Build AI Agents Without Coding! (DataHack Summit Session)

Key Takeaways

Hello everyone and welcome back to my YouTube channel. So today I have an exciting news to share with all of you. Recently I was invited to be a speaker at analytics data hack summit where I got to meet many of you as well as the founder of autogen and Josh Tama who runs the channel stat quest. Now there I was very very happy to connect with many people with the like-minded overall approach of solving problem and how learn that how they're using AI in their everyday life to solve problem in an enterprise as well as at a consulting level. Now I took a very important talk there on how can we use these no code tools to automate everything which I believe will be very helpful for anyone who is let's say in a company or if you are someone who is even trying to start your own agency as well. Now I have been running my own consulting based on all of these learnings and in this video I have made sure uh that I'm sharing each and everything with you. Now I requested analytics video to provide me with the video so that all of my audience can get fully uh benefited from this and learn what all I have shared. So with that please see the video. I am pretty sure that there will be so much things to learn. I have kept the overall video where I have also answered the questions in the end. So that if many of you will be having something after seeing the full presentation you can surely uh just learn that okay what exactly was the answer for that and if you still have anything please save in the doubts and yeah do make sure that you like this video and share with all of your friends. I've also covered that how can you use it in a company so if any of your friend is there who is looking forward to just uh understand these tools or understand the full flow then I think it will be a very good and a perfect video to get started with. So with that everyone now let's jump onto the video. [Music] So yeah, as you can see on the screen and the same is pretty nicely visible building automation and agentic AI application with no code tools. So this is our topic for today and I've not seen the other presentation but I assure you that this is going to be a lot more practical then I just telling you few things because I have designed the slides as well as I've introduced many practical things here which you will see that how using these no code tools the automation can be helpful to you right out of the box. Okay so just a second I forgot the switcher. Great. So yeah, this is a big photo of me and this is me in real life. As you can just learn a little bit, I'm a graduate of NSIT. I have worked pre previously with companies like Goldman Sachs, OO Rooms, Mind Tickle. Now I have switched the gears and currently I'm working on my own startup as a CTO where I am creating a platform where you can you know many professionals and students they can come and upskill themselves in AI plus do real world projects. So it is still in stealth mode but hopefully within the next 2 three months we will be launching the same as well. A little bit more about myself. So hopefully many of you must have seen me on YouTube where I share my knowledge on using AI practically. Along with that I also do lots of AI consulting for different different companies where I tell them that how they can integrate and use AI into their day-to-day work. Right? So that is a quick brief about me and you will surely find me posting more and more content on YouTube and educating you via the live classes I take where from students to professor alike. I teach them that how AI is not that difficult and how can you genuinely use AI into your everyday use case. So let's move forward everyone and this is the agenda. So coming to the agenda we are going to start from the very basic understanding about automation. Then we will see that how no code tools can help you in integrating automation into some common use cases. Moving forward we will just have the elephant in the room or I think in the whole world today which is AI. We will see that how can that be integrated right? How can that be integrated into professional use cases not for just uh like funny cases where we are just trying to run a workflow. So I have prepared awesome practicals for the same. Then I will cover a very good no code tool NAN. Pretty sure that many of you must have heard about the same. If not I will show that as well that how it is very easy to understand and integrate AI into the same. Right? Then of course industry use cases. So we have practicals which I was saying. So I have covered two three use cases from different different industry and then I will shed a light on how can exactly we can adopt the same because knowledge is all fine. this we can learn about many things but this specific topic I have chosen so that everyone in this room he can he or she can actually understand how we will be adopting it into our day-to-day jobs or into our business and how AI is now not something which is very difficult right so that is all going to be the agenda and finally yes we will have a Q&A where you can ask any question for sure regarding the topic with me and I will be happy to answer that okay so great everyone with that let's now move forward. So next as you can see automation 101. So how many of you have an idea about automation or say are afraid of automation in the cloud? Can I have a quick hands up? Afraid of automation. Let's keep the question afraid of automation. Great. How many of you have an idea about automation? Like what exactly is an automation? Yeah. So a lot more which makes sense. So pretty good oneline definition is written. Automation is the process of using technology to perform task without human intervention. So as the name suggest automation means to automate something. Now it's not a new concept though many of us will be seeing the same across different different articles different different softwares and we might be thinking that well this is something new. uh the very first project which I took up in my professional experience I was automating something right when I was working at Goldman Sachs 80 to 90% of the use cases they were where we were automating something which took a lot of time to do manually right so automation has been there we have been able to do automation via no code tools like I will show you via writing code as well right so automation as we can see it is following a fixed rule-based process and of course it is good for repetitive and predictable able task where we don't want to spend a lot of our time. Coming to the example, a very good example can be when we send an email to a customer as soon as they sign up on a website. So many of you will be signing up on a website. Either you will get an OTP, right? No matter if it's 4:00 or let's say in the morning or let's say as soon as you sign up, you get a spam email or something. Then a very relatable example which everyone would surely relate to WhatsApp automation, right? So you just do something uh you place an order you get a message. So it's not that someone is sitting and sending you that message right. So automation is happening around us more than we realize it in real world everywhere. The point comes that how can we understand it in depth so as to use it as well. Going