The Moment Automation Changed My Business Forever [#10 Oskar Kramarz]

configuration of AI agent is I would say 10 20% the other like 80 90% is how you configure the tools Kmarts is a automation developer known for his work in building apps APIs and business process automations not only has he helped multiple businesses thrive the use of AI he also teaches hundreds of thousands of people through his YouTube channel today he's revealing his groundbreaking process that will change the way that you approach your automation >> there are a lot of things that can go

wrong agent knows when to use this tool So basically what I presented is workflow that performs autonomous scraping. Agent was able to navigate for example through terms of service, privacy policy, contact us pages and so on to basically retrieve the profile of the company, physical addresses, email addresses and so on so forth. Those are like a very specific problems that probably every business has. >> You now have a way you can simply just ask the AI and then it reports back to

you exactly what to use in what use case. the world is full of those use cases that are just waiting to be automated. >> How do you bridge the gap with that? Do you guys have like a kickoff call? Do you like set up systems or what's your typical pattern for interacting with other developers a part of a team that may not have innate incompetence? Oscar, what's up buddy? Welcome to the podcast. >> Hi Dylan, good to see you. >> We got connected at the We Are Developers Conference. You came down and

you did a presentation that had some amazing workflow use cases. Can you talk to me a little bit about what you presented at the We Are Developers Conference? Basically what I presented is workflow that performs autonomous scraping. So basically what I did is I've created a puppeteer script. Puppeteer is like a framework for like headless browser where you can like programmatically create your own flows like to click this click that extract this information. So basically what I

did is I've created the script that yeah retrieves the text from the page like a full content and the array of URLs on this page. So basically I put it back to the agent. So I connected this as a tool and agent was able to autonomously select the website to actually retrieve information and in this case in we are developers it was agent for retrieving contact details. So basically our agent was able to navigate for example through terms of service, privacy policy,

contact us, sub pages and so on to basically retrieve the profile of the company, physical addresses, email addresses and so on so forth. >> So yeah, Publisher allows you to essentially have the AI be able to open up a web browser, go to a certain web page, look at the contact details of the user, be able to go in and say, "Okay, this is the email from the website. this is what they do, what they're about, and then you can feed that into some sort of legent system. Is that correct?

>> Yeah. Basically, Puppeteer in this case was actually responsible only for retrieving text and links. That's all on on the side of Puppeteer. So, yeah, the rest was made by NA10 workflow and agent. >> And for this, can you talk to me about the maybe the types of users that would use something like this, a system? >> All the people that are very into lead genen. So basically imagine you have I would say spreadsheet or a spreadsheet of hundreds of websites that you want to

simply scrape. What you do in this case is you simply create like a web hook with which connects you to this workflow and yeah you extract this data right so in this case it's like super simple and it's matter of like seconds to extract data from the page right so yeah lead genen people but the best part of this workflow is that in case of we are developers we extracted contact details but you can adjust it since you have access to the content of the page and

simply scrape. What you do in this case is you simply create like a web hook with which connects you to this workflow and yeah you extract this data right so in this case it's like super simple and it's matter of like seconds to extract data from the page right so yeah lead genen people but the best part of this workflow is that in case of we are developers we extracted contact details but you can adjust it since you have access to the content of the page and

links to retrieve other information so Let's say you are very into s applications, right? And you want to monitor pricing of those apps. So what you can do is basically adjust your agent to go through pricing pages of those apps and yeah store it into structured form in like spreadsheet database or whatever. How would you imagine a user being able to take this to the next step? Would they be doing like DM reachouts? Would they do email messages? So imagine there's a SAS

product and that they have a certain pricing model. You say, "Hey, I can help you increase your conversions with SAS." And you do that via cold email. How would N take this to the next level in order to take those leads and then turn those leads that you generated into maybe booked meetings or something? >> There are like a lot of ways to do this. Like the one you mentioned, this is like cold mailing, right? If you like want to create like a campaign, definitely it's

