Zapier and ChatGPT With OpenAI Beginner's Guide To AI Automation
Let's build with Zapier and AI (100+ videos)📅 2023-05-31
Transcript
[00:00] welcome back to web Cafe AI where we do daily chat gbt and AI videos for your personal and business life and today's video is going to be quite a special one I want to do a full-blown mini course where I essentially show you everything that's entailed with starting of AI automation understanding zapier as a platform and understanding how to integrate that with open AI if you've been a watcher of this channel then this video may be more on the simple side but if you're brand new welcome to web Cafe AI here we do a lot of videos when it comes to AI Automation and how to leverage that in different parts of your life or in different parts of your business so this video right here is
[00:31] going to be a full-blown mini course where essentially I show you from start to finish everything that's entailed with the zapier platform everything that's until the open AI platform how they complement each other and you'll get a fully understanding of how to use AI Automation in whatever specific use case you have finally if you like this kind of content that is more longer form mini course oriented let me know by just simply liking the video it's free and it helps us here at web Cafe AI hey let's go ahead and start off here so phase one is we got the Creator accounts to do that you're simply going to click the
[01:01] two links that I have in the description below and we're going to start with zapier all you need to do is start for free with Google or start an email and then if we go over to pricing as you see they'll have a tier that is completely free and you only have to have up to 100 tasks a month this is a great starting point for anyone new to automation as it's gonna allow you to play around a little bit with understanding the different stuff entailed with you know zapier flows and zaps in general the next link I have in the description below is going to be open Ai and this is going to essentially allow you to
[01:32] integrate Ai and all your zapier flows so all you need to do is come up to the sign up button right here and sign up with an account and once you create your account you'll get access to a dashboard that will allow you to see billing usage and your API keys that will integrate later in this tutorial now your next question might be hey Corbin well this is Grace sounds like it's free but what are the costs associated if I wanted to kind of scale or use more tasks here essentially how zapier works is it's going to be on task basis the idea behind a task is let's say you're running a flow this will make more sense as we dive further into the
[02:03] tutorial but just for now so get a general gist right off the bat let's say you have a full that has five different blocks those five different blocks would equate to five tasks once the flow has been run once so in this context as you see here for 20 a month you'll get access to 750 tasks if we move over to open AI open ai's price structuring is a little different so it is no initial fixed fee every single month they aren't going to charge you ten dollars a month they're going to charge you 20 a month no matter what their way of charging for their service essentially is based off usage so if you only use AI from their
[02:34] gbt4 or gbt3 like once or twice you're only gonna be paying one or two pennies the entire month so their structure and their context is purely based off usage rather than fixed fees subscription so that is good to know as we build up these models and these flows on zapier because maybe we want to use GPT 3.5 because it's cheaper than using gbt4 and these different models to clarify on that point as you see here in order for a thousand tokens of usage for gbt4 it's going to cost you around three pennies a thousand tokens of usage
[03:05] of GPT 3.5 is zero zero two pennies now right now you might be saying Corbin I don't even know what the heck a token is don't worry about that in this tutorial you'll see how to not only calculate that but also monitor that and really understand what each prompt does and how many tokens are typically associated with every single output that you have in any type of folder you build on zapier and open AI okay so you have officially now signed up created your zapier account you've signed up created your open AI account the next stage here is let's understand both of these
[03:37] interfaces so that you can know how to leverage and really understand how to like walk through each platform and you don't necessarily get lost when trying to find something useful so your homepage might look a little different than mine I'm currently enrolled in their beta program where they'll just give me new tools and new features in order to like beta test with them and give them feedback for example just create a zap tool which we did 81 on this channel so if you're interested in that go and check that out but for now all your homepage really will have essentially is going to be your recent zap so you worked on and recommended for you so different zaps that are pre-built
[04:07] in the community and then you know all these little footer options uh you don't necessarily have to care about those for now that is all you have to really care about when it comes to the front-facing dashboard on zapier if you take your cursor and you come over here you'll be able to see all these different options here starting with zaps in order to minimize or maximize this side door all you need to do is come up to here and say open sidebar and now it's maximized if you want to minimize it you just hit close sidebar alright so now that we're in the area of where our zaps are located there is a couple stuff we want to learn here first off you can search a zap by name
[04:39] right here second off what I like to do is organize zaps based off the context that they're being used so for example all the zaps that I've created for this YouTube account and I have shared in every single description of those videos I store in this specific folder so I know it's designated that hey these are you know YouTube's apps that I've shared and essentially I don't necessarily want these being deleted because I want you guys to have access to them in order to create a folder all you need to do is come over here to this plus sign and I could just type in test here