next uh this is the very easy flow of an automation like very dumb down. We have a input order trigger. So let's say it can be just you going and placing an order. Then there will be some action. it will see that okay what you have placed how much is the MRP what all you have to do and finally the output where you will be getting something right you get a message on your WhatsApp that's the final output so this is something as a very basic and understandable flow now coming to next I think this is the elephant in the world which I was talking about now let's talk about agentic AI right so agentic AI or AI what are all these things coming to agentic AI it is just an autonomous AI system, right? Which can act without our help. Okay? So, a very very important component of agentic AI is AI agents. Now, how many of you have heard about AI agents? I think most of us, okay, how many of you are afraid of AI agents? Like you don't know like it's very complex technology, something like that. Pretty good. Like, okay, I'm very happy. So, like very less people are there. So in a real sense and I teach that a lot and this is the best analogy I found. AI agent is just like a digital human. So just like I am giving a presentation here. Hopefully in 10 years if we are able to make an agent like a digital version of me he will give a presentation uh instead of me. So that is what we are planning to achieve with agent just that nothing else. So just like we human have a brain okay it will have an LLM. Just like we have a memory it can have a DB where it is storing everything. Just like I can act on my own right without based on my surroundings like hopefully if there is some natural calamity we all will run outside right I will be aware about my surroundings. Similarly that AI should also be aware about the surrounding the system and then take that action. So limited supervision of course will be there and a very good example is a travel agent which can not only suggest that what you should do. It can take action as well because if you remember back in 2022 when chad GPT was introduced yes it was very helpful but it was limited to giving you information. You if you let's say ask it to write a email for your manager for asking for a leave it used to give you that email but you have to copy paste that and we humans are lazy right? we still don't want to copy paste that as well. So we created a system in which we can connect it to tools and it can send the email for us as well. So all this AI agents came into picture because we are lazy like that's the real reason. So not just it is giving you the output it is also making sure that it can act on that. Going next a very quick comparison how automation is rule based how agents are datadriven right how automation we have the predefined process and agents can do adaptive learning then automation majorly we will see a sequential task happening so this then this then we can have some conditions as well agents can be smart they can take the complex reasoning if I am sending you a message where I'm very angry on your services agent will not just send a message for the next service he will try to understand and most probably it will send a message that hey what can we do to correct that right so this is the power or benefit which we get from having agents into our system and then of course limited decision-m autonomous action so this I think shall be very clear from the example we have taken now this is just an uh expanded workflow the one which I shared earlier so if you remember I shared uh okay back button is working yeah so this is a flow of an automation the same automation what happens when we introduce AI into that right so initially we just have again triggered event so it can be any trigger event the major thing happen in between where the action was getting taken now with AI into that you can see that how I've created three arrows in the right diagram where we have the memory tool and AI so AI is pretty simple we can have chat GPT gemini deepseek gro okay anyone we can use then we can have any tool so let's say if you want to send an email you will have to connect your Gmail to that right so it can take that step from view from your view and a very important thing memory so we can have a rag based system as well where it is taking some inspiration let's say from any company documents I will show the practical as well so that now it is a lot smarter and it can act on your act on its own on your behalf so earlier where for a smarter system you might need a person we have eliminated that person with the help of AI agents or this AI agentic flow so that's That's the whole idea. Pretty sure it should be clear. The clarification will be needed. But we will see the same in action as well. Now again real world example. You can see that how I've drawn two coffee machines. So the left one is a pretty basic one, right? It is where you will just go and create your own coffee. The right one it can be made a little smart. You can control that via Wi-Fi. You it can also make the coffee let's say based on the demand as well. So when you are adding that functionality of being smart into any machine that becomes a agentic system. Uh another example which I can give here everyone is of let's say a water sprinkler which we see in garden. One sprinkler can be which just waters the plant every day at 9:00 in the morning. Right? So it has been automated that every day at 9 it will just sprinkle water on all the plants. But what if there is a rain one day? Well, the second one if we make it smart then it will not do that because we have made it smarter and it will understand that yes already my plants they are getting all drenched in the water why should I add extra water into that right so these are one very good relatable example I always believe that relatable example are very helpful in understanding these complex things so with that now let's move forward let's talk about no code tools as the name suggest no code tools right we don't have to write any code right So they are just platform which can help you build any application without writing any single line of code. Now the reason I have chosen this specifically because no matter if you're in college where we are very active and very uh let's say very much open to learning and then if I compare same to let's say 10 or 15 years experience where we are little less towards learning new things right this no code tools they can be helpful to each and every one of us right and with AI taking over all the industries they can help you to integrate and apply the same in your work so we can just have a simple drag back end of interface where we have the pre-built components. Now features you can see we have the pre-built integration for databases APIs as well as services. So now if you want to again connect your Gmail right or let's say Google sheets you can just drag and drop that button or that overall flow of Gmail or Google sheets. Now there are some common one which you can see on the right NAN make relay zapier okay we will be focusing specifically on NAN but other one can also be used for doing all the work going next enabling automation without coding. So this good diagram which I have created this is kind of bridging the gap between a no coder and a coder right. So the