systems across three levels of difficulty. So if you've never coded before or you're very experienced, there's going to be something super valuable for you.
Now, in my last tech startup, I remember very distinctly, I was drinking some delicious coffee. I spoke to my CTO and I said, "I need to get an admin-level dashboard of the software we had 'cause we had so many customers. We wanted to see what was working." He said, "Great. We need to do that in a few months." Yesterday, I published a video on YouTube. I showed you how you can get a full admin dashboard built in under an hour. Now, that is crazy. So, the game has completely changed. So I want to just I can't home enough but this technology is getting crazy and the things that you can do with it is wild.
So system one, we're going to begin with something called Lindy AI. Lindy AI, if you think about it, is a really good example of how you can very quickly stand up many pre-developed AI systems. So for example, if you check out, we're now at the point where you can basically browse and deploy systems in seamless ways and there's many different platforms to do this: Natan, Make, Lindy AI, there's lots. But let's say that I want to remove the need of whenever I have an email with somebody that I just want someone to schedule it for me and I don't want to hire a personal assistant. Well, we have a system for that. So if you check out on the left, say I want to come down to emails, I click on emails and this shows me all of the templates that exist for emails. I can, if I want to, click on this meeting scheduler. This has actually going to blow your mind. These are the apps that it uses: Google Calendar, Lend Utilities, Forms, Lend Mail. I click on add. This is freaking wild. This is where we're at now with these AI systems. You're going to connect it to your email inbox. You can click on continue. Now, if you're a business owner wanting to deploy this, this is one of the best platforms to get agent systems, email responder systems, like anything from the bazillions of different things they've got set up, ready to rock and roll. So, you can be down here by adding a Meetup link that you want to schedule. So, we're going to enter in a meeting link that we're happy with. We click on continue. If you haven't connected Google Calendar, that will do it pretty easily. And look at this, guys. So the way this is going to work now is I copy this. I click on continue. And anytime I CC in, I love the good bit of confetti that email address into any conversation. That AI will then continue the conversation for us and book the meeting, which is crazy. And you can even see the flow editor. But did you notice at no point did we actually have to design any of the stuff? It just did it for us instantaneously, which is crazy. So for example, this is a message I sent yesterday. I sent an email to myself. I said, "It sounds great. Let's find some time next Tuesday to talk about the McMillan merger. Everyone knows one of the biggest mergers on the planet." And I CC'd in that link that it gave me. Then if I come over to my second inbox, this AI has basically gone out and said, "Great, let's now do this." And it will organize that automatically for us and set something up. Again, completely out of the block system that you can deploy right now. And the point I'm making here is that the skill of these systems is only getting stronger.
So if I come over to home, I can actually tell it to build apps. So, what if I said something like, "Hey, I want you to build for me a scraper. I want you to scrape for me 20 dentists in Leeds. I want to know their email address. I want to know their phone number. And I would like you to write for me a personalized one-sentence outliner for my agency that is unique." Okay, something really freaking simple like that. And I hit enter. We're at the point where we can build these apps by very, very simple prompts. This work for us in the background. And you can see this building in the background, which is really cool. So the way that it works on Lindy is that you have the text terminal on the left-hand side and then the browser on the right and then it will go through everything. It will give you a bit of an update and then you'll be able to chat with this to see exactly how it works. And you can see as this is going through, guys, it's doing a couple of things. So it's waiting for the page to load. It's deploying code. You can see everything that it's doing here in the terminal. It's actually searching the internet. So it's going online for us now to grab this stuff and bring it in an app. So the way that I like to think about this is it's sort of like instead of you doing all the manual scraping and things, you just chat to it and it does it. [clears throat] So traditionally, if you wanted to systematize it, you would go ahead and you'd build out, you can build out the full system. Of course, you can do this in many platforms. You go onto Appy. This is a really easy, simple entry way to get in and start scraping and doing great things by just saying, tell it what you want it to do and then it goes ahead and actually builds the thing. And just like that, guys, Lindy has completed. So we've got the full overview. We've got 20 dentists. We've got 13 that have been found with emails and 20 with phone numbers. We've got 16 websites cataloged. Your agency's oneliner: "We help dental practices 3x more qualified patients." Key statistics: phones, practices, private practice, email success rate 65%. And then it's given us a CSV. And actually, if you scroll down, uh, we've got it here. So now we could do whatever we wanted to with that system. And the point is, with these apps, you can build anything. And again, you just go through the library of things. It's probably the easiest way to get something built and get it started as a complete beginner.
