Zapier and OpenAI Full Guide For Beginners
Let's build with Zapier and AI (100+ videos)📅 2024-02-08
Transcript
[00:00] as you may already know we can start leveraging artificial intelligence in automations what this allows us to do is integrate with over 7,000 apps or other words do humanlike task for 7,000 different applications that could be possibly incurring in our business therefore understanding open Ai and how to leverage open ai's artificial intelligence for automations is going to save you time it's going to save you cost and overall make your business more effective so in this video which is probably going to be pretty long we're going to go over everything you need to know welcome back here y my name is
[00:30] Corbin for those who have never met me before and you just clicked because you're like who is this guy and I actually do want to learn everything I need to know about Ai and automation this is going to be your video I don't know how long this is going to be but I'm going to make this extremely in-depth now that being said if you want to integrate open aai into zapier EG get the key be able to access your automations I encourage you to check out that video right there the 48,000 zapier and chbt opening as beginner I'm going to link it up there as well that's not going to go over this in this video this video is basically episode two of this
[01:01] where it's like okay I have it integrated I'm ready to go the key is good to go which you learn all in that 46 minute video this video though is going to give you basically perspective from two years of experience condensed into like 20 30 minutes from me on how to leverage this AI one last little announcement here check me out on Twitter it's in the description down below I'm actually creating my own software using artificial intelligence so that's a pretty cool series I have going on here I do a bunch of other stuff with Chad gbt co-pilot perplexity AI everything to do with artificial intelligence I'm showing you how to do
[01:32] it in in your business and your personal life let's jump into today's video so I'm going to approach this as if you have zero to little knowledge on open aai and just artificial intelligence automation so you can probably skip through this a little bit if you're a little bit more experienced although you may hear little cool things that you didn't know before if you watch this in full let's just start the basics we're here at open aai open AI if you don't know what open AI is think of it like an AI provider what does AI provider mean basically if you have developed software or you're familiar with software it's like a backend it's like our way of
[02:02] accessing the power of artificial intelligence and they provide it through a thing we call our a API which in software all that means is basically data transferring so in this context we're basically asking openai to use the data we provide and give us a more effective output we can leverage later on Numero Uno of what you should know is that supposedly according to open Ai and what they say in their legal documentation is they do not train their models on the use in open AI API if you use this version of Chad gbt and you're not on the teams plan yes they do
[02:33] train their data on the inquiry the data everything you put into this they will train it on but when we access it through its API which you'll see in today's video it's not getting trained on so that's number one that people may care about number two you're going to be looking at this and you're going be like what the heck's going on all right let me go in the middle y'all so I'm not covering up too much of this you're going to be like what is all this so I'm just going to give you a brief overview of all this CU we can actually access this in a no code way using zier so it's actually pretty pertinent and relevant for you understanding all this also I
[03:03] know someone in the comments is going be like you're you're talking too fast Corbin put me at75 speed this is just how I talk I mean not in every video but especially in these kind of videos where it's like a lot of information I want to condense a lot of my viewership likes it when I talk like this just because they're like you know I just want the information let's go let's go let's go let's go okay first things first gbt 4 Turbo gbt 4 what should we know here from here on out we're just going to be like AV in gbt 4 and we're going to proceed with gbt 4 preview this has 128k context what is
[03:35] that Corman okay that's this th000 per token think of it as 750 wordss okay the next question might be okay well a th000 tokens is equal to 750 words that we're processing what do you mean by process basically AI is looking at and doing something with it there is context or the way they charge is based off input and output as you'll notice input so let's say I put in 5,000 words is cheaper than the output what the AI model says so I put 5,000 words here and then it gives me two sentences that's
