From LLM to GPT-4: Demystifying AI for Small Businesses

Anthony Karls of Rocket Clicks and Casey Shea discuss how AI, particularly ChatGPT, can boost small business productivity by 40%. Learn about the evolution of GPT, practical use cases, and how to create tailored AI tools to streamline tasks like resume screening.

00:00 Introduction to Revenue Road
00:31 Understanding AI and ChatGPT
02:42 Impact of AI on Small Businesses
04:20 Five Key Tips for Using ChatGPT
07:24 Live Demonstration: Using ChatGPT for SEO
12:30 Creating Custom GPTs for Business Tasks
19:59 Conclusion and Final Thoughts

Check out our blog here!

#DigitalMarketing #Sales #BusinessGrowth

Curious to discover your personalized roadmap to scaling a law firm, no matter where you are in the business?

Follow these steps:

1. BOOK A FREE 30-MINUTE AUDIT WITH US: https://rocketclicks.com/schedule-a-family-law-quick-audit/

2. CONNECT WITH US:

LinkedIn: https://www.linkedin.com/in/anthonykarls/

https://www.linkedin.com/company/rocket-clicks

Facebook: http://facebook.com/2311.karls.anthony

http://facebook.com/RocketClicks

Instagram: http://instagram.com/karls.anthony

https://www.instagram.com/rocketclicks/

3. TELL ME WHAT YOU WANT:

Tell me in the comments if you liked this podcast and what other kinds of topics you would like to see.

Transcript

Anthony Karls: 0:01

All right. Welcome, welcome. So this is Revenue Road back where we talk about sales and marketing for local entrepreneurs. I'm Anthony Carls, president of RocketClicks. So today I'm here with Casey Shea. So last time we talked with Casey, we talked about technical SEO. mentioned in that video that he's also one of our. AI champions here at RocketClicks. so we're gonna be, we're gonna be tapping him about that today. So we're just going to introduce the topic. We're going to talk a little bit about it. he's going to give us some pretty cool use cases. You learn something. So we're talking about AI today. so before we jump in and kind of do our little show and tell. So what is AI? What's chat GPT? Like, where did it come from? What is this thing? Everybody's, everybody thinks it's brand new. Cool. I know you have different perspective on it. You gave a great presentation to the U. S. community that we presented to a couple months ago. What's AI? Tell me about it.

Casey Shea: 0:55

So basically AI it's not in the sense of sci fi movies, Terminator, Skynet, you know, is going to be unleashed. Basically what it is it's a large language model. What it does is, and this has been around for decades. It's been used since the beginning of the internet. I mean, it's been around since the eighties, really. They've been using Large language models. And what it does is it just predicts the next word, the next character of a sentence.

Anthony Karls: 1:23

So how does that differ from what exists today? So this is kind of, it's been around forever. So like, why is everybody so excited about JetGPT as an example? That's what we're going to use today.

Casey Shea: 1:35

So some things changed back in 2017 when open AI they took a large language model and they just pumped it full of a whole bunch of information. It was like over 2000 books. And then they also took, I think it was around 83 million reviews off Amazon. And they just wanted to see what it would do. And then all of a sudden. This pre trained transformer, or as we know, a generative pre trained transformer, GPT, it started to figure things out on its own. And it started to figure out like the emotion behind certain words. It began to think, and then they're like, all right, well, what if we put more information from the internet into this thing, and it just continued to grow and get smarter and smarter as they did this.

Anthony Karls: 2:19

Okay. So, so basically it went from an LLM. a more advanced LLM where it's kind of thinking kind of putting a lot of things together and contextualizing it based on all of the information that it potentially has. It's kind of what I heard you say.

Casey Shea: 2:35

Exactly. It's just, it's starting to think on its own, not quite Skynet, but it's getting there.

Anthony Karls: 2:41

Nice. Okay. So on revenue roadmap, we talk about things that are impactful to understanding that will drive a small businesses revenue. So like, how does this fit in? Okay. To, to that what is, what does that look like? Like, why should I consider this as a small business owner? I have a small team. We have a pretty good business that we're running. Like, why should I look at chat GPT as something to use?

Casey Shea: 3:07

The biggest reason why is just reflecting on a couple of studies that have been recently done fall of last year with Stanford and Harvard. They took individuals from just any small business, put them in a room. No experience whatsoever with chat GPT with any kind of AI. They gave them just a quick run through, just like the five basics that we're going to kind of talk over today. And with that, they saw a 40 percent increase in productivity. So just within a week of just playing around with it and with no prior knowledge whatsoever, 40 percent increase. I mean, I'm pretty sure any business would love to see a, an increase like that.

