Why You Need to Try Out N8N Agent Workflow Builder

We already had workflow builder for years. But the thing is like it's still hard to use because like you have to figure out and [music] when you only have the prompt, it's just people just write whatever comes to them, but they don't even have to like spell it. They just type something and it works because AI understands everything. So that's the magic part. If you write a prompt that doesn't work, just change your prompt and redo it. Hi, my name is Demetri Bonichi and I'm a content creator, agency owner, and AI enthusiast. You're listening to the AI Agents podcast brought to you by Jot Form and featuring our very own CEO and founder Idkin Tank. This is the show where artificial intelligence meets innovation, productivity, and the tools shaping the future of work. Enjoy the show. Hello and welcome back to another episode of the

AI Asians podcast. In this episode, we are going to be diving into some interesting tools. We're going to be talking about the N8N agent workflow builder, why you need to try it out, and why you also might want to try out another super and fun thing that we have going on at Jaw Form. How you doing, Adicanin? >> Hey, how are you doing? Uh yeah, I was really impressed like this uh this new workflow builder, agent builder just came out uh I think like last week and I was really excited to try it out and um yeah, I think it's it's it's um it's a good thing to feature because it makes your job so much easier to automate stuff. Uh you just write your prompt and it just creates this agent. it creates this workflow for you, includes the agents in it. So, it's

just um really uh easy way to build an agent. >> So, I think it's I think, you know, uh it's going to be an exciting episode to feature them. >> Yeah, I totally agree. Um well, I mean, just to kind of kick things off, I think it's uh important to note that, you know, there's a lot of tools out there right now in the world of um AI agents, AI agent builders, especially in automation. So, what I'm going to try to do is just in the simplest terms explain what's going on, what's the opportunity here, what's the new kind of uh things that we think are, I guess, uh good about it, what um we do versus what they do, all that kind of fun stuff. So, um to kick things off, uh let me just pull up my screen here. Trying to get the

right tab shared. I think we're good. [snorts] Yeah, there we go. All right. So, naden um is essentially a imagine if you took make.com made it more advanced, made it a little bit better price per cost, maybe slightly more technical. And a big learning curve for people was the building. Okay. Now, uh, what we're talking about in this episode, as I mentioned, the name is the agent workflow builder. The AI agent workflow builder, right? So, what does that mean? Well, the AI agent capabilities of this tool are really powerful, but it's difficult and confusing to some as to how one would maybe launch the uh let me this tab, sorry, launch the uh workflows that they'd want because you have to get it off the ground somehow. You have to build it. And unfortunately for the average person, that's not the easiest thing to do.

But they've managed to make it easy. So the workflow builder turns your natural language prompts into working automations. So uh you can describe what you want to build and n will essentially generate a draft workflow by adding and configuring nodes and writing up the logic for you. From there you can refine, expand or adjust the workflow directly in the editor. [snorts] So for example, you can press build with AI. You see they have a little template here that's like build a pipeline that accepts PDF CSV or JSON files through an NAN form. Chunks it into token segments, generates embeddings and stores it in a vector database and then boom, right? We have a uh complete workflow made right automatically. That is the worst quality I've ever seen in regards to a video. I thought this would be better. Um but I think you'd get the

get the point. >> They show us that on the uh homepage they actually had like screenshots. I think they were even better. >> Yeah. Okay. I was about to say that was weird. Um, and we can show you I can show you an example real quick just in the app so that it's >> No, actually at the top like there there was Okay, show your example. That's better. >> Yeah. I mean this these are actually >> um looking great. >> Yeah. So imagine just typing a prompt for example when I say >> um this one you can be like when a new IT ops can onboard new employees. So when a new form submission is in NAN like imagine this was the prompt make a AI agent that onboards a new person on the team and um if they are manager add to some channels