[05:10] and then as you see here it will create that little folder there and with these three ellipses that's a lot of the navigation so I can move a folder into a folder I can rename that folder I can create a new folder within that folder so folder two and then you see right there new folder within the folder and then I can finally delete it only if the folders within that folder are deleted as your mouth full so let me go ahead and delete these real quick because I don't necessarily need them right now now the next cool feature here is that if you go from private folders to Shared folders let's
[05:41] say you have multiple members this is essentially where you'd want to put the folders here and this essentially allows you to share folders with other individuals on your team so you can both be working on a zap and you can kind of collaborate more in that context now don't mind the names right there this is just where I beta test a lot of different zap ideas that I think of and I just leave it in the home folder from here though you have a really good idea of the navigation on the zap panel the last thing you'd want to understand is this little button right here if you click this it will select all zaps Associated so if I go to YouTube's apps
[06:12] real quick I can go ahead and mass click this and then I could go ahead and say move to folder delete clear selection or select and pause all of them at a single time now let's go ahead and just create a zap We're not gonna do anything crazy here because we still need to do a lot more backdrop of this platform in general but I want to show you essentially how you create this app in order to do so you're going to come over here to create not new folder we're going to say create a new zap and then once you're in a new zap here you'll be able to start off essentially by adding that trigger that you'd want which could
[06:43] be any of this stuff obviously we get deeper into this tutorial I plan on telling you a lot more about what we can do on this platform when it comes to this part of it but that essentially is how you create a zap on the surface level the next little tab that you see in your sidebar is transfers this shouldn't be too important to you if you're brand new to the platform you'll rarely need to use this essentially the point of it though is to send data in bulk from one app to another this is useful from pointing records or completing a batch process this isn't too fundamental on the early stages of understanding AI automation so we're not
[07:13] going to go too deep into here the next thing you may see on your sidebar depending on zapier if zapier released it or not yet is going to be tables essentially tables is their way I think of it like a Google Excel sheet or Google Google Sheets within your zapier account but what's great about that is that now with tables you can essentially uh you know it's a level up of a Google sheet because now we can add automations within each little cell we can do different flows within the actual tables themselves this might be a little
[07:44] confusing we do a whole tutorial on that on a separate video but just a quick overview for that now the next little tab here is going to be interfaces think of interfaces as a separate web page builder the context you use interfaces is either for internal team use such as building out like crms that are internally used to drag to different flows or using it for landing pages and so on for your business this is where you're going to essentially connect all the different apps that you want to when you want to use different flows so for
[08:14] example if you want to connect your Instagram you'd connect it here if you want to connect your Gmail you would just simply type in the name of the app and you will be able to integrate it so now we'd be able to access it and you know either do actions or triggers with it so the next section we have here is zap history this is going to give you an idea of where your zaps are going and how they're being used and for example we can see zap runs here and this is going to give you a chronological idea of what is the most recent zap that was running your account so this context would be the IG to Pinterest and then you come over to tax task usage and you
[08:45] can kind of understand okay well this day we did 61 tasks worth of zaps and then you can kind of get an overall summary at the top here so you can understand what a zap is let's go ahead and jump into one of these so we're going to go to the IG printers which essentially is when a Instagram post is posted it will essentially push a Pinterest post with the exact same post and then use gbt in order to make the contents of that post and then as you see here you'll get at the top here how many tasks were used so in this context three why were three tasks used it's because we have three
[09:17] blocks post trigger one two three that is basically how they associate cost with everything so that means if you have a 70 block automation flow and just so you know when I say blocks I mean just any type of this is a block this is a block it doesn't matter what the program is that's just what I refer to as a block if you have 70 blocks and one trigger to trigger all 70s 70 of those blocks that's going to cost you 70 zaps every single or 70 tasks every
[09:48] single time that automation runs if that makes sense this leads me to the next thing which shows how many tasks have been burned so far in your 30-day billing cycle so for me since it just started only 99 have been burned and then right here you're gonna get the limits of how many zaps you can have on your account I believe when you first started it's a Max of either 10 or 20 so you'll see that here for me it's unlimited because we're on a Pro Plan but for now that's what you need to understand about your plan if you want to manage it all you have to do is click this and you can either go to a higher tier or lower tier depending on the