major thing is that it is empowering non-developers. So if you are the owner of a problem right this is how it used to happens 10 years back some nontech person or let's say some problem owner or project owner he used to come up with a solution he used to communicate the same to the dev team or the IT team they used to code everything well right then we used to have multiple iterations and a thing which might take a week used to take a month right this is the biggest gap which these no code tools are helping you to overall just jump where Now any non-developer he can just build and it on the workflow. This surely enables the speed and the agility. So now you can the problem owner right if I understand the problem better why can't I just create that why should I have that communication in between because if we communicate something is going to get affected right and of course the last thing is there will be no syntax error if you're not writing code there will be no syntax error for sure. Uh just drag and drop components if you just handle the English pretty well you will be pretty fine with that. Right? So these are the enabling automation without coding. Now let's move forward and see the no code meets AI. So if I would have to explain agent like the way I explained right I think it's pretty easy that I easily created a agent via one of the no code tools the N10 only where I can send a message then you will see agent here it is having a chat model which can be again LLM which is Gemini open AAI all of these then we can have a memory. So as of now I can just have any DB for the memory and I can provide it tool as well. Tool can be any application right? If the same thing I have to do with coding this is via lang. Now pretty sure that if we compare this and this slide half the half the code would have went away if I would have started with this slide right I have to explain that okay what is all this libraries what is lang chain lang graph here also I'm creating a agent it is a react kind of an agent now okay doesn't matter what is react what is this we don't need to worry because this session is on no code tools right so this is agent because of us even a person who is in school he can understand the yes this is an agent I don't need to understand what is this complex syntax I cannot make that error So just drag and drop all is done. Now let's of course quickly go through the benefits of no code automation and AI faster development lower cost it will overall be very accessible to you right you don't need to communicate with the team you can have less engineering team as well so that is of course a very very high cost saved then we can have so many services just via drag and drop you don't need to understand right you don't need to understand if Gmail gives this service or not you don't need to go to its documentation just integrate that as a tool and it will do what you want to do right similarly we can focus on the logic not on syntax logic is where I think we should spend our time not on the syntax and of course how it is flexible and easily maintainable. Tomorrow when the use case changes easy just drag and drop tomorrow when let's say GPD 5 has came into picture if GPD 6 comes just drag and drop the GPD6 component right you don't need to understand anything that's the major benefit now let's talk quickly about no code beyond the AI so as I said no code tools are not something which are very new we have no code tools everywhere right pretty sure all of you must have heard about Udu you must have seen Vix then coming to data engineering we have Azour data factory We have Amazon AWS glue where we do all these things. Right. Next, let's talk about quickly why NAN is a very good choice for automation. Okay. So, NAN is an application where we will see the demo as well. It is open source and extensible, right? And they have not sponsored me. It's just one of the platform which I like. Then they're having a very powerful workflow engine. Okay. So, they can handle any complex logic. They're having integration with multiple services. So all the common services which we use like again let's say Google sheets, slack, WhatsApp, telegram all of these services are just there out of the box and it has around 400 plus integrations easily right then of course how it is having the best support for AI related things. So when you want to have this agent something which has been I guess the major point of this full year you can just drag and drop that AI agent component and you can just provide whatever you want to provide. So that's the whole idea and that is that is why we will be seeing the same via any tech. Now of course we can use any other tool as well. The major thing is that we should understand the use case as well as what exactly these agents and AI is. Okay. So moving forward let me show you now quickly via practical. So as you can see this is no code versus code comparison. So first I'm showing you a use case. I will quickly tell you what this use case is. On the left we have a user. He will be asking a query and we have three agents right? So three agents which I have created they will be connected to a company source. So they just have to answer the question. Let's say if I ask a question about hey what is the leaf policy? What is the thing related to laptop? If I have to get a new laptop how can I connect to the VPN. Now let me quickly show you the same via the code as well as via this do code so that it is very helpful. If I just quickly go back and show you. So this is actually one of the one more session which I am part of. This is the full code which I have written here. Right. So this full code if I just quickly show you I have used a framework again for AI CU AI many of you might have heard of it I've used that framework very complex code right so the keys are deactivated if anyone is thinking of copying that again this is the same thing this is the exact policy eta which I was talking about right so again very simple system we are creating companies has some policy leave policy IT related policy finance related policy we are getting a chatbot which can help you answer that thus making sure that HR or your manager is a lot free. So these are the policies now which I have created. After that if I just zoom it out we can see that how I've created a rack config how I've created multiple agents something for memory right then task etc. All these things I have created and made a very very big code so that I can do what I want to do right so this is all if we don't have these no code or AI present in these no code tools. Now let's see the magic and yeah this is what we have in a no code tools. Here you can see three agents just like I showed you. We have a chat trigger. Now you can just change it via Slack or something. The first agent is a classifier agent. The next one is a retriever agent which goes through your talks and everything. And the last one just tell you the AI agent which tells you that hey what finally should be the policy. How easily we have created that right? I hope all of you can see the difference. So classify agent it has a Gemini brain. It is having the structured output which tells if it's a HR query, IT query or finance query. Then we have the retriever agent which retrieves those documents, retrieves the