[04:06] going to be pretty cheap because it's you know two sentences what is that you know maybe 40 words therefore you don't have to worry too much your output unless you're outputting your outputs are huge which I'll show you in this video is probably not a good thing to do majority of the time the majority of your costs associated when using an endpoint like this is going to be with the input so as you see as of now it's 1 cent per 1,000 tokens or in other words 1 cent for 750 words we process so scrolling down here we come to our next model side note as well
[04:37] this bucket hat is from zapier so this is a zapier bucket hat I think I needed to wear because this video is basically going to be the meta of AI and automation gbt 3.5 turbo Corbin why are why do we have 3.5 why do we have four what's going on here so four think of it in very human oriented task for example if I want to generate a draft email I'm using gbt 4 if I want want to generate a social media caption I'm using gbt 4 3.5 then comes into play of why would I use 3.5 if the outputs of four or better 3.5
[05:09] is your way of accessing complex code task but with lame IND dictation other words you writing a prompt hey in this paragraph what is the one sentence that has this word in it the way you use 3.5 is much more of data fil filteration and condensation take this large 10,000w document condense it into 2,000 words and then use gbt 4 in order to analyze it so you're going to look at 3.5 as much more of a data condenser and a data
[05:40] filterer which will make more sense as we jump into the later video and as you'll notice and I'm part of the reason why I'm making today's video is they slash the prices again we're looking at 0.005 for 1K token and 015 for output to give perspective I believe the input used to be 150% more when this just started with the API end points for now though it's getting cheaper that's a trend in the industry overall I go over that with my software development you can check that if you want to next assistance API what is this Corbin assistance API is pretty useful
[06:12] and I have more complex tutorials you can check out on this channel if you want to do so if you don't see them right away they're either in this playlist at the end here or you can type in Corbin assistance API Chad gbt and zappier in your search bar and I should show up if I don't then I guess I'm failing YouTube SEO assistance API is basically you notice in chat gbt how chats when you talk to it it it can remember previous context of what you said that's kind of what assistance API does in the context of Automation and accessing API
[06:42] this is a very simple way of describing it I don't know someone in the comments be like well no chill out I'm describing it in the sense of using automations in AI it's much more just understanding previous context which we can leverage this video I'm not going to go too much into this you can check out me other videos for this when accessing 3.5 and four in automations these are one shot outputs for example we're basically putting in one input and getting one output assistance API we can have a ton of previous context and get a more in-depth output hope that explained that
[07:15] well let me know if you were confused in the comments fine tuny models we don't have to worry about anding models we don't have to worry about base models we don't have to worry about images you may need to worry about Del 3 is really cool if you don't know what D 3 is this is the AI image generator I would say this is probably the second best AI image generator in the market first being mid Journey no I'm not sponsored by either of these companies this is just from two years of experience up to this point or maybe a year and a half Audio models we do a little touch on this a little bit we can still access this in the context of zapier it is not too relevant for
[07:46] what we do for most workloads but think of it as text to speech speech to text this handles uh basically audio or speech translations stuff of this nature we have two major apps that we're going to be integrating into zapier which you probably already have if you watched the other video I pointed out it's going to be open Ai and it's going to be Chad gbt Corbin why are they separated because of the fact that both fundamentally do really different things um in the sense of basically think of open AI integration dis zappier as every single endpoint other than gbt 4 in gbt 3.5 the
[08:17] image Generation The Whisper you know random stuff like fine-tune things for creating translations this is where you're going to use open AI but the majority of your stuff is unless you're really leveraging Del 3 majority of your stuff is going to be C gbt gbt GPT I say that with a B I don't know maybe I have a a lisp maybe I have a lisp but jbt is going to be our main one here what you'll notice is zapier does try to help us by basically adding a
[08:47] bunch of like stuff they can could do write an email you know make it a little bit more user friendly but for most stuff as you'll notice on this channel is we use conversation because we know how to Proctor it on top of that there is also reference to finding or creating assistance which I referred to earlier okay we caught up we good to go we understand what open AI is we understand the cost oh yeah that's another thing why don't I just use gbt 4 all the time why do I even care about 3.5 because it's a lot cheaper one penny for