Level two is Natan. I'm going to show you two really important things that you have to understand about Natan. Now, roughly a few weeks ago to a month ago, we had this feature released here, which is "Build with AI." Now, when it came out, it was exceptional and we loved it. Something's happened over the past, I don't know, week or so. But basically, it has gotten to the point where it's pretty, like, it's insane. You still want to learn how to build workflows, guys. But how good this has gotten is actually crazy. And I want to show you right now. So, if I click on this and I said something like, "Hey, I would like you to create me a Reddit scraper, okay? And I'd like you to scrape the top 20 posts from the Natan subreddit. And then I want AI to analyze those posts. And I want you to give me five best posts that would be relevant for my channel, which talks about how to turn AI automations into income, and my school community, which is about building AI systems and monetizing it with no faff. And I want that output to be article 1, 2, 3, 4, 5 with a hook on each of them." Okay, how cool is that? Beautiful. So, we're going to send this one off, which is great. And then the editor is going to build those for us in the background. Now, this editor in Natan is kind of like magic. It isn't perfect. For example, if it's a really big, complex flow, it does have limitations, which is where level three comes in, which I really want to show you, which is exciting. But for building certain flows, guys, it has got to the point where it's exceptional.
So the question is, does this mean then that we shouldn't be learning Natan? Well, the truth is no. You should be learning Natan. And if I show you actually, just I've set it out this way because this I believe is the best way to do it. If you think about your AI journey, the first thing that you really want to do is learn the foundations, right? Foundations are things like what is prompt engineering? What is, if I come down and show you here, AI safety? What is retrieval augmented generation? You want to get a good broad core overview of everything that you're doing. And then when it comes to automations, you want to learn how to build because you need to know how these things connect together. Now, the great news for you is the fact that building these systems has never gone easier. And whilst this is essentially magic in the workflow that it's building here, and look at this, it's crazy. It's absolutely crazy. And we're going to run this in a second. There's still going to be times where this reaches a limit where your ability to code and understand how it works is going to help you. So let's actually test this out right now. I'm going to come down to execute and refine. And the cool thing is, and look at this, it's done. Look, authorization failed. So we just need to change the model, which is fine. So we're going to stop this real quick. I'm going to show you how you debug this. I'm going to get rid of this. Add in a chat model like so. Come down to OpenAI. Beautiful. And then let's just add in our API key. Make sure we've got the right one here. Fantastic. And then once we've updated that API key, you can see it's now working in the background. And guys, look at this. I've got all the stuff on the right-hand side now with everything here. The Natan sub. And you can actually specify what you want it to be. And it's added all this stuff in automatically. Now, you'll need to connect to Reddit by creating an account and get an API. This has done it for me automatically because I had the actual credential in it. So, you'll need to do this. But then the content analyzer finished and we've got one output. So if I come down here, I can check out what it is. And then it's given me exactly what I wanted. And we just did that with text. We used the beautiful phenomenal Glider to do this, which is our fantastic text-to-speech startup that's kicking butt and rocking worlds and saves us hours a day. But fundamentally, guys, that's exactly what we've done. We've just built something completely bespoke and unique just by giving it a simple command.
And level three is going to be using Gemini and Claude Code to build apps that can do effectively anything that we want to. So instead of using any and crank scenarios, which is still amazing and that's still really important, what we can do now is just say to the model, "What do I want?" And it will physically build it for us. Now, there's still going to need to be a little bit of back and forth to these things 'cause they're not what we call one-shot. As they just say, and it happens, but the capabilities for this are literally crazy. Another way that I typically start these pressures off is I begin it in AI Studio.google.com and I'll give it a prompt. The first step that I usually make is I'll get a different model like Claude or ChatGPT to create for me an SOP, a big prompt. Then I throw it into Gemini. Once we've done that, we end up with something really cool. As you'll see in my last video, this dashboard here, which is flipping code for amazing, and it can build essentially anything that you want. You can go over to websites like uh dribble.com where we can find the most gorgeous, sexy designs you could imagine for anything we want. So if I wanted a dashboard, I could type in dashboard data, whatever the thing is. We can take a screenshot of that, give it to the model and get it to build anything in this particular style. You can also grab the HTML text behind any website and know exactly what to do. So this is a really cool one. Again, we got dashboards. We got different settings. You can effectively create anything you want to. And then when you've done that, you can then download it. And when you download it, you can put it into what we call a CLI or something like this. So there are a few different things that you can use for this. You can use Claude Code, the app. You can use VS Code. You can use Cursor. You can use Anti-gravity. There's lots of different things. I am showing you right here, Cursor. Now let me show you something I built earlier. So I'm going to type in Claude into the terminal. And don't worry if this doesn't make sense. It'll make perfect sense to him. This is going to activate Claude Code. So I tend to use a combination of it. Now check this out. This is something that I built in my last video. And I said, "Build it for me." If I say, "Hey dude, open up the app for me." Okay, boom. It basically, what's going to happen now is it's going to give me the localhost where this app is hosted. And you're going to see exactly how this works. And then the really interesting thing with this is all of the building and development is now done on the laptop. And all I'm effectively doing is giving it a prompt and it just does the rest. And there we go. This is the content feed. Now, bear in mind, we can connect this to Superbase. We can connect this to Appify. Pretty much anything you want to, just through having conversations.