[10:18] context you're dealing with okay so the next section here is explore so explore essentially allows you to kind of access the community of different zapier people and essentially the different flows that they have publicly made available so maybe you're you have a specific app like Twitter and you really need to understand better ways you can use automation with it all you need to do essentially would be clicking Twitter here and then you'd click one other app so maybe do Twitter and Gmail and let's go ahead and see if we can run it as you see down there it shows you different uh
[10:51] zap templates that are associated with the apps you clicks so here's a Twitter one here's a Gmail one and so on this is a good place to go if you need to get an idea of something to do on your personal account if you want to achieve a task of a certain app that maybe you can't find a lot of tutorials on although as you know here at web Cafe AI you you're going to be able to find a ton of different tutorials for you know preset templates when it comes to integrating AI with it and your final tab here is going to be the get help tab as the name
[11:22] says if you have any issues you can obviously contact zapier's customer service they have a great customer service they respond in time and really try to walk through and help you out when it comes to different issues you may face also they have a community board so maybe you can even post your question on a community board that people within the community can help you answer okay so now you have a general idea of the zapier platform what's associated with it and how to approach it let's go ahead and learn about the open AI platform which is going to be
[11:53] the AI in all of our different automations we do here once you log into your openai account you're going to be prompted with this kind of dashboard you'll have three different sections here you have start Basics build an application and build a chat TBT plugin for the purposes of what we're trying to achieve here which is Simply Having zapier and open AI work hand in hand these aren't too important as build application is more developer based where you know you're calling apis in order to integrate it to different stuff and then obviously build an higbt plugin that is much more developer based where
[12:24] essentially if you're familiar with my other videos on this channel where we go over plugins and be building out a plug-in for the chat gbt uh platform therefore a lot of these tabs you don't really need to care about in the context of what we're dealing with openai so for example playground is just a tool you can use to try different models and see what the output would be in certain context but you can still do that in zapier once we integrate it so that's not too important examples may be interesting to you if you kind of want to see some examples of prompt
[12:54] structuring such as converting movie titles into emojis right here convert movie Thousand emojis back to the future and so on so as you see the prompt right here is convert movie titles into emojis and then essentially it gives out that output you can even copy it here this is not too important stuff when it comes to our certain context but if you want to look at certain prompt structuring this could be imported there API reference don't have to worry about this because all we need for our context will just be the API key provided by open AI finally documentation more
[13:25] developer stuff not important for what we're trying to achieve here which is just purely AI animation with zapier so that is a brief overview of that part of the platform which is not too important to you what's important to you though is going to be coming up to your name right here and we're going to go to manage account this is where you're going to find your organization name and your organization ID that's hidden under me and that could be important in certain context where it requests and wants your organization ID you're going to find it right above me on your settings the next tab of usage is going to show you a
[13:55] breakdown day by day how much cost has been associated with the amount of AI you've used as you see here the bars you know fluctuate you'll be able to see a cumulative of this as well and then on top of that one thing I want to point out that is very important is this next little tab over here which is rate limits essentially any new account has a rate limit of 250 USD which in most contexts you don't have to worry about but if you notice yourself using a lot of AI because you're using it in a lot of different functions like we do here
[14:25] at web Cafe AI you're going to have to make sure you request a higher rate usage this is a lengthy process so if you know you're gonna be using AI a lot I implore you to essentially ask for a higher rate usage give your reasoning and then you know wait a couple days to see what the response is but this will make sure that you don't you know truncate yourself if the issue arises where you need to get AI but now you've hit your cap of 250 USD a month and it doesn't let you uh you know charge their model anymore okay so probably the most important part and what you have to care
[14:55] about when it comes to AI Automation and zapier is going to be the API keys so essentially I went here from coming up to your profile you can either hit view API keys or manage your account either one you're going to find in the side section here and all you have to do is press API key once you press API key The Next Step essentially would be generating a API key and let's go ahead and do that so we're going to hit create new secret key we're going to go ahead and name that secret key to test don't worry you're gonna see it but I'm deleting it after this tutorial so don't
[15:27] try to use it on YouTube alright alright so we're gonna hit test here create this secret key and then I'm going to go ahead and copy it to clarify real quick an API key is what gives you access to open ai's backend to chat gbt in the different models but also an API key is where open AI realizes like okay this API key was used x amount of times let's charge them this much so it's kind of like you know purely to you don't ever share your API key because if you share it and someone else gets a hold of it and they could charge your key out the Wazoo and
