AI Agents EXPLAINED In 13 Minutes.

AI agents is the biggest opportunity of 2025 AI agents and you must be seeing it everywhere on social media in news and in the startup space as well hi everyone I'm isan Sharma and if you want to start learning about AI agents if you want to build your own AI agents this video is your guide to getting started with AI agents in 2025 and by the end of this video you'll have a clear idea about what is an AI agent how do AI agents really work how can you build your very own AI agents without

writing a single line of code and the biggest business opportunities with AI agents in 2025 watch till the end hit the like button and let's first talk about what is an AI agent simply put AI agents are basically like an assistant which is doing your work on your behalf which is figuring out the plan of action to execute something that you have assigned and executing it automatically without you having to intervene in the middle think of all the things you do every day you are booking cabs you are ordering groceries you're buying your own clothes

you're creating a to-do list for what to do today and in general doing all of these tasks every single day repeatedly imagine agents who can be doing all of these things for you on your behalf because they know your preferences imagine feeding all of the past purchases of the food you like to eat of the cars that you drive in of the type of clothes that you wear and now this agent will be equipped with knowledge with memory that is really important and now it can make decision on your behalf this is an application which

is a little in the future but today we have ai agents which can automate a lot of the things that you might be doing yourself for example replying to emails for example repurposing content from long form to short form videos hear me out mid Journey reached 200 million of AR with just 10 people cursor reached $100 million of AR with just 20 people lovable which is a tool to build any website you want in seconds reached $5 million of ARR with just 20 people bold. new on the other hand reached $20 million of AR with

just 30 people the future of work and business are small teams reaching crazy Revenue numbers by using AI agents to automate the work for them it's no longer having 100 people teams it's having a small lean team which can do all of the tasks and automate the rest using AI agents and that is why everyone is talking about it in 2025 but till now if you've only heard about GPT deep seek clae or perplexity you might be confused about what is this term really about like what's happening behind the scenes well let me take you

inside an AI agent essentially under the hood it's a large language model like GPT you have Claude you have Gemini you have deep seek any of those along with a tool that it has access to now this tool could be anything like searching the web it could be executing a piece of code it could be creating events in your calendar it could be creating creating content using a particular tool anything that you want it to do but it now has access to that particular tool to execute anything that it has thought about that it has

created let's take an example let's say you are really hungry and you want to have a pizza you can simply ask your a agent hey I am really hungry I want to have a pizza please order it and put it on cash on delivery what is happening behind the scenes is that the AI agent is relying on a large language model to come up with a series of steps to execute the first step could be figuring out what is the nearest pizza delivery place in your locality the second step is to figure out what pizza

would you want to order the third step is to actually add it to the cart and then proceed to check out add your address and the last step is to actually place the order so it will create that list and then it will proceed to check out so it is having that reasoning to create those series of steps and then it executes on that using the tools that it has access to let's say you have a llm like Gemini which can access your web browser and can do tasks on your behalf and that is what

you call a browser agent which has access to your browser it can search anything that you wanted to search now let's say in the case of ordering a pizza it opens a pizza heart and it real that pizza heart is closed now it will reflect and it will iterate on its plan that okay the first link is not working let me go back and let me go to the next link so it is iterative and it is figuring out what to do if something doesn't work out so it's not a just executing all the things

which you've said but it is also creating its own new series of steps when something is not working or when it has to change its its plans putting it simply just like you operate whenever you are executing a task the AI agent is doing the same thing all thanks to the large language model and it just has tools like it has hands with which it can do stuff on your browser it has other tools that it can use to do other functions that's all what a AI agent is all about what is not an AI

agent is let's say you ask CH GPD to some a document it's not really using a planning or reasoning model it's not even having any access to tools it is just taking that input and giving you that output that is not an AI agent that is just you querying that chatbot that large language model to summarize an AI agent under the H is a large language model which has access to tools which can iterate on its plans and can create new plans from scratch and reason how to execute a particular task that is all what

an AI agent is about now imagine a case in which you have multiple AI agents for example let's say I've asked a AI agent to create a complete website which has a front end which has a back end which has rest apis which has a goodlook design now this requires multiple you know skills for example it requires a skill for you to build databases it requires a skill for you to create nice looking front end it requires a skill for you you to debug any errors that you might face it requires a skill for you

to deploy that app when we are done with the production the development part of it so this has multiple use cases and on top of it you need a AI agent which can assign tasks to all of them this is called as a multi- agentic workflow in which you have one agent which is passing on its output as an input for the next AI agent you can either have a one to one multi-agent workflow in which one AI agent is passing on something to another AI agent and that is all it is doing let's say