[09:18] 1,000 tokens or 750 wordss or alternatively are we going to pay5 for 750 warts that's why we need to understand and differentiate when using between 3.5 and four knowing this as a good rule of film and I've talked about this in the past basically if you have to have very complex logic where four four times you need to use some AI needs to be used multiple times always end your conversion event of whatever the end data that you want it to look like with a gbt 4 block
[09:49] everything before that can be 3.5 but it's very important that you end on a four block because this is like Insurance in the world of AI where you even if you could do it a 3.5 block you want to do a four as you want consistent outputs at scale if you run that automation a thousand times hope that made sense let's jump back over to zapier all sounds great Corbin but how the heck do we leverage it in the user interface of zapier let's find out so I'm going to do this J gbt on top of that let me show you a little trick here so when you're obviously dealing with AI
[10:20] or just dealing with uh making automations you might not know what to go or you might not know what to do so create test data so the test data I'm going to use today is going to be this pseudo invoice this is my test data today and then what I'm going to do is I'm just going to create a new folder called test data in Google Drive this is all free services by Google Google and we're going to go ahead and drag our test data into test data folder the reason we're doing this is so that I can just play around a little bit in this user interface and I
[10:50] can you know get some stuff going here so I'm going to do Google Drive and this part of the video y'all is I'm going to show you basically why I would use 3.5 in a context you're going to learn a little bit about prompt engineering you're also going to learn why I would use four in a different context with some prompt engineering so we're going to go ahead and do new file and folder choose our account here's going to be our courses account here at webcafe aai we're going to choose our folder which is going to be test data and then we're going to continue here test this trigger now if this all looks new to you and you have never used zapy before then you
[11:21] just found the gold mine check out this channel check at the playlist at the end here we I have over 400 videos yes crazy okay so let's go ahead and continue to selected record here and as you'll notice the reason I do that is because of the fact that we got our relevant information here but we need to get the text out of this dock to do so we're actually going to use zapier's new tool here um which is called files by zapier also want to point out big update that zapier did as of recently is they
[11:52] gave us they made it so that their toolbox what do you mean their toolbox uh their toolbox which is basically like their way of making it so it's easy easier to do logic uh without doing code I come over here and I come to zapier I think majority of these now are free they used to count as task but they're actually free now which is really nice on zapier's end giving them a Competitive Edge we're going to go ahead and choose our data here so we're going to do text from file this is our test data do file here and we're going to do file
[12:23] text and we're going to go ahead and hit continue test step here so in today's video is purely just to give you context of how to even talk to these models but if you actually want to see real world examples real world automations automations I've built in businesses backends you can check out that playlist hit continue and we're going to go ahead get our data here so as you see it's text invoice and scroll down Perfect all right let's go ahead and bring in that CH gbt block so first chbt block we're going to play with is going to be the 3.5 event here we're going to go ahead
[12:53] see there's a ton of stuff here so the only ones that I personally used and find Value in are going to be convers ation of assistant which I made videos on you can check out in this video we're just going to do conversation this basically allows it so we're not bowling or we got the rails up to make it so like we can't do crazier stuff we're going to bowl but without the rails we don't need the rails although those rails can be fun when you just just Chuck in this like goes like crazy all right so user message this is basically what you've
[13:24] heard on your Tik Tok what you've heard on your YouTube what you heard on Twitter new jobs in incoming prompt engineer in this is like this thing that you've been hearing okay uh that being said if you want to know how to monetize prompt engineering I have a video on that as well okay user message we're going to go ahead and basically put in the data this is how I like to format majority of my prompts we're going to do context going add it [Music] format parameters so these are all just different terms that we use here also we got important I rarely use important
[13:57] important only comes in into play it basically you've tried multiple times at an output and it's really just not listening we do important it furthermore if it's really not listening and you've proctored it a ton we have little tricks for that we use JavaScript and we're going to be able to remove certain characters that might have sounded confusing but don't worry it's actually pretty simple as I said before why do we use 3.5 3.5 comes into play actually let me make sure I wasn't covering this y'all okay I wasn't that would have been bad 3.5 comes into play