[15:59] without you even knowing but we have our API key now so all we have to do is come to my apps and we're going to go ahead and integrate it with two different things here so the first one is going to be open AI and you're saying okay why are we doing it open Ai and chat jbt why don't we just do one or the other because open AI is what's going to give us access to the ability for image generation and using whisper which is text to voice so it means some context you want to develop an SRT file might want a little faster but essentially just hit add new connection
[16:29] and then that API key that you just copied and paste here we say yes and continue and then I'm gonna go ahead and rename this as uh test account because I'm deleting it so don't try to steal it test account and then as you see here you can go ahead and click this you can test connection you can reconnect and so on if you hit test connections it's gonna be like is it confirmation is there a connection so we can want to do that right now testing test successful okay perfect now we're gonna go ahead
[17:01] and make one other connection here we're gonna go up here we're going to chat gbt so why are we uh getting Chad gbt is because this is going to be the main one you want to use when you have any type of context where you need to use the language model and not necessarily the image generation model so we're just hit add new connection and we're gonna go ahead and use the same key here say yes continue and then we can go ahead and rename this as test
[17:32] gbt all right perfect so we have officially connected AI with zapier now we can create AI automated zaps really cool stuff here so the next part of this tutorial is I'm going to show you some examples of zaps that have ai integrated with it I'm also going to show you how to understand these apps when they execute such as token usage such as timing such as you know block usage and so on so this next part of the tutorial is really going to show you how to operate when dealing with AI and zapier so let's go ahead and begin by creating a zap as we showed earlier in this
[18:04] tutorial we're going here and we're just going to go ahead and just do testing AI and let's go ahead and just learn a couple stuff here I'm gonna delete that real quick um the first thing you want to learn here is that open Ai and chat gbt don't have triggers reason is because what necessarily is going to trigger chat gbt and open AI to incur these are more of uh tasks or actions because of the fact that a lot of the times chat gbt or open AI are manipulating data that has been already generated before in the flow
[18:36] so just so you understand the idea behind the trigger let's go ahead and make a very simple one your trigger can either be a third-party app that you integrated or one of zapier's built-in tools as you see here if we come over here zapier has a ton of built-in tools and actually are very useful in certain contexts such as the formatter tool if I can find it this one right here for the context so maybe you need to format the output of the AI generated test and text and so on so you know understanding how to use zapier's you know basic tools
[19:07] really helps you to leverage your ability to build out very complex zaps but for the purposes of this tutorial we're just going to start with a very simple one which is going to be schedule and obviously as the name says a schedule or zap essentially will every single time uh the event occurs it will trigger so in this context it'd be either every day every hour every month or every week so let's just say every day this will trigger whatever is about an Cora after every day at let's say at
[19:38] noon if I can click this and we hit test trigger this will obviously work essentially a trigger will then occur so every day at noon this will basically gen uh incur a trigger to the action blocks now that we have made our trigger of every day let's go ahead and play around with the open AI Block in the Chad gbt block and understand it better starting with the open AI blocks we're going to hit open AI we're going to choose an event so we
[20:09] have a couple of events here so the first one being check moderation essentially in this context we could have text come in so maybe an email is coming in maybe a YouTube comments coming in and it wants to check moderation for like essentially as you see here hate hate threatening self-harm uh violence and so on create a transcript this is interesting so this uses uh Whispers API to create a transcript create a translation generate an image send a prompt API request so in this context if you know how to
[20:40] integrate API that maybe isn't necessarily natively integrated to zapier so what I mean by that is if you know how it says like zapier's 5000 plus apps there's still more than that right so maybe the 5000 third one is the one that doesn't natively integrate with zapper so you have to make a custom call for that which we'll get into later in another tutorial but then you can also search embeddings for specific things some of this stuff I encourage you to use the chat gbt block and at the open a block open AI block I encourage you to
[21:11] use the open AI Block in the context of creating transcripts creating translations and generating images everything else that has been outlined here and obviously the API either one everything else has been outlined here we're going to be able to use the chat gbt block for but in this context we're going to go ahead and choose generate an image and essentially this uses Daly's back end and as you see here we'll choose the connection we made earlier for us it's just test connect we're gonna go ahead and hit continue here now for the prompt here you're
[21:42] going to do traditionally what you would do for a prompt if you're not familiar with ballet prompts and all that essentially it's a AI generated text to image software so for our context we'll just say a dog at a beach obviously this can get a lot more you know Advanced but for the purpose of the tutorial we're just gonna do a dog and a beach and then it asks for the number of variations of an image typically you'll only say one you could say 10 and 10 different variations of a dog at a beach will be generated and then finally it's going to ask for a size we'll do 10 24 by 10 24. hit continue test that action and then