rules, right? and tells you that okay what are the rules the last agent it just tells you that in a simple manner like let's say if you if I ask for the leave policy it will just tell me what is the leave policy right so I can show you the same in action as well pretty nicely let's see the same if I just quickly push on this and open chat I can ask what is the leave policy and when I press enter it's very visually nice like we can see that okay now how it is using the brain it will go through the memory and understand that what are the policies and hopefully I will get the answer in maybe a minute. So that's a quick thing but I hope all of you understand how easy is it to create right when applying a medical or parental leave please submit your request for supporting documentation via HR portal at least 2 weeks prior to leave date. The thing which I wanted to majorly point out here is that it was very easy right do a prototyping to support use cases like this we can just create something like this and you can use that as compared to the previous code which I showed that was easily a I think 800 plus lines of code doing the same thing here I have done the same I can explain to my uh let's say managers I can explain to my stakeholders that hey this is what I'm trying to create let's quickly do a prototyping We can do the prototyping in a week as compared to me writing that code first in a week if that is successful right and of course the understanding is needed I have created a multi- aent system in maybe a day easily right so this is a multi- aent system where we are having multiple agents they're having their brain they're able to understand what I'm asking and I can ask any follow-up question because they have the memory as well doing all this in a code though can be done it is not at all easy you have to understand frameworks in depth and then you have to create that this even a 14 15 year old schoolgoing child can also make once you understand what is this right so this is everyone the very first practical I hope you are able to understand that how we are able to do that just by drag and dropping now if I go here I can even have multiple applications right so these are all the different different icons of application which it supports right Slack Gmail sheets excel all of these things are very nicely supported along with that your cloud technologies like AWS, Azour, they are also supported very nicely. Right? So let's go back onto the presentation and quickly see some more use cases. Now people who are from tech hopefully will be thinking that mang that is all fine but we are from tech technology in IT we have some very good use cases we cannot use that. Well that is where this has taken a lot of leap this technology right. So the next use case is automating PR code reviews something which I have spent days on and not the best work which I used to like but you can create a system right which can do this code reviewing for you which can handle the PR which you create now let's say any change you make it can tell that hey these this code is wrong maybe you have to name the variable better we can have all these things as well via a simple again system right we can see the steps so it's a PR trigger it is now connected to GitHub Right. So natively it has the GitHub connection. I for once have also never used GitHub APIs. Right. There are many things which we have to do here. It's just again a drag and drop GitHub API. It will show you the PR from whatever repo you connect to. Right? For the people who are in nontech and never created a PR that is totally fine. This is just a very good system and makes it very easy for engineers. Right? Then it will also make sure that it reports everything into a Google sheet. It is able to see if we have already done that. Right? an agent is sitting in between which is reviewing your code and you can just again create this system very easily to do a first level of sanity check for your company right and once you do this and you actually use that you will see that the results are also pretty nice so I have done this and I have integrated that for one of my clients where they wanted to make sure it was a small company that at least they are not sending very bad code to the production right so this was easily able to handle at the you edge cases and tell them that how can you create that Right. So we can see the same. Now if I go next, this is just the traditional code versus the no code approach that how we are having a trigger like so these are just different different steps. So we can have a trigger. If we were doing it via traditional code approach, right? We will used to have some API or custom web hook. Here again everything is there in the visual node. Then we will have to fetch the difference. So me as a engineer I have to go and see that hey what are all the changes which are created here. the AI agent or this workflow it will just get all the changes for me then it will do the review best practice context post review label the PR and make sure that it is able to also add the same into the Google sheets right so this is again the process where I've compared it to traditional versus the no code approach right that traditionally how it was very difficult but with this workflow it is easy to understand as well as very easy to do right we have these steps which I have written the things properly that first is GitHub trigger then it will be getting the PR differences is then how it will be making sure that it understands what all we have given so AI will now going to take part it will update the Google sheet as well which I have connected right so this is again all using the no code the second practical uh showing this will take a little of time but yeah this works pretty well now let's see the third one and this is pretty nice so where we can search the trends in real life and create a newsletter right so content I think every company is creating the content so let's see this is in action I think many of us will like the And I will explain this much in depth as well. So uh let me go back and what I will be doing search and send newsletter. Right. I am using multiple things here. The trigger is something which I will just click and it will start. Right? Then I'm having the tavly. So it is just searching for Google for different things. Right? Let me first show the same in action. Then we will understand it parally. So here what I'm asking this table to do I am asking it to query or search on Google about data hack subummit. Yes the same thing which we are a part of where we are sitting. I just asked uh this full workflow. It will be writing a newsletter right and now it can be article it can be block but it is real time. So on real time it will go on Google. It will search for data hacks summit and then it will try to write a full newsletter which we can use. So let us see the same in action. Now let me just quickly push this and you will see this visually happening. So it is for the timing it search for the net that okay what are the different data hack submit references an agent is now creating the sections of different different uh let's say whatever it has searched on. So let it work it takes a little bit of time right and then after that it will be editing the copy so because we want our newsletter