possible with an ATM. What you basically need to do is like just connect the SMTP server or something like this to like send the campaigns out, right? There are a lot of other options. The one of the crazy one that I came up with is for example send like traditional mail since you have like access to the physical addresses. Why not to use AI to create some kind of letters or something like this? Definitely maybe not like I would say a most modern way to do the

marketing, but definitely doable. I've actually heard of those services and it's like robo handwriting that you can like hook up to an automations. It'll actually send a hand a robo handwritten mail out to you. Okay, that's cool. I could see it because you're getting through this. There's a lot of different busy channels. I know email is a really busy channel, right? And so maybe this kind of a lumpy mail could be sent out. That's that's really neat. As a it seems

like a lot of these things that you focus on are real practical business use cases for business owners or other people that are in business. This lead genen scraping gathering and reach out is one system. Is there other what are the other popular workflows or automations that business owners might be interested in? I can say from my own p perspective because since I don't run like workflows which is my company is like one part of my activity the other one is familyowned company that is

completely not software related and it's actually the way I how I started doing workflows and what I do in this company actually and I automate a lot of stuff so I for example created like a workflow for I would say a currency calculation which was like a super interesting story. This is actually how I started with NA10 because we are using in this like non-software company familyowned business CRM system which basically is like super closed and we ask the

developers to create like a super simple function to do multiplication to do the ex currency exchange the calculation and yeah they did they gave us like a super high price for this which we were not willing to pay. So I decided to use N10 at that time to create like a own bot for this. Yeah, this is one of the use cases. >> What notes so funny about that too is cuz I just went out to Berlin for the we are developers conference and I track my expenses and working with init and one

of the things that was I was like oh man these are in euros and I only do them in US dollars and I'm like now I need a now I need a way to be able to track those and convert those over. So I could actually see that workflow being super useful even just for myself. Oh yeah, for this thing that you mentioned actually I've also created a workflow. So there is a super cool open-source software that is called actual budget and it's it can be used for example for

home budgeting right so you like record your expenses and so on. So what I did some time ago is like for my more personal use case it's but it can be also used definitely for business too is you do basically the photo of their receipt right and it's being parsed by llama parsed and all the data is being extracted and put into directly to this system. So yeah also super cool cool use case both for a personal I would say budget and yeah for the business ones

too. It's funny because you talked about how you got into inadin. You said there was a service out there that was really expensive and you're like I don't know if I want to pay that. What have you seen as like the common ways that people have gotten into in what's been that entry point? >> Price is definitely one one of those factors. This is how I got into NAN because I've actually started like automating things a few years ago. I think it was close around like Corona

too. It's funny because you talked about how you got into inadin. You said there was a service out there that was really expensive and you're like I don't know if I want to pay that. What have you seen as like the common ways that people have gotten into in what's been that entry point? >> Price is definitely one one of those factors. This is how I got into NAN because I've actually started like automating things a few years ago. I think it was close around like Corona

times something like this. And first I started like with Zapier which I didn't like then make which was like a short story for me at the time it was called integral as far as I remember and yeah eventually I stick with NA10 and for me it was like crazy that I don't need to pay actually such high amount of money for each I would say execution for each item that is being processed by the software. So this is like a number one thing. The second one is definitely flexibility and

possibility to do the things in the user interface. So basically what I do a lot of my clients do is basically we create the workflow as developers on the advanced level but then we hand it over to not especially I would say technical people that can still monitor it and like eventually do the very I would say basic bug fixing on their own right when it's something like I would say more difficult comes into game then developers is definitely developer is definitely necessary