[22:13] as you see here you'll get a nice little output with the output we can manipulate and use that data so what you'll let's go ahead and see the output real quick the output is a URL just so you know this URL will expire within 30 minutes to an hour so if you have flows that maybe will call upon this URL like a day or two days later it will have an error where essentially like we can't find a URL essentially that means that this dollar image has been generated uh this link is only live for like 30 minutes so let's go ahead and see what it looks
[22:44] like all right so I opened up a new tab and then just pasted that uh link there and as you see here we got our AI generated image congratulations if you're following along you've just done your first AI automated task which was every day on uh 12 pm generate a dog ring on a beach pretty cool stuff okay so now that we kind of understand the function of generating an image using Dalai we can go ahead and try another function that we could use with uh open AI here which
[23:15] would be create a transcript so we're going to continue here continue here and then essentially the way that this works is that for the file it says mp3 mp4 and Peg uh m4a I highly recommend don't use a video file for this because it's just going to be too large use an audio file it's a lot faster when it comes to processing it and then you might be asking yourself okay well if you're just giving me a transcript what is the purpose of a prompt function here the way you kind of look at the prompt function is that if you're familiar with creating transcripts from videos in the
[23:46] past or audio in the past an issue that always occurs is that some words that are said or you know written down or outputted wrong every single time so a context for me is that when I've done transcripts in the past for videos like this chat gbt comes out to like chat gdt like spelled completely wrong what I can do then is essentially say every time you hear chat gbt spell it as quotation marks I'm going to do chat gbt and
[24:20] really surprising this actually understands that so if there's certain nuisances when you create transcripts that you're aware of this is a great tool to use in the sense of making sure that every single transcript that comes out is highly accurate and as we know it's a language model so you'll be able to talk to it like a human structure it like this and then essentially you know you can even ask uh you can even like list more so if I had another word like dog that was so wrong do dog and stuff like that the next thing that you could do here is the response format so you
[24:51] see text SRT webvtt Json verbose Json typically you use the SRT and then the language of the audio you're going to want to use uh this kind of character and they even give you a whole list of the other formatted like en for English es for Spanish and so on now the next thing you can do with open AI That's powerful is creating a translation so this is the same idea here essentially you could put in the file as an audio make sure it's audio even though it's suggest video or you can do video do audio enter a prompt like I said before
[25:23] there's certain things that you want to make very specific then the response format we can have as you know SRT and then essentially for the prompt let's say in this context it's just a translation of the audio file please I'm gonna go ahead and say please generate Spanish and make sure Spanish words are spelled like in a semicolon and then you just put parentheses to Spanish words that
[25:55] you'd want spelled correctly that you may have had issues with in the past so when it goes from English to Spanish you know it's a pretty fine-tuned transcript okay so now let's get to the chat gbt blog before I continue if you felt like you learned something up to this point make sure to like the video it's completely free and it helps us out here web Cafe AI but now we're going to get to the very fun part of what I think AI animation has obviously Dale's uh Dale and Whisper has its role but that's very niche in specific roles but the best part of this integration is the fact
[26:26] that we can access chat gbt okay to begin we only have one type of event currently which is gonna be conversation we're gonna continue here we're going to choose the account we just created which was test gbt and one thing I want to point out before I continue here is that you can rename these little blocks so if I hit this and just said gbt test you can do it like that that might help you when structuring your flows but from here we have a couple of input boxes so let's Understand Each one and what it means for you okay so the first one user message think of that as when you're
[26:58] using you know regular chat gbt with a nice little UI chat bot that's the product you essentially input this is where you can essentially tell Chad gbt what do you want to do and what's great about zapier is that now we have access to data that has been previously said so for example for some reason if you want to put Tuesday in the prompt and obviously it's a variable so this was sent off at Wednesday that would turn into a Wednesday you can do that now and that's what makes this whole function super powerful is now we can generate or create prompts that have variables within it so that every single output
[27:28] that is generated is going to be different and very specific depending on the data that happens previous to it now the next thing that occurs here is going to be the username here which is going to be essentially you as an individual and then the assistant name what I like to do is typically is keep these as the default of user and assistant there isn't a lot of context here where you would need to essentially change the assistant name or the username the purposes I could see in some context is that if you wanted to do an assistant name of like bot one and