I have an agentic workflow in which my input is the starting of a story line it then creates the entire story using llm that is one agent which creates that story the next agent then turns that story into audio using 11 Labs API that is the next agent and then it combines it all together and gives me the output so this is an example of a one to1 agent another type of multi-agent system is a one to many workflow in which as I described before if you have to create a full-fledged app from start you

need to have one agent which is supervising everything right this could be the boss which is overlooking everything and this agent would make decisions for what to do when it will create a plan of action that okay first of all we will create the database then we will create the front end then we will test for all the input Fields then we will maybe start creating the rest apis and then maybe we will test and use the error debugging tool agent and then at the end we will deploy it so this is the state of

actions that it has created then it will assign those tasks to every single agent underneath that you can have multiple agents which are doing the work simultaneously and underneath you can have even more agents but this basically is what a multi-agent workflow looks like now that you have a basic understanding of AI agents let's talk about how do you build your own AI Agents from scratch two ways to do it the first one is using code and in this case you will have to learn something called as Lang chain there is a deep learning AI

short CES on Lang chain that you can take up and understand how to build AI agents and essentially once you understand Lang chain the next step is to figure out how do you make this particular agent interact with anything so that is where you need to learn about tooling so for example let's say you want your agent to access the web and the web browser you need to learn about playright or you need to learn about Puppeteer you need to know how to create these connections learn about basic JavaScript learn about Json and then at

the end just learn how to bring all of this together and build the app using code another way to do it which is what I would recommend all of you to start with is using no code apps there are three no code apps in particular the first one is make.com second one is crew aai and the last one is n8n now crew aai has a dedicated course on deep learning AI Called multi- aai Agent systems that you can take up and learn how to use crew AI to create that multi- aai agent workflow from scratch

and it is very simple it's about a three-hour course and it goes in depth about how do you build your very first AI agent from scratch but the next step which is n8n is really interesting because n8n allows you to do a lot of customizations it's a bit hard to learn at first but once you get the grip of it you can literally build anything that you want to you can have huge massive AI agent workflows which are doing the things on your behalf and the way it works on n10 is that you have a

task which is assigned to a AI agent it has access to a large language model as I just talked about which could be deeps could be Claude could be Gemini could be GPT and then it also has access to tools so tools could be anything as I just said it could have access to your Gmail your Google calendar it can access any other app that you want it to access after that it should have a answer a reply and that is basically what a AI agent node looks like so everything in n10 works on nodes

you have a trigger node which could be you sending a message on chat you connecting it to Telegram and asking it something over there on the telegram bot it then takes that message sends it forward you also have something called as a human in the loop feedback in which you can actually pause this entire workflow until a human reviews and gives its own feedback so let's say for example you have a workflow in which you enter one word and the AI is supposed to create an entire you know text a real script for that word

and once that real script is done it then wants to get it approved from you so that is where the human in the loop thing comes in that's a node that you attach over there in which the whole system will be paused waiting for you to approve it could be like a email that you can reply to saying that okay this is approved or you can give feedback for some changes that you might want it to do it will again go back it will make those changes and then send you another email asking you if

the current script is good to go or not and once you give a green light to that then it goes ahead and generates the audio from that generates the video from that but this is essentially what a n10 workflow would look like for creating AI Agents from scratch that is all that you need to learn to get started check out the Deep learning AI course by crew AI by Lang chain there are other tools as well that you can use but just in general be curious right at the end of the day that is what

you need to have and lastly let's talk about the biggest business opportunities in the field of AI agents the best way to think about this is IM imagine all of the companies and where do they spend most of their money on they might be spending their money on hiring they might be spending their money on operations on finances accounting legal and a million other things figure out what are those tools which are these companies currently working with and if you can build a AI agent version of those tools of those SAS companies this is from

a recent V combinator video in which they talked about the best thing you can do is just to look at the best performing SAS companies today and start building AI agent versions of those in which the AI agent is able to do all the things faster and saves a lot of time as well and since there is no human intervention it is able to do it for cheaper as well so just look at all of those SAS companies and start building those tools those AI agents for those particular problems and just start solving those figure

out where is the majority of the money going by businesses I've made a separate video which goes in depth and talks about the 15 top business ideas for AI agents in 2025 so check that if you are curious to learn more about this that's all from me today I hope you learned a thing or two let me know if you have any questions below in the comment section and I'll see you in the next video the description will have some learning resources so check those out if you're interested there's a report that Google published which

is on AI agents so read that if you're curious and I'll see you in the next one bye for