Anthony Karls: 3:43

Yeah, I'd agree. Okay. So that's, so it's impactful. and it's only getting better. So I know it's made a, from three to four was a big jump. I don't know if five's coming out this year or not. I've. Recently heard owner CEO of open AI talk about kind of when he thinks five is going to come out and that's coming out soon ish. So we'll see, but that one should be better than four. So we'll continue, but all right. So let's talk about, you see, you mentioned there's five things. So there's five things that, they gave these small business owners in this study that made them 40 percent more effective. So what were those? What did they. Like, how do I use this thing? How do I start thinking about it?

Casey Shea: 4:20

Basically what they did is they just ran through the five most important ways to speak to an LLM or a GPT. Those five things are going to be. First and foremost, how are you going to engage with this? You have to give it some kind of role. What is it doing? What do you want it to be? You want it to be an expert in marketing, advertisement, SEO, whatever it is. It could be a recruiter. It just needs some kind of role that it can really start to. Refine its knowledge into, because you got to keep in mind, this has the entire collective internet in its knowledge base from, I think it was April of 2023. So it's a lot to figure out. So it needs a specific role.

Anthony Karls: 5:01

It identifies role. Okay. What's next?

Casey Shea: 5:04

Second one depending on this one is kind of, it's optional. I always say, so this is what kind of tone do you want it to speak in? So if you're breaking down some like technical jargon or something like that, that you have no understanding of, you got to say, Hey, speak to me like I'm 10 years old and I know nothing about tech. So give it some kind of tone. And then also if it's writing an email for you or a report or something like that, you want it to have more of a professional tone, or sometimes you want it more conversational if it's a friend or a coworker.

Anthony Karls: 5:36

So who am I? What's my role? Okay. And how should I speak back to you? What's my, what's the tone you want this back in? So what else?

Casey Shea: 5:45

Ask, it'd be very specific. It needs to know what it's doing. It doesn't need, Hey, can you please do this? And I'd really like it if you did that. Nope, this is your little robotic slave. You just tell it what it needs to be done in as few words as possible, but also as specific as possible,

Anthony Karls: 6:03

So who am I? What's the tone you want to receive it in? And what's the task? What's number four?

Casey Shea: 6:08

ask questions, I can guarantee you that the first result is not going to be exactly what you want if it does, Hey, like you're doing better than me most days. But. Ask it questions just like it's someone that you're training, someone that's new on the job. Someone who's not very familiar with the project that you're working on. Talk to it conversationally, like it is a person. Once again, though, you don't need to have all the manners and the fillers and things like that. Just get right to the point,

Anthony Karls: 6:35

Okay. And then what's number five? So ask the questions is number four. What's number five.

Casey Shea: 6:39

Reworking your prompt. So sometimes it's just, you can have a conversation. You can keep talking to it. You can keep asking questions about what it's giving you. But sometimes you just need to start from the beginning and rewrite the directions that you initially gave it. The other one that kind of blends in with five is also understanding. GPT and AI has a really short memory right now. So if you have a big, long conversation, maybe you're going through 50 different resumes, you know, that's a lot for it to handle and remember through a full conversation. So there's certain points where you want to start over with the conversation because it has a limited memory. So you want to make sure that you start over and you don't go out too long.

Anthony Karls: 7:19

Okay. Got it. Okay. So that's the useful. So how do you use it? So give us an example, show us how you're actually doing this. So pull us up on, on the screen. Let's do a little show and tell. Let's do a little mad science experiment with Casey. So we can show you off a little bit.

Casey Shea: 7:37

So this is something that I actually just used earlier this week. So what I did, I gave it a role. It's an expert in SEO. And then what I wanted it to do, I wanted it to read and analyze. This page right here, it's going to go on the internet. It's going to look at it. And then what do I want out of it? I want specific content opportunities. I want to know what is done poorly and what needs to be optimized. This is just surface level. What's that?

Anthony Karls: 8:07

you enter, what is what is SEO Sage?

Casey Shea: 8:10

So,

Anthony Karls: 8:10

chat GPT, so what is this?

Casey Shea: 8:12

so Sage is going to be something this really shows the capabilities of what you have at your fingertips with a little bit of work, a little bit of prompting and just a little bit of hands on. So with. GPT, you are able to go in and create your own GPT that gives you step by step, and you can continue to just fill it with knowledge, instructions. With my SEO Sage, I've got a number of different documents in, in the back end of it that have everything from Google's guidelines, To what it wants out of the search results, which is 163 pages that I'm not going to read, but my AI will do it for me as it's going through content. And it will reflect on the documentation when it's giving me answers. So this is my sage for SEO.

Anthony Karls: 8:57

Got it. So this is basically a, this is on open AI, right? Or chat GPT and you've created your own version of it to do more specific work. probably need a shirt. So in this example, you probably don't need as long of a prop cause it's got a big reservoir of information cause it

Casey Shea: 9:15

Yeah.