>> so what you're saying is that like in terms of services one of the things you do is you set maybe a complex workflow or something that solves a problem and then you make some sort of front-end interface that allows the user or the company to be able to step in and whether it's execute the workflows or see the output of them or something like that. Is that correct? >> Yes. Yes. This is like the number one thing. So this is when we have clients when I have clients that want to make

the workflow accessible also for nontechnical people. The other part is when I cooperate with internal developers team. So this is a bit more I would say easier for me because I know there's always someone within the company that can like m perform maintenance or something like this. Yeah, it seems to me I don't know if you experienced whenever you work with a new company is there's always like someone on the team that is has some sort of developer role or is interested in it

but they don't have a lot of competence within it in you have to bridge the gap between what they know and what you're doing. So for example maybe there's a Salesforce integrator right but they don't really know a lot about init. So you have to have this handoff between what you do and then what they do like how do you bridge the gap with that? Do you guys have like a kickoff call? Do you like set up systems or what's your typical pattern for interacting with

other say developers a part of a team that may not have innate incompetence? >> So basically it works often both ways I would say they need to learn something about N10 and I need to learn something about their systems. So yeah, especially when people are quite cooperative, it's quite easy actually to do because yeah, nowadays we have a lot of content when it comes to any. So I can always redirect for example to official tutorials to show them how it works or

create like schedule like a quick call where we jump into this and I show the basics. It's like super easy to do on my part. Of course there is like often going through documentation of their like CRM system like the applications they are doing and so on to actually we have so we can be on the same page and start the works. >> Got it. So there's a little bit of education, but because there's so much documentation out there, whether it's YouTube videos or straight in it in

documentation, you can say, "Hey, this is what init. Here's what it looks like. You can watch a couple of these videos. I can walk you through some of these things and let me map out a bit of your system so that I understand exactly what you guys are doing and then we bridge the gap between what your current systems are to what n can do and then how to interact with it." >> Absolutely. It actually works the way you said it. And when you talk about these dashboards or these front-end

interfaces, I know that you primarily a lot of things that you focus on is more advanced use cases with the init and then making these dashboards accessible to your clients. Can you talk to me about maybe a couple of use cases that you see businesses using this system for? >> I can give you like a super cool example from my family owned business, right? where I used NA10 as a back end and created like a front end application like front end dashboard user interface

for this. So yeah, this is like a super niche business. So have you ever visited like a brewery for example or Yeah. >> Yeah. So there are like a lot of tongs there, right? So where the beer is stored and they have those like doors that are used for inspection, right? And cleaning and so on. So basically our family own business our family business sells this kind of stuff like those doors right so not are mounted into tanks and for each of those doors you need to actually add the ceiling right

for this. So yeah, this is like a super niche business. So have you ever visited like a brewery for example or Yeah. >> Yeah. So there are like a lot of tongs there, right? So where the beer is stored and they have those like doors that are used for inspection, right? And cleaning and so on. So basically our family own business our family business sells this kind of stuff like those doors right so not are mounted into tanks and for each of those doors you need to actually add the ceiling right

for this door and they are from various materials. So we had a problem because a lot of people have been asking us about those materials what kind of material they should use for the specific chemical because not only for beer right it can be used for water for I don't know any other chemicals like acetone and so on and on the other hand we have the database of table resistance right for those materials so this is like a very specific use case where I used NA10

because I've created like a workflow where I seeded the database with the information that we had about the resistance for those ceilings and I've created an AI agent basically that is retrieving the specific material according to the use case that a user is describing and it worked well for me as a internal use case I would say as internal chatbot because I was able to connect it to telegram for example and say hey what kind of material for this and that this kind of chemical is like

suitable in this temperature and it worked very well but we decided in company to make a step further and what I created is basically a simple user interface index.js where I connected it to workflow. So basically user has like a very pretty form where they can insert the chemical temperature and so forth. So basically like a very strict data that we use to actually generate what kind of ceiling is the best for this use case, right? Yeah. >> This is one of the examples.