to be pretty fine and it will be sending the same to my Gmail. So I will show you the result as well in just a minute once done. So again you can see the same visually. So that is something which I like. You can see the step as well. And I've collected an open AI model pretty nicely right. And finally we have send a message. So let me go to my email. Let me refresh it. And see we get dive into AI innovation at data summit 2025. And all of this is pretty based on what all is happening. If I showed you that what Tavly got right, it got all these things. So from district in it was able to get that data submit is happening. Then from LinkedIn somewhere it extracted that yes someone has added for data hack subummit and then for data hacks subummit it is getting the same from the analytics with the website. So something which I think analytics with the people will be doing and they will be spending a lot of time. I created this workflow yesterday just in an hour right. So all this I was able to do and if anyone from analytics with want this news data let me know but once this completes right the same workflow can be needed because if they will update that yes it was a success which I'm pretty sure it will be my Google search API it will quickly get those results only and if you want you can just automate this whole thing right you can just get the latest AI updates you can say that hey just get for the last Monday and you can get the same in the in your email so as I've shown you The best part now let me explain the component as well. The best part is that how it is writing different different sections. So if I quickly go and explain the same. The very first one I just manually trigger this one. This is a tavly which is an API which searches for Google and on the net for different resources. So I ask for data submit it is searching for that and got me the result like it is giving you the score as well that okay how close it feels the result is. So you can have a threshold. Let's say if I search for AI or some very specific technology, I want the threshold to be more than 90%. Right? Then let's move forward and see the next step. Here is the first agent which comes into picture. It is writing different different sections for me. Right? So again it's a pretty big one. I have made sure that it is writing the same in a easy to understand HTML manner. Right? So it has given all these things. Next if I move forward and let me just show you. So we can just go like this as well. I'm just aggregating all these different different sections. After that second agent comes into picture which tries to make sure that okay whatever I have given that has that is properly edited right just on the basis of whatever it was present. There are no repetitions between section it is having this flow. So just like how I will do it in a real sense when I will just research everything then I will try to write something then I will try to get it edited and finally send the same to my manager for a draft which will easily I think take a day uh you can do it in a 5 minutes if you try to create this automation and enjoy the rest of your day and finally it sends an email so I've just entered my email it is giving you the subject and the content now you can change it whatever you want you can easily change it you can add more option as well right we can just remove this edit and automation as And once I execute this, it just send me the email for draft. So let me go back and see I've got that. So embrace the AI future all these things. It is giving the link as well the source. So we can go on district and see everything. Similarly I can go on let's say this new thing. This is the second link the exact website. So it has scraped that in real time. Right? Tomorrow when there will be new updates it will be able to do that as well. So that's the whole idea with respect to creating this and everyone I hope will be able to understand how difficult it would have been if you are doing it manually or via code as well. 10 years back something creating something like this would easily have taken me a week. I would have to go through the APIs understand everything. Now even if you don't know what a API is you can just drag and drop and create this. So that's the overall benefit. Now let's move forward and talk about industry use cases. So the very first one is healthcare which I have taken. Coming to healthcare I actually helped one of the client uh who was a doctor which I known for a very long time. I helped create a chatbot for him right and integrate that in WhatsApp where his patient because they used to come to him for very pity things like okay how when I have to take this medicine given the full prescription or based on their previous overall diagnostics right so we saved all of that and we made sure that all these question can be answered in real time right saving thus saving a lots of time so when person used to ask that hey uh this medicine is not available can I take some other medicine it used to reply that yes this is of salt you can take some other medicine when I have to take this medicine I'm not able to understand your writing something which happens with doctors then we can surely it can just tell you that hey this is where you have to take this medicine in a very proper to uh understandable way for the layman person as well so that is something which I created now again we can have different different use cases because healthcare is on sector where technology or automation is never the major overall aim and they are very less on these resources right healthcare do not normally uh push and just research in this. They are majorly focusing on helping the patients right. So all your different different hospital they have the common problem with respect to your appointment scheduling with respect to billing all these things can very nicely be automated. Then you can create a virtual nurse assistant as well where in your let's say hospitals you can just make sure based on what all is happening you can easily just create a system which tells the person what exactly is going on how is the diagnostic and everything happening. Then you can also have a patient data summarization. Anytime a doctor comes he will be seeing that okay what is everything which is like let's say for the last 5 days patient is admitted he sees everything right? We can give the same in a summary to him and considering that how it can be very quick, it can be accurate and you can always have this AI thing or any API as well integrated. This is going to generally be a gamecher which we can use in each and every of the industry no matter even if you are a simple developer, if you are an analyst or if you are a CEO or some CXO as well. Next is again some use case very close to me finance. So I worked in the finance field for three and a half years. Now in finance as well we can have multiple use cases. The very famous being the fraud detection and alert right then how you can make sure that any any transaction which is happening right we are able to see that in real time tell the person about the same we can send in the offers. So this having this AI agent thing and quick prototyping help us a lot. Now I have covered a very niche use case as well here which is algorithmic trading because this is something