Become a Creator today!Start creating today - Share your story with the world!
Start for free
00:00:00
00:00:01
Ai Copilot: Your Next Partner for Decision Making w/Karl Yeh
297 Plays1 year ago
Welcome to AI-Driven Marketer, the podcast where marketers can learn how to leverage AI to enhance their productivity and efficiency. In this episode, host Dan Sanchez interviews Karl Yeh, an expert in utilizing AI for business strategy and tactics. They discuss the potential of custom GPT copilots for practical advice, content creation, and automating business processes. From automating call summaries to creating tailored sales offers, they delve into real-world applications of AI in marketing and business. Join us as we explore the power of AI and how it can transform the way marketers work.
Timestamps:
00:00 Building a business copilot for personalized support.
03:53 Custom GPT and language models for diverse tasks.
07:44 Create AI Copilot for specific business needs.
13:18 Chatbots for admissions streamline information for users.
14:30 AI chatbots like Chatbase enable data retrieval.
19:36 The gap in AI knowledge is widening.
23:21 Advanced ABM Copilot: Target prospect questions, supporting documents.
25:41 Interact with GPT, customize messages, use hotkeys.
29:39 Use prewritten outline, learn, write, repeat steps.
31:38 Custom GPT automates intuitive process, book on naming.
36:47 Respell tool allows efficient processing of audio.
37:50 Respell integrates with Zapier and Slack for insights.
41:34 Human review improves and speeds up tasks.
47:03 Excited yet scared about future technology advancements.
48:25 AI becoming real, implications uncertain, but hopeful.
Love these show notes? They're made by Ai with Cast Magic. Find out how to automate your show notes and more here: https://aimicroskills.com/castmagic
Recommended
Transcript
Introduction to AI Micro Skills Podcast
00:00:01
Speaker
Welcome to the AI Micro Skills Podcast where I'm curating the best lessons on AI for marketers who want to leverage the best of what's available now.
Meet Carl Yeh: Productivity and AI
00:00:10
Speaker
Today I'm talking to Carl Yeh about how to become more productive with AI. Carl, welcome to the show. Thanks a lot, Dan. I've been following you for so long and it's great to be on your show and connect.
Evolution of AI Commenting Systems
00:00:22
Speaker
Dude, that's the fun thing about LinkedIn. You could be building a relationship in the comments for years, and then when you finally meet face to face, it's kind of like you know each other already. It's a fun experience. A experience that's probably maybe even going away with people using AI commenting systems now. I'm like, have your bots call my bots? That's kind of how I feel like where it's going now.
00:00:43
Speaker
But hopefully it doesn't get to that point.
Dan's AI Experiments and Tools
00:00:45
Speaker
I'm excited to have you on, man. I'm taking the deep dive into AI. I've been started really drinking the Kool-Aid. I mean, I've been playing with AI consistently over the last year, and I'm using Cast Magic to automate stuff. And I have my open AI.
00:00:58
Speaker
pro account and I use it probably daily now, but only over the last couple of weeks. I'm like, that's it. It's time to roll up my sleeves and really, really dive into this and start learning about how people actually leveraging it and getting the most out of AI as it is right now or in the near future. For sure. So to kind of kick it off, I'd love to know like, what are some of your best workflows for chat GPT specifically around just productivity? Okay. Specifically around chat GPT, I think there's one piece
00:01:28
Speaker
Now, it all depends too if the person has a GPT Plus account or just using the free version of a chat GPT. Highly recommend if there's nothing else you would spend on, maybe cut your Netflix and go to the Plus version, although OpenAI has paused their acceptance of new Plus users just because it's overwhelming.
Creating a Business Co-Pilot with GPT
00:01:51
Speaker
But for me to show you how good it is, right? People are like, it's like, Hold on. No, no, don't give us money. Don't give us money. We're not ready. Which is crazy. Business says that don't give us money. Yeah. So if you do so and you can do this with normal chat, GPT, but it's a little you'll have to do a bunch of workarounds. But for me, the number one thing for any bit I feel this for any business is building your own co pilot.
00:02:18
Speaker
So when I talk about a co-pilot, like we've created a co-pilot for zero to 60 AI, the company that I'm co-founded, but we're talking about imagine a business mentor, coach, even investor that will be on your shoulder every day that you can ask questions to related specifically to your business.