>> So essentially what you said is that you don't always know what materials will match up to what you can use for these for your business. And so what I loved about this is you started with a problem like we had this problem. I'm not always sure what exactly materials match up to the doors that I need to have and what needs to be put together. So you essentially created a chatbot that gives you easy to find information. And so you don't need to go through and look

through all the materials and cross reference them and see okay these are what lines up. You now have a way you could simply just ask the AI and then it reports back to you exactly what to use in what use case. >> Yes. And the very cool thing about this is of course it's using our internal database as mentioned but AI also is what I've noticed is good when it comes to applying the specific materials for the use case. So when we don't have something specific the specific use case

in our database AI will recommend from own knowledge I would say. So this is also cool for us because we have actually extended our knowledge here right. >> Yeah. So you're using the I would say the internal knowledge base of the company and then you're extending on that knowledge base with AI to go outwards and say okay this isn't actually fitting into the system but we do know from the world's knowledge or the AI's world's knowledge that here's a couple of materials we would recommend

for this specific use case. >> That's awesome. >> Yes. Exactly. >> You mentioned also with I think it was like NexJS might have been what you're talking about. Did I don't know if you ever use I know Figma just has an MCP right now I believe that you can hook up to they have an MCP so you can use Nadin and the Figma MCP to actually create websites. I don't know if you're familiar with that at all. >> No, I haven't heard about this. Definitely you need to drop me a link after our call here.

>> I just found out the technology is moving so fast and so when someone else comes out with it I have a friend who's an N developer and he was showing me did you know Figma can do this? So he was building and generating websites with Figma and the MCP with NAD and I thought that was really neat. So I don't know if you knew that. Do you do anything with MCPS at all? >> Currently I'm working with one client that is using like implementing MCPS like heavily. So basically I cannot tell

tell much about it but what we do here basically is we create also a kind of dashboard for agents where most of the agents are actually connected to NA10 and what we also do is we have I would say the custom coded engine which we connect directly to MCP servers in NA10. So basically it's super cool especially in terms of like delegation of specific I would say MCPS. So this user for example from employee from marketing has access to this specific MCPS and those

can be connected. This is like a super cool because we have a custom dashboard custom like in gine for for AI connections I would say custom framework and on the other hand we have MCP servers. So basically user users can connect own credentials like super fast and yeah it works really well. >> So like you have job specific or domain specific MCPS that certain individuals get assigned to. So for marketing, there might be MCPs for maybe lead generation or creating some sort of social media

content versus operational or operation person who might have access to maybe some sort of MCP that hooks up to whatever plaid that you can get financial records and so you make them domain specific and then they can have access to the tools that they need to move faster. >> Yes, absolutely. And the also cool thing is that for example users can connect their own calendars for this. So yeah, it's also like interacting that way. So both I would say the job to that needs

to be done and yeah the I would say personal data from like those calendars in books and so on. So yeah super cool. >> Love that. Speaking of MCPS, I mean there's a lot of AI tools and features that are coming out very rapidly with it and we know it's we move very quickly. What are the features that you've seen come out whether recently that are really helpful or really useful that have come out recently from Naden? I haven't tested yet unfortunately but I

suppose it can be super useful of course agent as tools so we can easily on one canvas create like multi- aents I think this is something a lot of people have been like waiting for so I definitely need to give it a try mcp is definitely useful cool cool feature and something that I appreciate probably the most from recent updates I've also underlined it one in one of my LinkedIn post is evaluations so this is like a lutely I think one of the most important

features in like recent month like in in the change lock of an A10. >> Let's look at these one by one though. You talked about these multi- aent features. So previously what one of the things you had to do in order to create a multi- aent was you had to create an in workflow that had an AI agent and then one of the tools had to be accessing another workflow and inside that workflow you had a you had an agent and then the output would be sending that back up to the original workflow.

Now, with this multi- aent feature, you have the ability to create multiple agents inside of one workflow so that you can see everything at once and they can all interact together. And do you have any ideas around what you want to do with a system or workflow like that? >> Not yet probably or maybe other way. I have a ton of ideas because I have whole lists of the agents that I want to build and a lot of like products that I want to build and so on and so forth. So, I

have all list for that. The problem is I don't know which one to start for this. Currently I'm like at this specific moment I'm focused on e-commerce use case in this family owned business which doesn't rely heavily on AI agents at the moment. But yeah I think it's a matter of few weeks where we are going to implement some of those features too. >> Yeah I know one thing I like about the multi- aents is that sometimes it's hard to keep everything in your head when