[14:28] when we're condensing data finding data and basically not doing complex outputs that think of it like another human's going to read that's kind of how you want to use 3.5 so let me just give you an example of this we got an invoice here let's just say the specific data point that we're trying to grab here is how much is due on this invoice all we got to do is this context we received an invoice for our company and then here's what we do so we're going to refer to it this is like kind kind of like coding
[14:59] but not really because like it's like we're identifying variables but without the constant stuff so all we got to do is this we got to do invoice text perfect also want to point out if you're player new text semicolon doesn't show up it isn't red it's only the invoice the actual text itself right this stuff that just popped up the amount do we're just going to do a very simple prompt here 3.5 memory key also side note we go to models here as you'll notice we got 0125 0613
[15:30] 0613 what the heck does that mean that is January 25th June 13th of 2023 uh June 13th so those little dashes that's that that's like um outdated models but what that means for us is we're always going to be basically pointing to the one that just looks normal so with this context we're just going to do 3.5 turbo that's fine memory key the most important thing that you need to understand and I made a whole video on it as well this is extremely important this is what allign allows for
[16:01] consistent outputs it can be anything random 32 character string all right we're going to come down here a lot of this stuff you don't have to worry about too much continue test step and let's just see what this first output is going to look like so we should see 750 USD perfect the amount due on the invoice is 750 USD I don't like that why don't you like that there's too much talking so we're going to do the amount due just the amount due no text before or after after okay we're going to add a one here
[16:31] refresh the memory key slot and we're going to come down here we're going to continue test this step so basically if this comes out to be 750 USD and is locked in with that memory key then we can ensure every single time data that is pushed through this system we're going to get just the 750 USD 1,000 USD 2,000 USD whatever it may be there we go that's 3.5 for you we understand 3.5 the ability to to grab data condense data Etc let's go ahead and add a CH gbt
[17:02] 4 block here because I want to add a little bit of a layering effect which is pretty cool here we go we're going to go to conversation continue continue we're going to do model go to gbt 4 as our endpoint here we go and here's what we can do so assistance API does help with this a little bit better but sometimes you don't really need to use an assistance API and you can just use previous outputs from Chad gbt to make your later outputs from Chad gbt better so let's just assume that in that first one there we actually just condensed a bunch of information for our gbt for
[17:34] Block to internalize we're going to go ahead and just say this context based on the amount of this invoice and let make in parenthesis I like using parenthesis um maybe it's just because it's from code uh or kind of we're going to go ahead and do an output here it also just keeps this so it makes that the data point is uh easier to be read by the AI model context based on the amount of this invoice 700 us we're using previous chbt output here generate a one sentence or
[18:07] four sentence email to a potential client that this is the amount due so it doesn't have context on the invoice yet I could add it here as a text block but I'm just going to go ahead and just give a little bit more information here we are a lawn service company named Apple and bees Applebees hold up Applebees I haven't been there like for years I don't know
[18:40] apples and bees okay Apple bees is that like a okay we'll let that slide uh generate for okay generate memory key random 32 character string continue test this step so we're going to go ahead and get a output that we could in theory basically the end user would see like the end user would read this we could add tone parameters bullet point list old disc HTML output whatever it may be this is where we leverage the AI language model in the gbt 4 context so as you see subject apple and B's Lawn Service your
[19:11] invoices du de valued customer we were writing to inform you that Apple B's lawn service has completed the job and the invoice of our work is currently pending so knowing this knowing that we can basically only leverage J dvt4 for outputs that are read by humans 3.5 for condensing data you kind of have a good comprehension out of when to use different models you could in theory still use 3.5 for a email output question is do you want to give that much discretion to something that could possibly lead to a failed conversion because of the fact that the way the AI