AI as a Business Advisor
00:02:45
Speaker
So imagine that that ability
00:02:48
Speaker
24-7 and literally can answer things like, I don't know, like advice on handling clients, business development, how to charge clients, technology insights, how to use AI for internal business operations, finance, accounting, human resources, all these things. If you actually create, like we created a GPT specifically on that,
00:03:16
Speaker
Hold on, pause, pause. You're getting into the deep end real fast, Carl. Slow down a little bit. You're talking about, I got a question for that. It's question number four, but for people who are new to this and are like, maybe they fiddled with it, give me some prompts, give me some productivity stuff of like, okay, building a co-pilot, this is a bigger undertaking, but if someone's getting into the shallow one, what does that look like for,
00:03:44
Speaker
for it because I'm with you on the custom GPTs. I've only started digging into that. We'll trade a little stories here in a second, but give me some basic productivity stuff from the shallow end.
Effective Prompting for Business Strategy
00:03:54
Speaker
So let's say you can actually do this without building a custom GPT. I think the only difference is that a custom GPT has a knowledge base where you can upload information and where it can pull from.
00:04:09
Speaker
Now you can, if you don't have chat GPT plus you can actually, or what if you use other language models like Bard or Claude or Lama or whatever you're using, you can prompt your way to a, a copilot. So I know the, the immediate uses of AI would be.
00:04:32
Speaker
task to complete a task. And for marketers, a lot of it is content creation, sometimes some strategy, maybe analyzing financials or reporting. But if you actually think about what an LLM could do and
00:04:50
Speaker
the ability to, what kind of prompt you can give it. It doesn't have to be specifically about tasks. It can be about, hey, providing you that strategy, providing you the advice. So for example, I'll give you really simple prompt. This is how we start our, our co-pilot. You are a business co-pilot and strategist for our AI consulting firm called Zero to Sixty AI.
00:05:18
Speaker
you provide business advice, strategy, tactics on building our business. And then I lead with your world-class expert in business strategy and development, especially for startups and consulting firms. You are an expert in understanding the challenges and opportunities for small businesses and how they can navigate AI to make an impact. So that line itself sets up whether you're using free or a GPT sets the context of what this
00:05:50
Speaker
co-pilot will do.
AI's Role in Business Operations
00:05:51
Speaker
And then I add a couple little things like you have decades of experience, buildings, businesses and scaling them and understand the power of AI to fundamentally change how a business operates. Like that alone should set you up. And that's like just the starting point, but there's obviously a whole bunch of things you can add like context criteria and other stuff to build your GPT or your co-pilot.
00:06:16
Speaker
It's funny. I literally I'm using it's probably, I started with just basic prompts. Like most people do, especially if you're in content marketing world, you're like, Hey, like it's super normal. If you're a podcast thing to just take the transcript and use that to tell it to make a blog posts, social posts, tweet, tweet thread, like all that stuff. Cast magic now automates that whole process for you, but that's, it's pretty basic. And I've been doing that for awhile, only within the last, I'd say two months. If I started like.
00:06:42
Speaker
using chat GPT to help me think, which is like, in my opinion, like a different level of using chat GPT. I even did it this morning and I was like, what did I prompt it with? I'm like, can you give me some business models, even around this, this podcast and this market? And I'm like, what are some target markets for it? And it's, it spit out some target markets. I'm like, Hey, Google me, Dan Sanchez, marketer, and it'll find my website. Here's my resume based on my previous experience, which market should I go after?
00:07:11
Speaker
And it did a fantastic job of analyzing what I had already done and said, and remember talking about and telling me like, based on where you're at, based on what I'm seeing, you should probably focus on this industry. I was like, huh.
00:07:24
Speaker
Actually, that's really good. It helped clarify my thinking. So you're talking about using chat GPT to really essentially help you think. For sure. Which you don't have to build a co-pilot with. Or I mean, you don't have to get into custom GPTs even to do that. Nope, not at all. And I think there's more layers you can add on to it, right? You can give it your target audience, like who are you trying to reach. Yep. You can give it a description of your business, what your business goals are.
00:07:49
Speaker
And then that sets the context of it. And then give it some examples. You don't have to give all the examples of how you plan to use this. And the beauty of it is once you put this in, you'll now have a thread that you can continuously ask questions against it. So you've built this AI co-pilot for your business.
00:08:13
Speaker
Or it doesn't have to be your business. If you're a marketer working for whatever company, you can be specific to your industry, be specific to your team, be specific to a project. You can actually build multiple co-pilots on certain things. And the
Content Creation with AI
00:08:29
Speaker
awesome thing about a co-pilot is it can create you content.
00:08:34
Speaker
because it has the context of what you're looking for. So it's a really simple way to really use, to enhance the use of chat GPT going beyond just, you know, create certain pieces of content or accomplishing a task. So before we get into custom GPTs, tell me a little bit more about what kinds of roles you were having AI speak to.