[28:01] then bought two as the next chat TBT block maybe you want to separate the blocks for your own organizational benefit but for now there isn't a lot of use case or reasoning of why you'd want to change the assistant name or the username now the assistant instructions this could come into play a little bit I kind of see this as more of adding a cherry on top of a Sunday it isn't the whole Sunday the whole Sunday is going to be really a user message the cherry on top though is that let's say in some context you really want to build out a model of an SEO article
[28:31] generator in this area for assistant instructions you could say you are a SEO blog writer all this does really is it fine-tunes Chad gbt a little bit further than your original user message so it really kind of lasers down I'm like hey you aren't just a helpful message you're trying to make me an SEO article you're a SEO blog writer trying to make me that article that I'm trying to get right idea here essentially is that in some contexts the output is better than if you didn't give an assistant instruction
[29:02] and then I'll be honest with you in some context it's actually worse than if you were just to leave that as a helpful insistence so go ahead and test with that don't be too hard stuck on making sure that this is going to be changed every single time the next block here that we have is going to be your model and you can choose between gbt 3.5 and gbt4 and as I said earlier in tutorial GPT 3.5 is a lot cheaper than gbt4 but gbt 3.5 look at that more as a model that's used for grunt oriented task and
[29:32] what do I mean by grout oriented I mean in the context of summarize this gbt format this gbt translate this gbt that kind of context not necessarily in the context of like hey gbt write me an article hey gbt write me an email subject line with an email body with XYZ XYZ XYZ if you want more content oriented type of outputs you want to use gpt4 you you can use gb2 3.5 it will be fine it'll be able to function maybe you got to make your user message a little bit more intricate like it's still
[30:03] possible but if you want the best type of output the most high quality type of output you're going to want to use jpt4 okay so then the next section we have here is temperature and top p in most context you're going to keep this as defaults but as you see here a higher value will locate or reveal a more random output while a lower value will have a more consistent output honestly keep this at one as you're going to learn how to make it consistent every single time and exactly how you want every single time with this last little block here which is a memory key for top
[30:35] key it essentially does the exact same thing but in the context of probability using pre-context data that it has it's going to be like well based off the last three months of how we typically would output it this is how the user would typically like it that kind of thought and reasoning as you see here so a 0.1 means only the tokens comprising of the top 10 percent probability are considered so in that context you're dealing with only you're limiting the amount of output possibilities if you have it at a lower top P so I suggest
[31:06] keeping that in one as well finally and probably the coolest thing that's been added and why I really suggest chat DBT block natively rather than the open AI send prompt is the memory key the purposes of this tutorial let's just show you what a memory key is live so a memory key is this essentially playing a memory key I think it's best I just show you from the chat bot side in order to do so all you need to do is this we're going to say generate five captions for a company that sells
[31:40] dog products enter perfect and then let's fine-tune a little bit more and let's say each caption have a emoji okay great so you might be asking yourself how does this have any correlation the idea here is this what was the first caption you generated and there you go unleash the joy elevate your furry friend's happiness with our
[32:10] premium dog products unleash the joy same caption why that's important is basically when you set up a memory key on zapier's end you're essentially creating a small little brain or a little database for it to understand what it was what was said previously in the chat this can be very powerful in a lot of context such as your you've set up a AI automated article generator but you don't want the Articles to repeat or use the same content well guess what now with a memory key you're able to achieve this because every single one or every
[32:43] single memory key that you identify is a chat on its own knowing this I can set up a memory key to like uh cap 10 and that will then set up a chat on their back end that essentially everything that's been generated every time this flows ran it will know what was outputted and what was said previously and you can actually take that into account when writing out your prompt up here one thing I want to point out is that this is going to be 32 random characters don't use special characters and then
[33:14] essentially that's what creates a memory key it could be anything it could be this anything you want if you want to make it so it's easy for you to remember maybe the type the the intention or the purpose of this block is social capture so you do social cap zero one and go from there one other thing I want to point out here real quick is that when you're choosing between the models and you see gbt 3.5 turbo zero three oh one zero three fourteen that represents March 14th and this represents March 1st don't ever use these models as as you
[33:44] know gbt continues to update and increase they're going to slowly discontinue though so if you build out a ton of ir automation on the back of these two you're gonna have to reset it because now you have to choose the ones that are consistent and Universal Universal which are the ones that are just simply named gbt4 and gbt 3.5 turbo now that you understand all the different input values on a chat DBT block let's just do some real quick format lessons so you can really understand okay well how do I talk to it what's the best way to talk to it what's the best way to approach it when