which I'm again in connection with one of the client where helping him to understand if we can just have some summarization and understand the same via the no code tool. Right? So this is something which I'm doing and yes that is possible. Now let's move forward and just have some key takeaways from this industry example which we have covered. Now surely there can be retail, there can be sales, there can be CRM, there can be different different industries or rather all the industry where AI can be integrated that is for sure and these tools they make it very easy for you. So they have democratized the overall AI experience. Even if you don't know or have very less idea about AI, even just what I've told you here about agents, you can create an agentic workflow for yourself. Right? Then again it is very very fast. We have already understood that how I just created that newsletter thing in last just I think two hours one hour yesterday right. So it is very quick as compared to writing the code for that. Then next come about adaptability. If you tomorrow want said that hey charge GP is not giving me a good option just remove that block just drag the Gemini one and you're all set. Hey, Claude is now very good in understanding the use cases. The new Claude model, that's a game changer. Just remove the Gemini one, drag and drop the Claude one. All good. Hey, I want to send the output to Gmail as well as Slack. Just drag and drop, connect that dot, and everything will be there in Slack as well. Right? Then next is a little thing scalability. Now, traditionally, no code tools have been this bottleneck where they were not very scalable. But with the new one which you can host onto your system as well, they are pretty nicely scalable. They can handle multiple requests and I personally believe that you can easily create a system for a prototyping and you can scale it to let's say 100k users easily right it will be able to handle all those request and it will be able to give you fabulous result in a very quick way then augment not replace specifically included this point the whole idea is that we are not here to replace humans we just want to make sure that anything which can be automated and done smartly by AI they can be done and then humans can focus on more better problems Right? And that is I think the major role of AI is if you are trying to do something which can be easily done then AI is going to replace you. If you become very good and adopt these things very nicely very quickly then of course you can just focus on other things focus on the major task which require logic and let AI handles the rest right great so let's now move on to the second penultimate slide and this again may everything is good you told us for the last 20 25 minutes we understood the idea but how would I use it in my work how would I use it in my company so that's why I've included this slide because it is very very crucial that from this presentation you take away that okay how you have to do and actually use that else it will be not very helpful so the very first thing is to start small please don't go and have your core banking logic or the core logic of your system to the no code tools then mail me that bang I followed your advice everything is failed I'm fired I don't want that please try to pick up a very very small use case very high impact use case try to automate that try to see how that is happening right understand that yes this is something which is costing me a lot of time. This is a simple use case, high value use case which I can automate to provide a better experience. So if you are sending email to your uh let's say customers when they sign up, make that email a little bit personalized based on their location, based on their uh let's say details, right? You can just reach out to your uh let's say customers, potential customers, if you have any past data, let's say if a customer has not purchased anything for the last 6 months, just easily get a system where you can send this mail, right? So don't try to just have the whole logic of your company on the first day automated. You will surely get fired and it will not be on me. I made sure that I am warning all of you. Next is the executive team. So you will also have to make sure that once you do that hopefully after uh getting a little motivated uh from my side, you will have to make sure that your manager, your leadership team, your stakeholders, they understand things, right? So tell them just educate them a little bit. Show them that hey see this is something which I have quickly uh created uh let me do this let us not go to the tech team for this or let me just quickly create this prototype then they will surely be happy they will try to understand tell them that how this makes AI easily integraable tell them that how competitors are doing it because believe me every industry is doing this competitors when it they does uh managers or stakeholders are more likely to understand and show them the same in action with a particular use case then after that you have to make sure that your Teams are also empowered. Everyone knows from your use case. So once you do that, do share that with everyone. Right? Don't try to keep it like hey this is doing my work let me not enjoy. No please share that with everyone so they understand right? They are also able to do the same. You can create a common repository and don't try to add or end up where we end up 50 years back where we used to create Excel macros for everything like hundreds of same Excel macros are there right in the company or same code is there. No, share your learnings, share your no code prototypes, share your templates so that they can use the same. If anyone wants they can reach out to me. I will share that newsletter va so that you can just easily create a newsletter on anything. So this is very very helpful have the hackathons and everything then after that once you have enough people in the company then create some center of uh excellence. You can have some expert I know one you can reach out to me and just have some sessions understand that how your use case can be done because many time we have this thought that no code tools they are just for basic things not for the advanced use case which we might have in a company but that is not the case in the last couple of years there have been lots of advancement so they have genuinely made it possible for complex use cases as well right so after that once we have done this introduced now how to scale so if you have now introduced used. If you have all the last tips which I talked about then we can expand on the use cases. You can slowly try to understand what are the new use cases or the previous one which can be introduced. You can easily add AI into your existing use cases as well. Just try to have them onto no code tools. Just do a AB testing where some users are getting overall from the last one which was getting used and the new users they are basically getting served by the new overall no code automation which you have done. Always always make sure of security and compliance because in the industry in a particular company this thing is very very important from the start. Make sure that you are taking your IT team your comprise team together don't just develop in silos where you are just creating