Anthony Karls: 9:16

it is and it's supposed to do.

Casey Shea: 9:18

So with this one, it's a very it's built to be just a general expert. And SEO. So therefore my prompts for this are very short. Otherwise with something like this, I would have to say, you're an expert in SEO. Here's the project that we're working on, you know, giving it some more context behind like exactly what it is so it can refine what its task and what his job is. But with that, I've got it already done in the backend. So this simplifies it. And then in the future, I can just go into my left menu and I can just click it for whatever tasks that I want in the future and not have to redo the prompt every time.

Anthony Karls: 9:53

So what is. So that's super cool. I got a question about that after. So what did it do? What are we trying to do here? What question did you ask it? You gave it what?

Casey Shea: 10:01

So I wanted to know what content opportunities there were on this specific page, so on this website with this specific page, which is their commercial page and then it also included a link just in case I wanted to click it myself. Okay. And then it gave me just some high level specific content opportunities, certain areas of optimizations, and keeping in mind that this is reflecting strictly on exactly what Google asks us for. So if you've heard of EAT, if you've heard of the guidelines that Google continues to put out, this is reading that. And then it's reading the page and then it's putting them together. And then as far as, you know, other business owners, this might be a bunch of technical jargon, you know, maybe you want to run this on your own website, but you don't understand SEO, jargon, nerd talk, anything like that. So, explain it to me. I know. Explain to me I have no idea what digital marketing is

Anthony Karls: 11:04

Okay.

Casey Shea: 11:05

and it's going to start getting a little bit of goofy. So you can continue to simplify it, refine it in a way that you can understand it. So this is really useful, especially if you have really long documents, say it's a contract or something like that. You can load this contract into GPT, give it the specific role, the specific instructions and what you want to, what you want out of it. And it's going to produce that, and it can continue to refine it more and more as you're asking questions and getting exactly the result that you want out of it.

Anthony Karls: 11:37

Okay. So this is, this example, this is great for you and kind of, you've been able to download your years of information and like you're thinking about SEO into this, and then it can can be your admin assistant for doing some initial research, which is awesome. How would I use this as a business owner? How would I use chat GPT as a business owner? What's a, like a use case would be like it's really hard for me to like screen resumes. When I post a job, I get hundreds of resumes. I got to look through all of them. It's a pain in the butt. I don't have time to look through all of them. I'm probably missing stuff that I should be catching. I gave you that as a potential assignment. So what did, like, how did you think about that on the documentation we have that we have our own admins run? So like, how can we make that simpler and use chat GPT to you to do it?

Casey Shea: 12:30

So, first and foremost, let's start over, and let's go out of my Sage, and let's just go with regular GPT 4. One thing you may be asking, GPT 3. 5, What's the difference? 3. 5. I use that for like maybe organizing some numbers, calculating some things, very simple things. Think of it like a middle school education where in chat GPT four, that's more of a, it actually, it ranks in the top 10 percentile for college graduates. So it passes the bar exam. It passes the medical exam. I like to go with four, but there is a limit as far as how many times you can use it within a time span, just because it uses a lot of computing power.

Anthony Karls: 13:13

Got it.

Casey Shea: 13:14

Something that we worked on at RocketClicks for screening resumes is creating pretty much like a checklist that it gives it once again, very specific instructions for when it's scanning through resumes, and this is something that a recruiter, you know, could do, I'm pretty sure that nobody Enjoy scanning through resumes, trying to give it a point system for each one. This is a lot to go through and it's a really long document. So one way you can do this, if you don't create your own is just copying this, bringing it right up. Yeah,

Anthony Karls: 13:52

admin would from their HR person or maybe you're a smaller business. You don't have an HR person Maybe that's you but you're responsible for recruiting. So like maybe it's your paralegal. Maybe it's your admin. Here's the things We're looking for Go look at all these resumes, which is that's a, that sucks for a job, right?

Casey Shea: 14:08

not be happy with that job. That's for sure. But luckily enough, we can take all this stuff right here. And just, I mean, that's a lot to go through. So what it's going to do is it's going to run it back through just to make sure that it has a complete understanding of what you're asking for. So it's breaking everything down as we can see. And then what we want to do is put in just a little dummy resume that I just pulled off of Google as just a sample resume. And all I did is drop it in there. It already has the prompt. It already knows what to do. I dropped it in there and it's going to break it down according to exactly what our original prompt had. So the way that it's scoring it, the way that it's looking at different PPC strategist criteria and really assessing how they fit into this role. And then at the end. Gave it a total score of six. You can continue to ask questions. So why only a six? And that's also why you got to double check too, because it missed one. It looks like

Anthony Karls: 15:30

Nice. So it's basically taking the instructions and then it's taking the instructions and comparing it with the resume. And it's basically giving you a score.