if they're not a manager update their profile in Slack right so it's uh pretty interesting how you can come up with the ideas like for example this is one I made and um it's actually pretty easy to to make these but I'm going to press save so I don't lose this. But I can go to here, press create workflow. Um I apologize. Sorry. Um I already did this. So I basically said let's make a news summarizer. So what this does is it made an AI agent that uh essentially summarizes the world news and tech news from the last 24 hours. Skip your comments. Title should be world news, etc. Limited to 10 bullet points. Start it today. And then it has a open AI model that it connects to a the AI agents in general. How they kind of work is you have the ability

to connect it to a specific uh chat model. Like here we have um GPT4.1 mini which is like free with open 8 end credits. Obviously you can add your own um API key and whatnot but uh this is kind of the default that you get. And then uh you can also connect what are called what's called memories and tools. So memory includes things like storing stuff inside of naden which is really convenient. Um referencing uh other different things like MongoDB chat me uh chat memory um Postgress um a couple other things that you're probably familiar with. The simplest thing is to just like add the memory inside of here, but I don't really need this for the context of this agent. >> So the AI agent is kind of like an um like an employee. You can think of it as like an employee. You

can give it any task that you want to give and as long as you can give it the tools that could help uh that AI agent to to accomplish those tasks tasks >> it can actually do that. So basically what you are actually doing here is you are programming but you are programming without actually writing code. your programming by just uh if you think about it, programming is like uh basically it's a language. It's like you when you're writing code, you're translating um your language, human language, English to machine code. Um and basically you are telling computer what to do. But in this case now we have this AI that's that can both understand humans but also can turn that into machine language and execute it. So this means and it's basically an automation. It can work anytime anywhere. Um so basically what we are

doing is we are writing code here but you know it's the ultimate no code that without writing code you're able to kind of instruct AI what to do. Um so that's why I feel like this is like the the the li um the possibilities are endless. Uh there's just like so many things you can do with this. >> Yeah. No absolutely. Um, I think you're you're spot on uh with all that commentary. And um it has the ability, like I said, to have an AI model, uh the memory, and then last but not least, tools. So, as you can see, it automatically added a couple different tools here, which are RSS. So, reads data from an RSS feed. So, the Verge is a common I don't know if you guys are familiar with the Verge, U, but it's essentially a I'll pull up the website

real quick just to show what it is. Um, it's like a tech blog. Oh, did I I pressed close? Of course I did. Uh, it's a tech blog. Sorry. See if you see this now. >> Mhm. >> Okay, cool. Um, so it's a tech blog as you can see. All right. And then let's scroll through as you can just take a look. Right. So it's pulling it from here. And then it's also pulling from if I'm not wrong um I think it's pulling from another one. Trying to share this tab instead. Sorry, it's my screen sharing is bugging out. There we go. Uh it's also connected to another tool. So feeds.b Oh, BBCI, which I'm guessing is a specific like world tech news blog as well, right? So it's just getting the RSS feeds. >> BBC International. >> Yeah. >> Yeah, >> you're right. Yeah.

Yeah. BBC International News World and then that's what it added by default. And then it has other options for tools. Um you can code a tool by writing a tool in JSON or Pyth or JS or Python. So JavaScript or Python, which is really cool. You can call an NADN workflow tool. Um you can use MCPS which is really big. You can uh use different vector stores. Um you know referencing different databases is always a thing um in anything like this. Uh and that's why I think it speaks to your comment about is this an employee, right? Cuz you know you add you know obviously something like what's it called? Anthropic just released their new SDK capabilities the other day, right? Um, I feel like this is something you could add tool-wise onto uh what you're doing here. You can reference an Air Table database.

You can reference uh um Notion database. Um you can essentially do whatever you'd imagine. Uh has the ability to for it to take action and get info, right? And then parse it with the AI model. And then from there at the end it'll output a news summary for this example. So let's just run this workflow so you guys can see how it works. Tech news world news. It's running through. So as you can see it grabbed that stuff from the tools. Did this here executed successfully. All right. So let's open this up. Um all right. So this is the uh input output. So let me just I think I sorry I actually only pressed the one part. So I'm going to press execute. Right click here. I should have just pressed execute workflow. I just misclicked. So as you can see this is the output.