something having no idea and your whole data is getting lost. No, always make sure that whenever you automate a use case you are having the approval you are having the oversight of your compliance and the IT team. We don't want to create a system which breaks or cost us more just to save time. Okay. And then of course culture change uh as I said automation as of now as well is seen something which cannot be used but I personally know people I have helped brand I have helped companies startup basically and major companies as well where I have created their use case and they are scaling pretty well. So that culture change will be required. One very important thing which I would like to add here. Don't uh be afraid if let's say after some complexity you have to move from no code to two uh no code to code because many people have this problem where they say that hey I I might get 1 million request why should I move to no code? Well you can do a quick prototyping you can easily serve the customers and you can make sure that they are getting till let's say the 1 million hits after that if it becomes complex you can easily move on to a codebased solution as well. So this is something which I have seen and it is no problem because at least we have done the prototyping pretty easily. We were able to get to the market early and serve the customers. So yeah that is I think where I have covered about how we can also easily introduce and use that and thank you. So I think that was a very good timing thing. So that was the last slide everyone and uh I hope you got all I've understood and we will be open to Q&A now. So >> so yes we have 10 minutes we can take some questions. Yeah, sure. So, everyone I think you can easily ask the question. I can help you answer that if anyone has any. >> Hi. Uh I'm here Ashish. >> Yeah. Hi Ashish. >> Um great session. >> Thank you. >> I've actually tried to use NA10. >> Mhm. >> And uh there are areas where I'm stuck. I'm an decent codable person but no no code is my forte. Mhm. >> So when we are trying to do chat bots >> right uh we want a respon you know sort of a looping system where you have a chat that is where I struggled. >> Okay. >> So is that something these n platforms can handle because I couldn't get it to work. >> It is easily possible where you can have a loop kind of a thing it is working on the let's say whatever response is coming. >> So n supports that or rather any tool will support that. Uh one thing which I forget to add is that in n you can also have a code as well. So just in case if you want. >> So that becomes code right? >> No no no it is the use case which you are supporting which you are telling that and many more complex use cases they are out of the box. >> Okay. >> In case if there is still something it is supported by code but you will not be requiring that for 90% of the use cases the one which are present they are pretty good. So the looping thing that is just out of the box. >> So you you want the chatbot to take a decision based on the response and then loop. Yeah kind of. >> So I can show you that as well. >> Is that a Yeah. Is there a GitHub or some >> anything? So we have just have a simple loop object which we can drag and drop. So if I go and let's say go here just have a loop >> and this I'm trying to do via Telegram or WhatsApp. >> Yeah. So we can have Telegram, WhatsApp anything which you want to add that should not be a problem. So this is a loop thing. Let me just delete this. But like that's the idea like you can just easily connect that. So let's say it is going here. Now I can easily go and connect it like this. So depending on your use cases, it can easily handle those things, right? So looping is supported out of the box and it works pretty well because I've used that in one of the workflows where I was iterating and seeing that yes, at least the agent should spare two three times before it comes up with a final answer and I have genuinely like I handle most of my personal things via personal telegram chatbot where I just send the message, it adds something to my calendar, it can send the email for me, it can tell me what are the events. So I've created a personal uh full manager for that in Telegram. >> Yeah. So just to add and we can connect offline basically uh if it's a flow-based system where I have question one get an answer based on that answer take a question two >> is that also >> yeah yeah so we have the memory right our agent has a memory so it will understand that what you asked so let's say if I ask you that hey uh what is the best place to visit in India in this weather it tells uh let's say for example it tells Bangaluru only weather is nice now let's say if I ask can you tell me where all I can go it will have that context that yes it gives the bangaluru as an because of agent it is having that memory it is having that brain so just like a travel agent will do your AI travel agent will also do the same >> awesome thanks I'll reach out to you >> no worries you can actually I create I will be covering those use cases on my YouTube channel as well so you can just quickly search that I have created a full master class there so more than 100k views so if anyone just wants to get started into this n I have created that as well okay >> hi hi nice session uh actually I'm also looking for this kind of we said 98 is kind of open source. It means >> there is no license restrictions and we can >> no you can host it. Yeah, it's open source. It's on GitHub. You can host it onto your system as well. Like if you have let's say you can easily just have it's a docker image. You can just get it. You can host onto your your full infrastructure as well. >> There is no license complications and all those things for commercial purpose. >> No, it has anise thing available. If you want to have let's say support but overall it's open source. You can easily just host it for support and everything. It has thing. So no commercial issues as such that should not be a problem. >> It mean the source code and everything is a extensible and all that >> it's extensible >> and you also mentioned about that be ready with the code with the complex. So we have 1 million hit and all that stuff. So the there are two kind of personas right one is like who primarily look for more API level and one is uh >> how you balance it between these two things because uh a lot of people as you rightly mentioned I have let's say 20 agents they need to talk about it and this these guys will not be able to scale and all that >> versus edge cases are you guys able to handle those edge cases properly versus my graph state and all those things so those questions I'm keep on listening from my developers saying okay this is the uh foreign initial use cases it's good >> no so it is no it is not the case as I said this is something which is there in the industry as of now and I know that because I've done that again with the industry so I do consulting AI consulting overall where I help companies in that now it has become a lot mature genuinely I can just have a node I can connect it to my backend API even if I want the logic to stay with me I can have the whole logic I can just send it any response I want in JSON so that can easily be scaled and handled plus uh with the way it is getting developed very nicely