you're building out in workflows. there's sometimes so many working part parts that if you have all these sub agents and things like that, you're trying to track what's everything happening. So if you can see in one bird's eye view inside the workflow like all of the agents moving together and you can see them interacting, you don't need to hop around and check out the executions of multiple workflows. You can see everything operate all at once, which I know is a really cool ability to

debug and track things. So, one of my favorite things about Nitin is the ability to see the executions happening in real time and what's being processed because sometimes with code, you don't know what's going on. You have to see what's you have to have some sort of like trace back to what's going on. I love that new feature coming out. I haven't played with it yet, but I just did a I just did a meetup last week where I saw a gentleman show that to me and I was like, "Oh, that's right. I

need to get on that. I haven't done it yet, but it looks super useful." to be honest like multi-agents are super important because of the reason that you mentioned so debugging right when it comes to AI agents they have sometimes then can throw a lot of errors right because they are not deterministic so that's why I mentioned like evaluations are like super important part for me to make workflows actually good and production ready right this is like a super important step for me and yeah

having like now when you build the multi- aents and having them on one canvas where you can track it in one spot and you don't need to click like through tabs to go to separate workflows to see what actually happened and errors when it comes to AI agents on development states happen like very frequently a lot so definitely it helps it helps to keep the flow what do you think in terms of you talk about being production ready because there's it's easy I won't say easy but a lot of

people make fun multi- aent systems that aren't production Right? They're cool concepts. They neat little use cases, but they don't actually cross the chasm from neat little use case to actually being something that could work at scale or become production ready. Do you have a process or a system for bringing these multi- aent or any types of systems into becoming production ready? >> To be honest, at this moment, I don't have production ready multi- aent system. I haven't implemented such a

thing yet. Of course for me personally I did such such a things and I have like a testing phase in company like one or two probably I wouldn't call them at this point production ready when it comes to the agentic workflows where I have like several agents one after another then definitely I have created such workflows and for me it's always like a matter of like time spent on sometimes it can be frustrating because you spend so much time on this one single node which is AI

thing yet. Of course for me personally I did such such a things and I have like a testing phase in company like one or two probably I wouldn't call them at this point production ready when it comes to the agentic workflows where I have like several agents one after another then definitely I have created such workflows and for me it's always like a matter of like time spent on sometimes it can be frustrating because you spend so much time on this one single node which is AI

agent and you like tweak the prompt and what I personally believe is what is super important like a configuration of AI agent but this is my personal opinion configuration of AI agent is I would say 10 20% of the whole work the other like 80 90% is how you configure the tools and this is like super important part because when you especially when you have a very custom tool for your I don't know CRM or something like this where you do even like a theoretically simple

HTTP request There are a lot of things that can go wrong there. So you definitely want to configure it properly. So actually agent knows when to use this tool, right? So I put a lot of effort to configure the tools properly and I definitely then spend also a lot of time for the for example for system prompt and so on to and run evaluations to see what works what doesn't whether I should use the large larger model or I can go for example with GPT menu yeah

>> got it so you're saying like setting up the general workflow we'll say architecture right so you have an agent they have tools like for example Slack, Gmail, Google Calendar, Salesforce CRM or whatnot. That's just the initial 20% get going. But when you talk about configuring the tools, what you're talking about is like the Slack, like maybe making a custom HTT request for a tool or putting the prompts in place. When you're talking about getting these configuring right where you spend 80% of

your time making sure that it is actually operating the way that you want, what is your process for making sure or any trip ups or mistakes or common patterns that you've seen that happen that makes these tools not function the way that you want them? I I think the number one mistake is like of course it depends on the agent but limiting myself when it comes to the queries that can come in into the agent because even when I run like evaluations right I have for example the queries

that I have and the response that I want to like have from the agent right for those specific queries but when you have created those agents and you created those evaluations like perfectly right you eventually want to give it to the end user and you are it's completely unexpectable what end user can actually put in into the agent, right? So what I do on a regular basis is I try to cooperate with the end users or at least with people who know how the agent should behave and what queries can go