So um world news. The US announces new sanctions on countries that I will not state to avoid awkwardness on the podcast. Uh France is uh reportedly backing away blah blah blah. So this is like world more world news. Um >> you can probably read it easily like if you click on that table at the top instead of JSON. >> No, it is the it is actually table. Yeah, JSON's this one. >> So um let's talk talk more about this. So, by the way, just to point out, I could totally I think to probably clean this up because this is what the automation made. I probably would just for the sake of this remove world news um so we can focus on tech for us. And then like if there's any um >> big AI blog RSS feed um that we could find to add um

we totally could do that, right? It's like something that I think would be um would be something big to do. So, we could always find that and replace that. But if I run this again, it'll just focus on the tech news. And then just to point out, I obviously had this set up automatically, but I can have this running on a recurring basis, so I'm getting news reports, stuff like that. Right. So, no recent world news. Um, Fujifilm announced a new camera. That's cool. Pokemon introduced mega evolution cards. Nice. Um there's new uh MEV1 cameras that released. Um Leisure unveiled the Nano Gen 5 signer device. Open AAI's chatbt Atlas browser team is working on improvements including profiles, tab groups, ad blocker, and enhanced AI agent performance. Espresso 15.6 portable monitor reviewed as transformative for remote work with a thin light. You know what? I

actually own this. >> That's funny. [laughter] That's actually funny. So, the Verge made an article on this guy. I I do and it came out five hours ago. This article >> make this more powerful is um kind of like if you addit the memory to kind of know about what you care about and then just filter out and which the AI can easily do filter out all the news and just show you like the stuff that you would be interested in >> and how can you find like what you're interested in? I don't know, maybe from from your browser history, from your um emails or something like that. Um then it can probably just show you the interesting stuff like you know just like this monitor uh that whatever products you use uh you would probably want to hear news about them. Yeah. >> Yeah.

Like for example, if you're doing a, you know, if I had pulled up some other blog or some other article getters um and I was doing a physical tech blog, right? Or not tech blog, physical tech podcast, like I could get updated on it through the NAN tool because a couple things just got released like Apple announced the M5 MacBook Pro. This could be something to reference. So, if it's all like a physical tech blog, yeah, you could totally change the workflow to be um focused on the [snorts] physical side of things. And for our podcast, obviously, we're focused on AI agents. So, it's all about what you want to do. Um, and adding more memory or changing the prompt so that it's focusing specifically not on world news and tech news, but just specific types of tech news. And, you know, obviously upgrading the

model rather than using the free model would be helpful, too. So, it's going to do a better job. But yeah, uh overall, that's kind of how you can set one of these up. And it even, which is cool, gave me an example um bonus and optional one. So, this is a little tag. You can add like little tags to uh the um the the spot here. And um >> so what what can you add to an AI agent? Um I see that um ch uh chip or chat model. I see memory and I see tools. Are there any other types of things that you can add here or are these like um okay these are all the things that you can >> the tools inside these are all tools right these are all tools >> but are there any other like types like um like

AI model and then the memory and tool. Are there any other things that is possible to add to an AI agent on NA10? >> No, it's that's pretty much it. It's just the the [clears throat] tools are extremely extensive. >> it covers everything. >> It covers everything, especially with and this is where MCP kind of comes in, >> right? The ability to reference endpoints and then have it with a model break down the information extensively, right? That's the the key thing to note here in the capabilities and why and why it is such a positive thing and then after the agent runs you take the output you put it here um or in the case of this right what you can do is an example that it gave us is we can give the output summary to send it to our own personal email as

like a daily newsletter for ourselves or something like that right um and the rest of the tool has the ability to um do a lot of different AI capabilities. Obviously, more AI agents. It can connect with models. It has the ability to AI classify stuff. It does a really good job, I think, as a product adding little bits of AI here and there. Um has human in the loop. A lot of really cool um capabilities that other tools don't have right now. So, if I were you, I would definitely uh check out this new AI agent builder and mess around and try to build something um that can get you some quick wins in your in your day-to-day right now. >> Yeah. Um so, I'm going to also share um something interesting. >> Uh at Jot Form, we have been working on something very similar.