[34:15] generating a prompt so there's gonna be a couple things here one thing you could do here is do a context block you're gonna do a context then you do parentheses this is typically how you format or how we format at web Cafe AI when we're dealing with variables and so on context block think of this as you're giving context essentially to the bot of what its purpose is and what it's about to do so maybe the context is you are a social or you are a customer
[34:45] support agent for web Cafe AI and now it understands that whatever you ask it to do so you'll The Next Step maybe generate an email template for a customer now it understands that okay when I generate this email template for a customer the context is I'm a customer support agent for webcathy AI so entail that and put that more into its output let's do another type of block here
[35:15] which can be a parameter block parameter block identify limitations and ways you want to structure your output so for example parameter could be Max use two sentences that's an example of a parameter another parameter could be Max use two sentences and make sure to include the words XYZ and so on think of parameter as a way for you to really structure the output and fine-tune the output a little bit better another block you can do here is going to be a format block so maybe in
[35:46] some context you need to format the output into HTML so I have the output in HTML so for example email we want that so for this example we're going to go ahead and say have the output HTML bold important parts and then I'm going to go ahead and increase the model to gbt4 okay so as you see here we got the formatted email template in HTML one
[36:17] thing that may be annoying right off the bat is the fact that it says sure here's an email template and then you know gives the generation in the context so maybe you don't want text before it what I like to do is say generate email template for a customer no text before or after and then this allows gbt to understand okay when I output it I don't need to give some type of pretext or context of essentially what I'm generating now that you have a really good understanding of the gbt block and the open AI block there is going to be one little trick here I want to show and then we're going to dive into understanding what it looks like when
[36:47] you start executing these at scale and understanding the back end of it trick I want to show you real quick is understanding how to use other tools in zapier to manipulate the GPT data so in this context we're going to go ahead and use a formatter tool and what this is going to allow us to do is manipulate the text we go ahead and choose text is the event I'm going to hit continue here and then from here we're gonna go ahead and hit transform and we're going to hit replace and the input data that we're going to use is the GPT output
[37:17] and we're going to go ahead and click that so for this context if you ever want to replace a certain character so a lot of times what DBT does is has quotation marks in its outputs which can be very annoying but just for the sake of this tutorial and for the purpose of showing you how it works let's just say I want to get rid of this little piece right here this special character all I would need to do is put the input in say find that specific character replace it with empty space and then as you see here this new output it will have none of those so let's go and see what it looks like perfect so as you see here all of them
[37:49] have been deleted and gone use that as you will but this is just one small example of how to use a formatted tool to format the underlying data that you're getting one other thing that I forgot to mention I want to show you and I kind of mentioned it earlier but not really was the fact that the prompts are fixed text but probably the most important part of what makes a prompt good when it comes to AI automation is understanding how to use variables in the prompts so for example let's say you know generate an email template for a customer let's say we have the customer's data you can do customer
[38:19] semicolon we're going to put name location and we put product or service wanted each one of these are going to do a semicolon and we're going to put a parentheses and then assuming let's say we had a landing page or Landing form and we had this information previously given so maybe every time a Google form is submitted and it provides all that information you would just jump into it like this and you would simply find the specific data so let's just assume that the name is
[38:49] Tuesday let's assume the location in that form was uh you know the time and let's assume that the service in that form was like right here so essentially these are live data feeds that will be inputted based off the uh initial trigger obviously in this context I can't show it because I don't have an initial trigger of a Google form there I just have a scheduler but if you're interested in learning more about that kind of stuff and really using AI automation to the next level make sure to check out our other videos here as we dive deep into a bunch of different use cases when understanding how to prompt
[39:21] structure including variables okay so before we jump into understanding cost and usage one thing I want to show you real quick is how to share a zap in some context you may want to share zaps in the future so I'm going to share this zap in the description below but all you would need to do is come over here to these three ellipses and we're gonna go ahead and hit share template and then you get a little load screen here and it looks like it can't share it because we actually have to make sure everything within that template is been tested or skip tested to make sure it's been confirmed
[39:52] so in this context as you see here it doesn't like it because I provided no information for this underlying event so I'm just go ahead and generate a prompt again all right so I went ahead and just real quickly made the prompt again I'm gonna skip the test just to show you you don't have to generate every single time and as you see here since we skipped the test we're good to publish but in our context we're just going to share this app so I'm going to go ahead and back out here coming back down to here we title it testing AI we're going to come to these three ellipses and all we need