Casey Shea: 15:37

exactly. And then something that you can do similar to what I did.

Anthony Karls: 15:41

all you did was log it into, all you did was load it into, load the PDF here and that for you.

Casey Shea: 15:48

Yep. I just dropped the PDF. First I dropped in the prompt and then it ran back what the instructions were. I threw in the PDF of the file and this can, it can take, you know, PDF, TXT, a number of different files, word files. It'll tell you if it can't read it and then you can format it accordingly.

Anthony Karls: 16:07

Okay. So what if, how do I, so how do I create, How do I create the expert stage or whatever you would like to call it so I don't have to reprompt it and I can continue to feed it better and better information and keep training it on screening these. What does that process look like?

Casey Shea: 16:22

you'll see right here. Explore GPTs. There's a number of GPTs that people have already made. So for all I know, there's probably one already made. You can search for it, but I want it specific to this exact way that our company does it. So with that in this top corner, we're going to create it. And this is where it walks you through everything. So now that I already have this prompt document, so I'm going to go right into configure right away. And I'm going to drop that in there. So I'm going to take this, I'm going to download it. So I'm going to download it into a TXT file, just because I don't need to worry about any kind of formatting or something like that. And TXT is really small file for it. And then we're going to upload, just going to hit prompts, and then it'll load up in there. And then that's where you go over to the create. Usually what I'll just do is take this first prompt just so it has that recognition right away. And then you're going to put it in here.

Anthony Karls: 17:31

So this is kind of the 5 steps that we talked about earlier.

Casey Shea: 17:34

Exactly. So act as a human resource recruiting specialist and then breaking down all of the steps. And then you'll see it's updating the GPT. And once it's done updating, you can continue to refine it. You can continue to come back to this. You can save it. You can share it with your coworkers. You can share it with whoever. But it walks you through every step of the way. And then just to test it out too, while it's asking me all these questions that. Oh, it's going to make me do it. Yes.

Anthony Karls: 18:07

so what did it make you do?

Casey Shea: 18:09

Let me give it a name. Gives it a name. Gives it a picture. It's going to say, is this picture great? I'll say yes. It's great.

Anthony Karls: 18:19

so it suggested a name and then now it's going to suggest a picture.

Casey Shea: 18:23

Yeah.

Anthony Karls: 18:23

It's going to create your, the same thing that we hit, we saw earlier with your SEO sage.

Casey Shea: 18:27

Yep. So this just has some computer monitor. I have a similar one, but it's some old lady that's in some dark room because that's what I imagined when I think of someone just scanning. And then this is where it'll just continue breaking things down further and further as you go. Here's where you can test it out on the right hand side, just to make sure that it's operating the way that you want it to, if it doesn't. You go on the left hand side and refine it further. Tell it exactly what it did wrong. That's something that I find a lot of people don't do as they're like, ah, it's wrong again. I want blah, blah, blah. And they'll just repeat the prompt in a different way where it's you have to just like a child, you have to point out exactly what happened, what it did wrong and what you were expecting out of it. So you have to walk it through just like you're training an individual. And then as we can see. All I did is drop in the PDF. I gave it no prompting whatsoever, and it's breaking it all down for me. And then this way I don't ever have to put in a prompt ever again. I can just open this GPT, drop in a resume. I can drop in several resumes at a time and it'll go through each one of them.

Anthony Karls: 19:36

Nice. So it's going to just, it's going to come up and score it. Now this is basically a new app, a new GPT app that you could use.

Casey Shea: 19:44

Yep. Just created it. So this all works. And then right here, you can share it. You can have only me link. You can even throw it in the store if you feel like it. I'm just going to say anyone with a link and then it's saved forever.

Anthony Karls: 19:58

Cool. Well, this was awesome. This was awesome. Casey. Really appreciate you walking us through this. It seems like there's huge opportunities here. You just kind of have to sit down and think about it a little bit and. I think I think when I was trained on this the instructor said, you got to do a lot of FAFO. want to know what that means? Go Google it. You'll find some funny videos. I think that's honestly what you were talking about and you just got to keep playing with it and improving it and you'll see some pretty cool results. So appreciate your time and guidance here. Any other last suggestions on thinking about using AI or chat GPT for our audience here?

Casey Shea: 20:32

The biggest thing is use it for everything. Just even if you don't think that you need to getting used to it, because this is going to be integrated into everything that we touch. So it's going to be. Everywhere. So getting used to how to use it on a day to day basis, but then also keep it simple. The biggest thing that I always see people doing wrong is they over complicate the prompt, just short and as simple as possible. And you're going to get the best results.

Anthony Karls: 20:59

Well, thank you, Casey. Appreciate it.

Casey Shea: 21:01

Yeah. Thank you.