Uh so, I'm going to share those as well and I'm going to compare it to this one. Uh so this Jatform already has this feature called workflows because like uh even though Jatforms in the forms business um uh we call it we say like you know every every journey with starts with a form but there is like all these next steps right uh so just from workflows kind of help you with that. Um, you know, it starts with a form. Maybe you need to get a payment. Maybe you need to do approval. Maybe you need to get it signed. Uh, get the documents signed. Uh, maybe you need to send an email, do some conditional stuff. So, we have this approval uh stuff. And we have we actually have been uh on a beta uh on beta. We actually did exactly what uh the NA10

did. And this is what our looks like. And uh basically you can just put your prompt here. Uh so my prompt is uh when someone sends a feedback on our launch party feedback form on Jot Form analyze the rating and and feedback and then send an email to itform.com about it. So this is my prompt. I send it um and basically it's just going to build me a workflow. While it's working on this, I'm going to show the exact same prompt on NA10 workflow builder. So you can click on build with AI or you can use this uh you know NA10 AI on the right side. So there are like you can ask questions to it like just like a help bot or you can just ask it to build stuff for you and there are some examples here and you have also shown us

one example but basically I will answer the same prompt here when someone sends a feedback on our launch party feedback on jot form and basically and I'm using the free version um so just I I still have some credits left and I will go ahead and do that and basically this is also working on it. While this is working on it, let's go back to Jot form. Okay. Uh so we can see it. Actually, it's very horizontal. So I'm going to change this a little bit. Uh basically, how do how do I change? Okay, I will just do this. Um so it looks better. Uh maybe zoom in a little bit. Okay. So starting with the feedback form. So we have this feedback form, launch party feedback form, and we have this rating field basically. and and it start with this feedback form. It analyzes

feedback. It's asking me to, you know, uh improve it. Enter some details here, sentiments here, good or bad. Uh I could enter them here. And um okay, there's some error. Uh this is still still in beta, >> but I wanted to share it because just NAT just released something very similar. But yeah, I'm not I'm not going to focus on that. Uh but um so so basically you can add an AI field here. Basically it added this AI field here. Custom prompt here. This is a custom prompt. It's analyzing the sentinent and then the uh analyzing the rating because in the prompt I asked both sentiment and the rating and then it's sending an email to me and I can do all kinds of stuff here. I can improve this workflow. I can also integrate with uh all these other products. So pretty much you

know if you already have a jot form you probably may not need uh NA10 J form already covers all these stuff but it also covers uh all these you know included features like approvals jot form sign all these things uh like this or document generation um so this is what it looks like at jot form let's go back to any okay looks like it it finished it uh so I asked for the prompt searching notes editing notes connection note updating parameters how to set up. Uh so basically once you do that you go here and you can actually connect it with with your J form account. You can click on it enter your you know form ID uh first like maybe enter your credentials. Um the same way it connected with Gmail because I wanted to send an email. Mhm. >> So as you can

see um I also need to connect with Chip PT. I can add a tool here if I want to. Um and basically this also uh created me this um workflow for me. Um so I think it's it's just beautiful. uh like just automation is so simple with all these tools and whether you use like N10 or the Jatform workflows uh just like you can automate stuff very easily and um just save so much time. >> Yeah. No, absolutely. I think um it's interesting that you know we're we're starting to build out more and more capabilities that uh we hadn't had before. Uh I think it's uh exceptionally cool that I guess the world that we're living in with AI as we started the podcast was agentic uh in name only maybe and now we're continuing continuing to see improvements from our own product where you