[40:24] to do is press share a template once you have pressed shared a template you'll come to a screen like this essentially showing you the blocks you can change these app title you can change the description of the zap you're sharing but all we would need to do is press share here and then we're going to go ahead and get afforded that link there we go all I need to do is hit copy and I'm good to go you'll find that in the description below so we are back at the IG to Pinterest app now I'm going to show you how to understand costs associated with every single AI run so
[40:55] this zap specifically has been run a couple of times so you'll see that for the data out when you press data out essentially this shows you you know the initial data in and then essentially once the zap runs what was pushed forward so in this context this is what was pushed forward for this zap one thing you'll notice though is if you scroll down here you'll get a token used amount which is 6459. now in this context for this one Zap for this one run this did not cost me 649 or 600 459 tokens this is the amount used in the
[41:28] lifetime of this zap if that makes sense knowing this and in order to understand how to use this we're going to go ahead and just write down this number so we're going to say 6459 use uh use tokens I'm gonna come back over here and as you see here we have a ton of Zap runs so let's go ahead and count how many are on a page and we say this so there are 10 on a page therefore as we see down here we have six pages so
[41:59] we're gonna say in theory this zap has been run 60 times so let's just go to our handy dandy Google Calculator here and if we type in six five four nine divided by 60 it's around 109 tokens used every time that flow executes so now the next question you might be asking was great now I know the amount of tokens but I mean this is a pain I don't want to do this every single time and I don't want to essentially you know you know go through this process every single time
[42:29] to get a calculation well you're in luck here at web Cafe AI we built all right so went ahead and put this in the description down below use this as your zapier and open AI usage Excel sheet so the idea behind this essentially is that once you know these certain parameters behind your usage you're gonna be able to calculate at scale how much it would cost every single run so let me go ahead and clarify that a little so in this context we have a form entry response a receive customer inquiry so this is where you put your you know example flow name your objective and I will get into
[43:01] the part where you would just need to input your information so the first most important thing to input here would be the cost for a thousand tokens and then the overall cost of your plan so under the context that we're paying 58.50 a month for a thousand five hundred task it looks like every single time a task is run it costs around four cents and then on top of that for the open AI tokens we're assuming that it costs around two pennies every 1000 tokens in order to change that we would just come over here
[43:33] okay so since we're using gbt4 we're going to increase it to three right here in order to Encompass that every single token cost us around point zero or point zero zero zero three cents and then from there what we need to do essentially is input the information that we had from our flow as we saw the flow has three tasks used so let's go ahead and make sure that's added okay so in this context the three tasks have been added we call them zaps in this context and then finally we did the math for the usage which came out to basically 109
[44:04] tokens used per flow or do per run so I can go ahead and decrease this to 109. and then from here for one run it looks like it's gonna cost us around 12 cents to run let's go ahead and find out if this were to run a hundred times All We need to do is come over here to uh this one zero zero there we go for me to push an Instagram post at a AI generated caption and then put that over to Pinterest if I ran that
[44:35] all for 100 times it would only cost me 12 bucks this really shows you the power of AI Automation and what's even crazier is that this is 12 bucks under the context that I'm on the zap plan of 58.50 if I got a higher zap plan and it actually cost me maybe two cents per zap as you see over here two cents per task it only cost me six bucks something to keep in mind so that's it you now understand how to create a zapier account create an open AI account connected to and officially
[45:07] access the ability for AI Automation and I took it one step further showing you essentially how to calculate the cost of every single AI automated flow in the context that you want to scale and maybe you want to do a thousand articles and so on if you like this kind of content make sure to check out the playlist at the end of this video as it's a whole playlist when it comes to mastering Chad gbt and zapier and really understanding AI automation when different specific unique context we go over stuff like social media posting article generation and so on you're going to learn a ton of
[45:39] stuff when it comes to prompt structuring data manipulation and AI automation as a whole as we have been entering into a whole new business sphere in these last couple of months if you're a business yourself and you just want to get AI automated integrated to your business as fast as possible go ahead and check us out at web Cafe AI as we are an AI automation service agency that deals with clients across Industries and build out these flows for them lastly if you want to get a higher education when it comes to this kind of workflow make sure to check out our courses here at web Cafe AI we go over a bunch of plethora different topics such
[46:09] as our business Master course or our social media management course and so on but without further Ado make sure you check out that playlist at the end as you're going to be able to get a bunch of different Niche opportunity cases for AI automation but lastly and without further Ado I'll see you in the next video thanks for tuning in NGS surprise I'm an AI Avatar make sure to explore more here at web Cafe where we demystify AI for your personal and business life until
[46:40] next time all right