Before I test this, I want to show you something over here, which is The Rundown AI newsletter, which is a great example of a good newsletter. Now, when I gave it this prompt, I was like, "Look, this is the website." Actually, when you land in for the first time, it asks you to sign up. What you can do is come over to this and basically say something along the lines of this, which is: "I would like you to create a strategy for scraping the articles from this page. What I want is a unique card interface with the image scraped, the title of the article, and then some information underneath it. And then I want an AI summary of the thing, whatever." And you can give it the URL. Give it a prompt like this. And then hit Shift and Tab and put it into think mode. Plan mode for Claude, if you're not familiar, basically just lets it plan out and create a big, beautiful, big-brain strategy, as we say, to do it. And that's exactly what I did to this one. Because if you look at it, guys, we've got all this interesting stuff. You can click on the articles, you can go into it and see there's loads of different interesting things. And I want to grab all that stuff. And that's all I did. I gave it that prompt, a little bit of back and forth. And if I come over to newsletters and I click on refresh newsletters, what this will do is go ahead and scrape that. And again, you can do this from any newsletters you want to. I just gave it this one as an example. And you'll see this will populate for us as soon as it's grabbed the information. And bam, look at this, guys. "OpenAI trains model to confess when they cheat." I come back over and is that true? Now, what's really interesting here is that usually what it would do, like typically speaking, is just give you, if I come back here, this service level information. So the scraper, guys, is clicking into it and it has already determined what the sections are. It knows that this is an H1. It knows that this is an image and this is text. It did all that parsing work for us. So all I get when I'm thinking of my content right now is a beautiful breakdown from this great, you know, great series of things. And again, we've got Reddit, we've got YouTube, we've got Twitter, we've got like everything in one place. Then I can just simply save this if I want to. It goes into my save file. It's not saving because I've just disconnected it from Superbase. But the point is, it will all go into my saved file and then I can review that, make my content, and just kick butt. Like it's saved me like a crazy amount of money. It's so, so effective.
And so with all that in mind, what is the thing that you should be doing next? Because what I'm doing and how I'm building workflows is changing. So if you are a beginner, it is still really important that you learn the Makes and the Natans to know how to workflow build. There's still a very strong future where you're going to need those skills because even though we're getting to the point where we can give prompts now to things like Gemini and Claude and it can build it, you're still going to want that finer level of granularity and fine-tuning to make the models do what they want to do. So there are still scenarios where I would 100% use Natan and I wouldn't use Gemini. So, it's different tools for different purposes. All this is really saying is that Gemini and Claude have come, they've kind of blown a lot of the vibe coding stuff out of the water, to be honest with you, because they're just that proficient. And what it's done is taken a small number of circumstances, or probably a medium-sized number of circumstances, and meant that it is now the go-to tool to do it. That said, when I'm building workflows and automations for my clients and my business, I'm still using Natan in many different scenarios. So, you still need to learn if, if you want to be great and develop the world's most valuable skill set. I'm even hitting this u this pro way to learn that sort of stuff.
Now, if you are a business, I I think it's most important that you implement something into your business that is easy using a tool like Lindy AI or building a workflow so you can start to get and understand what your bottlenecks are and deploy against that. And for everybody else who's looking at the future who's a little bit more advanced, my honest take on this stuff is that the technology is moving extremely quickly. The delta, the difference between you having an idea and then that idea becoming a reality is getting increasingly short. And if we look at the data, that trend is only heading in one direction, and that is up. So it's going to be easier than ever to build the app. So the only real key differentiator becomes a couple of things. Number one is how quickly you can build it and secondly, what you build. But in addition to that, it's other things like understanding your prompt engineering, understanding different architecture, and building the skills so that when you do get something that actually changes the game, you have enough context and understanding and knowledge to leverage it to crush it. This is a really exciting time and I'm pretty much focused only on showing you the AI tech that will actually help you move forward in your business and crush it. But now you understand exactly how to build this in three levels of difficulty. The next thing you need to look at doing is learning how to build these AI systems from scratch using these tools that I shared.