00:09:01
Speaker
It depends on the situation. So like one example would be,
00:09:09
Speaker
We wanted to, this is kind of outside the marketing scope, where there is a submission and the submission creates a PDF and then that PDF gets turned into code and then applied. So instead of say, you are a marketer, the situation is more on business analyst
00:09:35
Speaker
more on specializing in, let's say, grant applications. So what I did is you are a world class business analyst specializing in grant applications for whatever industry that sets the context. And then you can add lick fillers like you've had decades of experience, you've had whatever else that you can put in. But that does set the context of it. But you can, from a role perspective, you can be anything just think about
00:10:05
Speaker
What role would this task or this action normally fall under and apply that? Does it really make a difference if you add the words world-class or you have decades of experience to the role? I mean, I can understand how the role you are X, give me feedback as X, but have you tested that? Like I see lots of people are saying it, but I'm like, does it really make a difference? Does it,
Custom GPTs and Practical Applications
00:10:31
Speaker
does it change it? Does it go to a different level saying world-class, you know?
00:10:35
Speaker
I've actually tested it more so on, you are a world-class, let's say, copywriter. You are a junior copywriter. There's a big difference with the outputs, right? Now, you may not, obviously from a copywriter perspective, I've yet to meet anyone who wants to put junior or intern copywriter.
00:10:58
Speaker
There are probably situations where you want it to be at a specific level. I just haven't really thought about that yet, but imagine you can set those levels, which is a power in itself. I'm like, why wouldn't I want world-class everything working on all my business? Exactly.
00:11:23
Speaker
I actually have an example for this. I just thought of it. Imagine if you put, like, let's say you're in school and you're like, I want to be a world class, I don't know, engineer who does this, but I need to write a paper. The output of that paper, imagine like, I am a graduate level engineer. I need to write this to very, like, that's how one way you can, I guess you can control the output of, you know, using AI.
00:11:54
Speaker
That's a good point. Actually, I can think of times, especially if it's giving advice to you. Sometimes world-class advice is just highly impractical to somebody who's in the beginner stages. You know, world, like if you're studying jiu-jitsu, it's a very nuanced sport and you want some feedback. I'm just kind of shooting since we all know like jiu-jitsu has belt levels. Getting feedback from a like third degree black belt, unhelpful some often.
00:12:19
Speaker
getting feedback from a purple belt, who's like three steps up from a white belt, probably way more helpful. Yeah. That's like a great use case and, and you can test it, right? You can always test these things, but it'd be fascinating to see what the, what the outputs are. So that's kind of an interesting case. I'll have to think about that as I go and get it to give me feedback. I can think about like what maybe a role, a step or two ahead might be more prompt, more
00:12:49
Speaker
accessible for me where I'm at now. So that's cool. Let's dive into custom GPTs, only something I started playing with maybe two weeks ago. I discovered that it was there and it blew my mind. I knew the applications as soon as I saw it because I've been working at tech companies and they're all about uploading documents, just even for chatbots. You have support chatbots and they're like, well, you can train the chatbots on your operations and your docs so they can provide better answers to your customers.
00:13:17
Speaker
obvious use case, right? That's where most people are spending their time on right now.
00:13:21
Speaker
Um, the funny thing is even as I was working at that company and their training, they're doing chat bots for like higher ed institutions for admissions, lots of details to know, even humans don't know all the details about all the admissions details. And it's really helpful for a chat bot who knows the, all the schools ins and outs, you know, the whole product catalog, which is usually how schools operate. But even as we were development, the HR director said something different. She's like, man, you know how many people ask me about like HR things that I answer over and over again.
00:13:50
Speaker
And it really opened, when she said that, I was like, man, this has so many implications to be able to upload information and be able to ask it questions based on that information, because this is the new search engine. Um, to be able to not just answer a question that's hidden in a doc somewhere, but actually do analysis on it, actually create content from it, actually have the con full context.
AI for Data Analysis and Marketing Insights
00:14:13
Speaker
It's a game changing application. That's what chat G that's what custom GPTs allow you to do is to create your own custom.
00:14:19
Speaker
GPTs your own custom chat GPT based on the information but tell me Howard like what are the different ways that you're using it? Because you can upload documents to it, but you can also train it. Mm-hmm. Yeah there's so one of the things that
00:14:35
Speaker
You know, I've been using before and actually client wise, you know, setting those up were custom AI chatbots, like tools like chat base, retune, Dante have actually built these already. And there's differences between custom GPTs and those chatbots. There's pros and cons to both. I think from a GPT perspective, the big one right now is that only chat GPT plus users can engage with them.