into the agent to see what can go wrong. It's definitely not possible in many cases. I wouldn't say it's impossible, but in many cases it's not possible to cover like all edge cases, but we can definitely do our best to make our agent high quality. Yeah. >> Yeah. Let's run through this a little bit because I think this is actually important. Step one, right, you figure out the use case, the problem you're trying to solve, right? Another step is you figure out, okay, what are the

systems or tools that they're currently using like Slack, Gmail, or whatnot. You then build out the workflow with, let's just say, a single agent in place. you make sure that it is configured properly, right? So you have the right prompts in place for the use case, right? You have the tools configured properly for each of them to make sure that the agent knows how to call those tools properly. And then another thing you're talking about is making sure that the queries, whatever the end user is

going to be using, what they're being put in place, they're actually using the queries properly. So they know how to request the agent, the AI agent to be able to make certain requests, whether they're pulling all the emails out or adding things to the Google calendar. and you say, "Okay, these are the types of queries and the types of prompting you can do for the agent." And then it sounds like that is the general flow. And then on the back side might be an AI evaluation system or some sort of

pattern to make sure that the not only is the human asking the right types of queries, but the with the AI evaluation, they're actually the agent is responding in the way that is expected to give the user the the results that they want. Is am I getting this correctly or am I missing something? Yeah, this is actually the flow that I'm using. It may be like sounds quite general especially for people who are more into the coding coding sphere because basically what we

do here is like a testing, right? So we give the user instructions on how to use the agent, right? And yeah, we want to cover as many edge cases as possible. Of course, I would say reasonable borders, right? So yeah, that's the flow. And when you're talking about giving to the edge cases, it sounds like it's updating the system prompt or maybe giving a few shot prompts for the agent in and of itself. So you can see these are the types of requests that we want to have

someone from the marketing team be able to say. And then we also say, "Ooh, Mr. Marketing sir, this is the types of queries you make." And then with the system prompt, you're like, "Oh, here's some edge cases that we want to take." Do you do you then train the system prompt or do you give uh fshot prompt examples inside of the agent in of itself or is there anything else that you do to make sure that they speak well to each other? >> Of course, this is my approach but I'm

not sure if it's like the best but what I want to do always is to have a clear path on what went wrong, right? For for specific for example execution, right? So since AI agent are not deterministic for edge cases or for example for errors I want to have a strictly deterministic path on what went wrong right in N10 is super cool because I can for example for errors right I can for create like a specific path like error path from AI agent where I don't know I can call it

like I don't know send message to my telegram that something went wrong in the workflow right or I can put this specific workflow execution to database and examine it later. So for this kind of cases, I want to collect them as much as possible and then analyze what actually went wrong. So I can go through history of the agent and see how the conversation was went whether it was for example I don't know sometimes we have this I don't know open AI outage right

and it was like the thing on the model side or whether the query was not clear enough for the agent. So yeah those are the things I I prefer to do. So basically collect those errored or or not necessarily errored but also cases where agent didn't answer like correctly where couldn't answer right. So I can after the agent I can do a simple chain to recognize whether actually agent provided an answer or not in case of answer has not been provided then I also can collect this. Right.

>> Got it. Yeah. So what you're saying is that there's you can always have that inside the settings of the agent or the output say always error on output which creates another path or another route inside of the NA in workflow and on that new path not only could you fade in what the error status is you can do a real-time feedback to notify you that oh my god open AAI's outage happened yet again and it's on them or for some reason there's some sort of issue inside

of NAN that you need to be aware of so that you can one respond immediately because there's nothing more frustrating than having an automation go out on real live business use cases, but then also you can maybe add that into some sort of spreadsheet or database that lets you to be able to track all of these issues to see is this a one-off that happens one out of a thousand times or is this something that's consistently happening over time that might take a priority status for you to fix?