[19:42] language model wrote it wasn't up to part of what it should sound like which leads us to our next part of this video which comes to cost Association and how do we optimize these flows once we build them out so as you'll notice if I go to actions here I come down to test you will notice if you scroll down here we have a little bit of History context tokens used 317 and then history context tokens remaining what you need to worry about is the tokens used this tells you that for this prompt Len all this text
[20:12] all this data for this user request message it had costed us around 317 tokens of usage coming back over here this is where this cost comes into play Let's go back over here then we come over here and we go to this one go all the way down here and as you'll notice the amount of tokens used for this one was only 61 for gbt 4 now your next question might be okay how the heck do we effectively gauge so once I build out automation how the heck do I effectively gauge how much that's going to cost me each run I got a solution for you so I'm
[20:44] going to leave this in the description down below but this is our Marketplace a web cafes software.com I'm going to go to AI usage calculator here and there we go all you need to do is select the majority model that you use or the model in one of the flow parts s you want to put in the cost per zap your task which you can find out by watching this video you can put the number of tasks per run which is basically every single block here right except the zap your ones now and then if I come back over here we put the tokens per run which I just showed you and then we can just spitball it
[21:15] okay so hypothetically if I did this a 100 times how much would that cost me it'll give you the AI cost and the associated zap your cost and here's the best part this is free to use so proceed if you felt like you learned something up to this point make sure you leave a like it's completely free it helps me out here now if you liked everything you just saw in this video and it scratched a part of your brain you were just like hold up I need to learn more this playlist is going to pop up at the end here look how many videos for you to go through okay let me just point out something that that are really cool that will get your brain really flowing I
[21:46] like this one this shows you how to do like code like Logic for no code in a no code way this one's amazing if you want to check out Banner bear which is basically think of it like canva but now we can do automatic stuff in canva no more design please make my thumbnail no no no Banner bear coming over here jumped into every single possible app that I thought was really cool in the context of Automation and AI scrolling down here this one shows you how to automatically deal with Excel sheets okay no more manual data entry which is the bane of
[22:16] most people and we got basically how to integrate every single app in the world with zapier check that one out coming down here this one does an automatic ebook so if you want to create automatic ebooks you can check that out AI this one is an in-depth video on the memory key keep scrolling down here this one is probably the most useful video that I made in this playlist it only has 2,000 views but basically if you don't want to write emails anymore watch this one it's called guided to Aid driven email Solutions extremely valuable video if we
[22:48] keep going here we got this entire right here so we got 50 plus files ease this is the assistant API I was talking about so if you see my face like that this is what I'm talking about right here that is going to give you context on assistance API and there's a ton more the only other playlist I can suggest to you if you want to do stuff other than Automation and AI but just like AI Tools in general is I would suggest you check out AI tools for modern business I'm not going to link this one but you can check it out on my channel here ton of different stuff when it comes to different tools you can start leveraging in your business so that just about does
[23:19] it you have a fundamental understanding of how to leverage artificial intelligence with a no code solution like zapier as of course this can apply to other automation platforms like pipe dream make I is prefer zapy for multitude of reasons I jump into like a bunch of videos of why but all you need to know is this right now in this modern day and age Ai and the ability to access it through API like this in this manner has only been capable in the last 1 to 1.2 years therefore what does that mean that means there is a bunch of possibilities and potential for your own business to
[23:50] start leveraging in your back end and really streamline stuff I'm telling youall majority of tasks that you do right now could probably be automated with the use of AI that is because AI has become so fundamental that a lot of stuff a lot of outputs a lot of stuff we do already doesn't necessarily need a lot of creative thinking it's a lot of just data reformatting repurposing and stuff of this nature I'll see you in the next video so that is the playlist I was talking about jumping into zappier in all 5,000 apps that's a random video not too sure on that that's my face see
[24:22] y'all later