it is in proper development you are able to have those edge cases etc easily handable and it is scalable as well like the free version which I just showed you it is able to take five concurrent request if you host it onto your system and you scale that up let's say on any EC2 or anything it can go further than that should not be a problem and anyway you can have a major way major case on the API onto your system as well and it can hit your API so you can just have a node which hits your API so all that complexing all that scalability if you want if you don't want to it outside you can handle on to your system so should not be a problem at all. >> Yeah, of course it will support the opensource models hosted locally. >> Yes, it is supported. Uh even nit is scalable like I know about others thing as well but let's stick to nit that is scalable because it can be on your infra so if your infra is good it will handle that. >> Perfect. Thank you so much. >> Thank you. >> Hi. >> Hi sir. Yeah. >> Okay. It was a nice session. Yeah. >> Uh I was quite following N for uh few some time. uh it's uh I think some of the answer you already given. So uh we are actually basically from a cyber security uh space >> and compliance is the biggest issue for us. So whenever I propose the solutions with NAT >> and how secure how we can make it secure so probably you said one of the thing is that we can actually host it in our inhouse. >> Yeah. Yeah. >> So that our data won't be gone out. >> Nothing will lose from >> nothing will be outside. So and then our LLM models and everything we have we have hosted inside. So it uh >> that can be supported local LLMs can be supported. So once you host it onto your infra then you can have a local LM which is powering that. >> Uh I don't have it ready but yes I can support that via Olama pretty nicely. Just let's say if it's on your local host it will the data will be yours. The LLM will be yours. Nothing is going out. Just have a quick compliance oversight and you should be good. Again solved a use case where that was the problem and specifically that comes into finance domain and the cyber security domain. So don't use the pre the given one on cloud >> just have it on your uh infra that should be fine. Uh uh and then the second thing how so if at all if it is going for NOS's so how the cost effective how much it >> uh NA10 is pretty cost effective right with respect to other ones and when you host it on your onto yours you will be just giving majorly the cost of the infra plus let's say any of the LLM cost which you are incurring uh when compared to competition I think it is easily 70 to 80% cheaper I had that slide I think I missed that but yes it is having with with respect to Zapier, Microsoft automate and others it is pretty cost effective and I have seen that cost has never been an issue into anything which I have automated and as of now because since it is open source it's it can be on your infra is able to handle that migrate to other >> so we can easily migrate it to other third party uh systems also correct right like that uh third party integrations also >> yes yes that is Outlook, teams and other stuff. >> Yeah. So it supports that out of the box otherwise it can connect to any API. >> So something that even if you host it in know so all these integrations would be available. >> Yeah. Yeah. Yeah. So as soon as you even get it on your local these integrations are supported. You might just have to get a web hook so that it can connect to your edit but otherwise these are supported that is not a problem. So they are there out of the box. >> May >> here. Uh so talking about the uh doctor's use case you right. Uh if the patients are chatting with the agent >> example uh but the agent has to respond back to the user directly >> yes >> is can there be a human in loop while the doctor approves the response >> we can have the human in the loop so what we can have and what like I will share from the use case only whenever there was any query which let's say agent doesn't have an idea or it was let's say very complex or it was out of overall where the person has been diagnosed etc then we made sure that that message used to be sent to the telegram of the doctor. If he says that yes this is what you should do or if he tells the answer that is what is in said. So human in the loop supported out of the box easily. Your workflow will wait right till the doctor responds and for your agent uh sorry for your c end consumer it will just be sending the message once doctor responds or it will not be sending anything or it will just after some time you can make it send that hey uh this is a complex use case uh doctor will be reaching out to you easily or let me like if you want to just easily imitate a doctor we will send the reply and some message seems like we have to talk on that. So human in the loop can easily be integrated. MCPS human in the loop agents all these AI related hot things they are they are supported the good thing again that is not very difficult but good thing is that they are supported out of the box so yeah that is there all this is there >> hi >> we'll take this as the last question because time's up yes audible >> great session thank you um I just want to know when you create a workflow with nit Okay. And how do you deploy it? You know, if I have to give it as a solution, >> uh how do you deploy? >> It has a so I had a manual trigger where I was just triggering it manually. I could have it triggered based on different things. It can be time based. It can be when you get a new Gmail, when you get a new email. >> It can be based on a web hook. So let's say if you hit that API then it is when it will trigger. It can be based on new WhatsApp message, new telegram message. So the overall first row which I showed on the input or trigger, right? There can be different different triggers. So based on that your workflow will get executed. It's totally on you. You can have multiple triggers as well that hey either it will get triggered every day at 9:00 or if let's say I hit that API or I I get a email. So it will do that like just as an example I have at 9:00 all of my email it just creates a draft reply for me so that at 9:15 I can sit and reply to everything. So and if I want I can just trigger it from my what's telegram as well like telegram is normally easy WhatsApp it required business meta business account. So based on those triggers it can get executed anytime and any way you want. So that is supported. >> Okay. Thank you. >> Okay. Thank you. >> Thank you. >> Thanks a lot everyone. Thank you very much. >> So great everyone. I really hope that you like the talk. Now if you have come this far pretty sure that you are very serious and I can assure you that you are going to have a very good career and overall understanding in this field. If you have any doubts again and if you would like to ask the same, please ask the same in the comments and I will make sure to answer that. As always, please share this video with everyone and do subscribe to my channel as I will be bringing more and more content into this space and into the overall AI space as well. This is going a lot exciting and based on my expertise and knowledge, I want to make sure that I can help you each and every one of you. So with that, thanks a lot.