>> Yes, exactly 100%. >> I love it. I love it. When people want to get started, businesses want to get started in it in start automating some of their processes, what do you see as ways for them to get started? what are the patterns for them to get up and running. How do they figure out what's their first step from going I don't have any kind of automation in my business to really fully getting up and automating so that they can save a bunch of time. >> So first of all the problem that I see

is what I should automate. So basically I I was at the same place. So what I've decided to do is to start identifying the potential problems in my company as mentioned for example with those ceilings and the apps for ceilings or currency exchange integration something like this. Those are like a very specific problems that probably every business has. So if not business then you personally also have some kind of things that you can automate. For example those parsing receipts that I've

mentioned and putting it to actual budget. I think the world is full of those kind of use cases that are just waiting to be automated. So identifying them is like a number one thing. >> One thing first any business owner can do is start to list out problems. Okay, what are the things in my business that cause the most time, the most friction, the most manual labor? What are the things that I'm maybe not paying attention to like responding my leads in less than 3 minutes or whatever it is,

but just make a list of the problems of your business and go, okay, which one of these would give me maximum value? And then maybe cross reference that list of here's my top problems to here's what's easy to be automated and then we can create that nexus point between those two points to be able to be automated. Is that correct? >> Yeah, definitely. We can do this that way. >> You were going to say something else. Please continue on the next point. Yeah. And then what I would simply do is when

I haven't any contact with any of them before definitely go into like tutorials for example those official or to YouTube creators for example like my channel and yeah just learn as much as possible and I would say a lot of people also are trying to delegate creating workflows to to external tools like asking chat GBG to create workflow for them. I think the better way is to learn by yourself and try to connect those notes, those pieces together. It always brings so much

satisfaction and I think you can learn a lot more along the way. Yeah, the prompt text to workflow. It's it's great for kind of zero to oh my god this feels amazing but it might you need to have some sort of at least semi-technical skills to be able to say know what's going on because if you do that out the gate if it breaks or if it just doesn't do what you want you won't actually know the core foundational skill sets of how to solve any of these little bugs because in the world of whether it's

automation or writing code one comma can break the whole system and it can be really demoralizing over time and so I totally 100% agree with that. And so getting up and running, figure out the business use cases, go through the process of solving those one thing at a time. One of the ways that you talked about this was maybe going and looking at YouTube tutorials as one piece for seeing, okay, has someone else already solved this for your own channel. What have you seen as

videos that have performed very well? or what are the ones that you see as the trending uh video content on your channel that people are really uh find a lot of value from watching? >> So for just for the record at the moment as we are recording this I haven't uploaded any video for a few months now. I'm a bit retired but hopefully not for so long but what I see in the statistics actually is top trending video of mine is the web hooks videos. This is like a super basic tutorial on how to use web

hooks. So what it basically shows is we have a lot of new users coming into NA10 that are looking for like a basic information on how to create workflows and how to connect them with external apps. So this is most probably why they are watching this video so heavily. Yeah, that makes a ton of sense because part of the automation is essentially when you talk about automation, it's merely moving data across different applications and being able to transform and manipulate that data in some sort of

hooks. So what it basically shows is we have a lot of new users coming into NA10 that are looking for like a basic information on how to create workflows and how to connect them with external apps. So this is most probably why they are watching this video so heavily. Yeah, that makes a ton of sense because part of the automation is essentially when you talk about automation, it's merely moving data across different applications and being able to transform and manipulate that data in some sort of

way. And HTTP requests and web hooks are the ways that you can send data out of inad if you want to call that to external applications whether that's a say a type form or any of these other applications that are out there. So you can send data out with HTTP requests. You can bring data in with web hooks and then with that then you can take it and manipulate it through using inadm. >> Exactly. >> Love it. What advice would you give to anybody that has seen a lot of these

way. And HTTP requests and web hooks are the ways that you can send data out of inad if you want to call that to external applications whether that's a say a type form or any of these other applications that are out there. So you can send data out with HTTP requests. You can bring data in with web hooks and then with that then you can take it and manipulate it through using inadm. >> Exactly. >> Love it. What advice would you give to anybody that has seen a lot of these