know when you have little bit by little bit right adding to the autonomous nature of this it shows you more and more we're going to get to AI agent capabilities I think not only in terms of people like me being able to build out like real nice cool AI agents but I mean this is like very accessible stuff nowadays I feel like when I when you know we started working on this it wasn't as accessible but now we're getting to the point where it's like wow this is uh this was pretty dang this pretty dang accessible now. I mean we are actually getting some real um uh real work from AI now like >> so there's this whole discussion about either AGI is coming or AI is a bubble but I think just the truth is somewhere in the middle the truth is AI is

improving all the time and uh and now all these like different products like Jot form NA10 all these different um sales products uh or new startups are using AI to build real products that help real people do real work. So, it's not the AGI or the AI bubble. It's actually the real work that's important. And these tools, like these workflow builders are amazing if you're just trying to automate the everyday tasks that are like, you know, burning your day and like wasting so much of your time that you can just focus on like uh things that really matter to you as opposed to uh all the repetitive stuff. So I think this is really exciting. Um and just it takes time like like people are expecting everybody to start using AI and every every tool to become AI but it's like it takes so much

time to like build all these products and like you know uh make them work it's really hard to make them work like just uh you can slap something together but like to make them work like there's really big uh work behind and because AI is not magic AI is like uh it's amazing but it's not magic. So you still need to get it to work and I think it's just uh all these products that are making AI work. Uh you know uh it's it's amazing. >> Absolutely. No, I think that you you're pointing out a a really key thing here which and that this kind of came up in a lot of the conversations we had with people about the MCP type stuff. [snorts] Um, it's not a magic bullet, you know, like it yet. Um, and I think that the reason because of that

is cuz it takes a lot of coding, takes a lot of effort, takes a lot of time to make things how we'd want to make them, right? Um, and I, you know, all due respect to to people who are like, "Oh, well, I think we should just have it work." It's like, "Well, we have the capabilities. We just need to apply the tech better." It's like, well, I mean, you know, practically speaking, it's it's getting there, right? It's just you have to be patient in this or else uh I mean, it's it's not going to really work out in my opinion that amazing if you're just like overindexing always on like rushing to to make everything perfectly handsoff and and whatnot. You have to do the most with the tools that you have. And what I really like is that this it seems like NADN

and obviously us at Jaw Former are trying to make the tools that we have more accessible rather than pitching it as um a magic bullet. though. >> Yeah. I mean, um, we already had workflow builder for years, but the thing is like it's still hard to use. Um, because like you have to figure out um and and when you only have the prompt, uh, it's just people just write whatever comes to them, right? They magic part. uh is that like you don't and you can always uh like you know if you write a prompt that doesn't work just change your prompt and redo it and uh like just uh it's just so flexible. So it's it's it's not rigid like the old days like the software was rigid in the old days. To use software you had to learn it to use software you had

to like follow its like structure but with AI it's you know you can make it work as you as you want it. I don't think we are still there. We are still like all these products are just coming out with these features and improving them. But in a few years, I can see u you know products being so easy to use because of AI. >> Back in my day, I used to have to download all my software. No. Um seriously though, back in back in back in your day, >> back in my day, we had to like we would get this floppy disc and put put it in the computer to actually run the software, >> which is wild to me that it's changed so much so quickly. um you know so kudos to kudos to all the good work we're doing at Jaw Forman

that we're doing everywhere else. So um I guess you know with that being said is there any other comments uh you got you have before we kind of close this episode out. >> I think it's really exciting times. Um yeah definitely uh this is a big improvement nit uh agent builder uh they did a good job and um you know we are going to try to catch up with them with our workflow builder as well. Uh but it's all good. Yeah. >> Cool. All right. Well, with that being said, thank you so much for uh spending time here with us today. We appreciate each and every one of you and we'll see you in the next one. Peace. >> See you the next time. Bye-bye.