How To Create AI Agents For Beginners (DeepSeek, Claude, ChatGPT) β
Let's build with Zapier and AI (100+ videos)π
2025-02-01
Transcript β
[00:00] we got a draft created through an AI agent let's learn how to create AI agents that require no code and can do anything we ever wanted on the internet welcome back y'all today's video we're going to be creating an AI agent don't worry it's not going to require any code I will be making another video that shows you how to do this with just raw code well that's a whole separate video this video I want to show you how to do it with no code show you how I can access any action within the entire internet and on top of that how do we let these AI agents do the work for us when we're just asleep so sound good
[00:30] let's jump in so right off the bat I'm going to leave a link to this in the description down below it's free to start they're called zap your agents what zap your agents allow us to do is access over 7,000 different applications found on the internet which means in this context it allows AI to do actions on our behalf across the board whether that is creating new rows within Google Sheets with data sending emails on our behalf specific workflows when it comes to notion I'll also make sure to leave this in the description out below so you can see all the capabilities that we'll have functionally with a zap your agent
[01:00] so to best showcase this I'm going to show you two different ways we can leverage AI agents first one being actual applicability within a business and then the second one will be a schedule basis where we can have the AI agent do any type of task at any time without our input therefore for whatever you want to do you're going to typically hit create custom agent but the steps and workflow I'm going to show you you can apply so we're going to do the support email agent for the first example here we're going to say use this template so here's what's cool once we set up this agent and it's live anytime I receive an email to my business this will automatically reply if a draft if
[01:31] it's a customer oriented email so let's do it first thing we want to do is make sure you connect your Gmail so for me I'm going to do my contact email we're going to hit save here so far so good that's essentially what's going to get the AI agent to start working the next thing we're going to want to do is provide data so that when it does answer the emails with a draft it's actually accurate so we got some suggestions here so what we're going to do is we're going to provide some data of a PDF for it to analyze therefore to provide this data we need to give context on the business itself and how to answer questions so a
[02:01] really great example of this could be using an FAQ section this right here is just bump UPS you can chat with any video local or on YouTube you can actually chat with this video right here if you get stuck link down below provide the YouTube link start chatting I'm going to scroll down here though we got a nice little FAQ section here obviously we could go a little bit further here provide more data but for now I'm just going to open every single one of these and I'm just going to do a simple copy just opened up a Google doc here I'll paste yes it doesn't look beautiful the formatting is not amazing but that doesn't matter when it comes to AI you'll be able to internalize all this information regardless so for me I'm
[02:32] just going to file export this as a PDF in theory we can actually keep this as a Google doc if for example we did a Google doc here we could in theory just update it live and then it would always have a live data source to reference therefore to have a live data source we can reference always because end of the day as a business things are constantly changing we'll go ahead and instead connect our Google doc but as you can clearly see there's a ton of different stuff we can connect here make sure you choose your correct Google account so so for me when I connect to my account it's going to be Google doc 2 and then the
[03:02] file name so bump UPS FAQ add data source so that set up here I'm going to go and delete this section here as it is requesting here we're going to go and move up all these instructions so far so good so let's go and read these instructions and then you can apply this kind of logic to whatever workflow you want to do with an AI agent when I get a new email look at the contents and determine if it's a question from a customer if it's a customer question look up the answer in and then we're going to provide our data source here I'll go ahead and just move this down here I'll reference the doc here as well bump UPS FAQ if you find an answer to
[03:32] the question there we're going to create a draft reply this is so it doesn't automatically send a customer email make sure to click that as anytime it does an action outside of just the chat itself or within the ecosystem of zapier it's going to require your input so for example let the agent find the specific thread let the agent create the body value to the CC this is a ton of information that if you chose to make something fixed you can so for example like label on your inbox we could set this to a very specific value like like any label we've already created within
[04:02] our Gmail so for example we could put important I could make a whole separate tag here that says chatbot etc for now though I'm going to let AI just handle it all but what's cool is that we do have discretion on whether or not we want to add very fixed variables here save now when working with AI agents you really have to narrow down how it approaches a problem so that's why we add stuff like important do not attempt to answer the question if you do not have a good answer because we don't want it to hallucinate in this makeup stuff here is 101 on instructions and when working with AI agents I want you to
[04:34] approach this step right here as if you're treating it like we're training a human to do it therefore treat the AI agent like an employee in your company if you wanted to give that employee a standard operating procedure what's step one what's step two what's step three all the way to end conversion event of what you care about so for this the end conversion event would be a draft reply to a customer email don't leave room for a big ousness or unclearness when dealing with AI agents because it will assume stuff that you may not want it to