videos on YouTube and they want to get started and they're just trying to figure out, okay, what do I do? >> Experiment. Experiment heavily. This is what I do on a daily basis and don't limit myself. And I recommend also to our listeners to our viewers not to limit themselves to basically try to push the boundaries and build try to build things with an A10. So I personally at this point I can build probably like I I've built a lot of workflows and I can I think I can build

a lot still but just I don't know two years ago I wasn't at that level of advance and creating like the workflows that I create on a daily basis was like a struggle for me and just because I did this on a regular basis and try I was trying new things I was curious about those things I was able to simply elevate ate and yeah learn a lot of stuff along the way and this is actually that was part of my workflow because I've tried to do something I wasn't able to find a easy way to do this and I was

trying to find a way to solve this problem and then I was recording a video about this that was for example for PDF generation I was not able to find a service that I could self-host by myself to generate PDF documents and then I found Goenberg. Goenberg is basically like a service that you can use to generate PDF documents or from HT HTML like a source code, right? So you can generate like invoices or something like this and I was struggling with this. I had this problem. I solved this and

eventually I recorded video about this and I see a lot of people are appreciating it. So yeah, this is my workflow and I recommend it to you too. >> I love it. Yeah, exploring, experimenting, playing with inn, seeing what's possible and you can level up really quickly. You talked about two years from being struggling to given a high level of competency within. So, what's cool about it is that you can get up and going pretty quickly with Naden and start to really get some skills that

you can then serve you for any of these problems that come up in the future. And then what's really cool and I think about content creators too is that you have a personal pain or a problem. You're like, man, how do I do this? I got this problem. And then you solve it and then you naturally want to share it with the world and like oh my god anybody else who has this problem I just want to tell you do this Gutenberg or whatever the things might be this if you

have this problem just watch this video and which I think is one of the great things about content creators that are really educators helping people solve the painful problem. They're like it was my pain please nobody else have the same pain. Yeah, exactly. And this is actually the power of the community, I think. And I also recommend to connect with other community members whether there are like whether there is like official NAN community or private communities within

N10 that popped out recently. A lot of them just connect with those people and I think there is like nothing better when eventually you meet in real life. Like for example, like we did in Nathan Fest which was super cool. People that just a few days ago were only avatar for you. You meet them personally and you can talk. Yeah. Nothing feels better than this. >> It's so fun. Yeah. You have this can I completely agree. It's the power of automation, the power of community

coming together and then you meet up and you're like, "Hey man, I'm working on this thing. How do you get this to do that?" And then we're all kind of like working on problems together and solving each other's issues. You're like, "Oh, I've already solved that." You use this Gutenberg thing or whatever the things might be. Because that is really the power of the same thing like the automation platform brings all of these different applications together but the community brings all these people

together and allows us to group solve so that we can accelerate our knowledge, learn faster, learn smarter, learn better, which I absolutely love. >> Absolutely. Absolutely 100%. >> Is there anything else you'd like to let people know about before you tell them how to get a hold of you? Is there anything else you'd like to share? Anything else you'd like to let people know about? Any messages you want to give to people before you can tell them how they can find you?

>> Just build stuff. This is like something that that I would say this is like my life domain building things and yeah I recommend it to you to build stuff and and share it with people share what you are working on and you will see how many good you will you will have in return. >> Cool. And if people want to find your YouTube channel or get a hold of you how do they do that? >> I have a YouTube channel called workflows. So basically when you type in Oscar nan Oscar with K in YouTube you

definitely see one of my videos. I'm also on LinkedIn which is Oscar Kamash, Oscar Kra, you can also pronounce it. So yeah, those are like two channels where I'm probably the most active. >> Oscar, it's been an honor and pleasure my friend. Thank you so much for sharing your love, your passion, your knowledge with init and I will see you on the other side, my friend. >> Oh yeah, thank you very much for hosting me and yeah, have a great time. >> All right, take care now. Bye. >> Cheers.