[05:04] assume this comes with time this comes with practicing this comes with testing so now that we got that out of the way let's go ahead and check out an example here from bubbble here let's just provide a question from our FAQ here and see if our agent can actually reference that data source so what we're going to do here is we're going to ask a question is Bump UPS free to use and the correct answer is yes we provide a free 60 minutes so you can just see if you even like the platform so for testing send yourself a dummy email here so for example is Bump UPS free to use and then if I come over here and hit test behavior let's see if it works what we
[05:35] should see here is one it knows it can respond to this and do a draft because that is relevant with the FAQ data that we provided pretty cool to see a live right I'll check the bump UPS FAQ to find information about bump UPS pricing and whether it's free to use and it will show you the workf live here now the whole point of an AI agent though is for it to work autonomously so in theory we wouldn't really see this if we were like cooking a steak and then an hour later we like oh wait our AI agent did something an hour ago what is it and they did this so coming back over to the
[06:05] inbox here if I go to drafts what you'll notice is that there we go we got a draft created through an AI agent hello thank you for reaching out yes pump UPS is free to use you get 60 Minutes free processing no credit card required to start simply create an account to begin using all features if you need more video processing hours you can always upgrade your plan notice that it stays right on par with our FAQ Doc and in addition any typ response here or the way it's structured for example the best regards web Cafe AI we could put in the instructions always sign off as best
[06:36] regards XYZ best regards Corbin Brown Etc so once we're good and that's ready to go we can simply turn it on and then it's just live we have a AI agent going just a real quick shout out if you want to see how to set up an AI agent that is focused on web scraping EG we provide any type of website link and it gets all the data associated with that description down below card right there I did a whole video dedicated to this topic let's find out though how we create an AI agent that allows itself to fire off at a certain time so I'm going
[07:07] to name this AI agent stock news we're going to do here is going to say start from scratch name of the game here is that we need to configure our AI agent and we're going to add a behavior so we're going to create a behavior the trigger of our AI agent is going to be what we call a scheduler this is really nice and you can use this in any context right so we can do every day every hour every month let's assume that we're doing this every day and we want to do this at Market close so we'll come down here to 300 p.m. as I am CST and we'll hit save from here we can provide
[07:38] relevant instructions of what to do next so an action I wanted to take is to get the most recent news found here take action and we're going to do a RSS feed create item in feed if you don't know an RSS feed this allows us to listen to other websites when new data appears and obviously the new data we care about is going to be relevant to the stock market here so let's provide the correct feed URL simply put in whatever you care about for example stocks RSS feed we can do like RSS feed from market watch here we can do top stories and then all you got to do is come up here to the top
[08:09] link here and copy feed URL we're going to paste it here everything else can just be relevant to what the AI wants that's fine so we'll say based on the top story please give me the stock ticker and a short summary sent to my email add some clarification if no stock is mentioned just give short summary this is very much up to your discretion that's why this is plug-and playay and you can just kind of type whatever you want you get very very granular with the steps here but for now we're going to say sent by my email take another action anytime you want to do something outside
[08:39] the ecosystem of Zap here we're going to be taking actions here uh we have a couple different options here we can do Gmail or email by zap here we'll do Gmail we do send email and then the Gmail account just be your contact again we're going to make sure we're doing it to ourselves so we're going to set a specific value here of myself and then let's just let the AI agent come up with all the rest of the variables here when it's save very simple very easy now we know how to proceed in a schedule manner save instructions and test so when testing sometimes it'll prompt you to ask whether the data it's dealing with looks good for now I'm going to hit
[09:10] approve this kind of approve logic won't be necessary once we have it live but because we're testing we need to make sure everything looks good I want to also point out that because of the fact that this does have web scraping abilities you don't necessarily need to use an RSS feed here in theory you could ask the agent itself to search the internet with that simple setup though here's the email I would receive top market news vertex Pharmaceuticals breakthrough nice little email from the AI here is the latest top story from market watch this news is significant as it highlights a major development in vertex Pharmaceuticals provided the
[09:40] ticker which could have substantial impact on the stock market there we go and then in theory every single day at 4 p.m. it'll give me the top breaking news that I care about so that covers how to create AI agents in a no code way I'll be creating another video here on this channel of how to do it in a code way so make sure to subscribe here you're able to get insight on how to do this but if you wanted to code it all out which is a whole separate Beast so we'll jump into that we need to jump into that without further Ado I'll see you in the next video AI agent two random videos that's
[10:12] my face I'll see you in the next video