00:15:04
Speaker
And a custom AI chatbot, like a chat base, you can embed on websites and embed on pages, which then allow a broader use. But the one thing that I feel... Let's start with basic first. What you're talking about is data retrieval, right? So you ask a question, it retrieves data, it provides the answers.
00:15:25
Speaker
It's helpful. There's a lot of use cases for that, but let's take that one. Let me give one use case real quick. Cause I just did it. I just got elected to like my HOA board. Okay. New neighborhood. They needed a board. I was like, sure. Why not dude? I don't, you know how long the, the bylaws are and I don't read legalese.
00:15:44
Speaker
I'm like brand new, I'm a marketer, I'm a business guy, but I don't read legal. I uploaded the whole CCRs, like the whole bylaws, JAT GPT, made a custom GPT. Now I just ask it what I want. I'm like, wait, what's the process for this? What are the rules on X? How long before Y? It's way, and it interprets it for me. It's way smarter. Of course, I have to go reference it just to make sure, but like, yeah, it's way easier for me than to go and
00:16:08
Speaker
look at the table of contents try to decipher what the heck it's saying it's much smoother so that's the one application i found just for data retrieval for myself and one thing that you have to think about too because you're using with a custom gpt you're now using like the the advanced versions of gpt4 and when you create them you can have
00:16:30
Speaker
you have code interpreter or advanced data analytics, whatever they term, that it can analyze structured data like CSVs, reports, and so on. But the one big piece that I really love about it is you can use GPT-4 with vision.
00:16:49
Speaker
So you can upload images on a variety of things. So Black Friday, simple example. Now, you can use this with GPT-4. So you can download the app. If you have GPT-4, you can use Vision. What I've been doing is taking pictures of price tags.
00:17:10
Speaker
So I took a picture of price tag and said, okay, is this the best deal for, I don't know. I think I was at furniture store, wherever that was for this type of couch. I took a picture of the couch too. Um, best deal in Canada. Give me, give me alternatives. So analyzes the price analyzes the couch, gave me a bunch of like alternatives for that couch and I'm like fantastic.
00:17:36
Speaker
So now I was looking at the different things. So I mean, it's doing multiple steps. It takes a picture, analyzes the picture to understand what the heck this is and the price that you're talking about, then goes and uses because it's attached to Bing. So it did a Bing search, research options, and then did the analysis to figure out what the better options were based on what it did. Yeah. It's a lot of steps that it's doing really quickly. Oh, for sure. And like from a marketing perspective, think about the reports that
00:18:04
Speaker
Like I just took a picture of Google analytics for GA for like a page. I took a picture of that and I said, give me the top 10 insights. Give me the top 10 recommendations based on just the picture. And it gave me everything about it. So then I took a picture of our quarterly reports. That's like, it's a pretty significant, pretty big, you know, spreadsheet.
00:18:31
Speaker
broke everything down. And then you can just have conversations about specific things like source, pipelines, and all that kind of fun stuff. But to me, that is the one untapped piece that I think people are missing out.
AI Use Frequency and Technological Advancements
00:18:46
Speaker
Again, chat GPT plus, but that ability alone, I feel it's even stronger than code interpreter because it can analyze what you have on screen rather than trying to figure out what's the
00:19:01
Speaker
what the fields are and all that kind of stuff. It just takes that. And there's so much use for that, that I think we're just starting to at the moment. I mean, talking about productivity, I'm like, dude, marketers aren't going to have to write executive summaries or reports anymore.
00:19:16
Speaker
Or maybe like AI is going to be writing all of them and we're just going to be looking over it cleaning it up Maybe shortening it down exactly like kick like think about how much time we freaking waste on this every month Just writing reports so that someone above us can have an understanding of what we're doing and what we're behind or where we're ahead like AI is good to do this yeah, and I think the the really the big piece that I
00:19:42
Speaker
we're missing out on is just thinking about ways that you could use it, right? Because at the moment there's, yeah, there's prompting courses and there's a whole bunch of how to use it for creating specific tasks, how to use it for making money and so on. But this is where I feel there's a big gap and it's widening is that
00:20:06
Speaker
There's people who use AI frequently versus people who use it infrequently or not at all. But because of the pace of change, the people who are using it frequently get upgraded with hate. They know this came out, this came out, this came out, this feature came out and just keep upgrading their skills. But the people who use it infrequently or don't use it at all, like they don't know that these things are coming and they may have just used
00:20:34
Speaker
Let's say a chat GPT a couple of times are like, the answers aren't very good. I'm not going to use it or I use it sometimes. I think like there's that big gap and I feel the gap is because everything's moving so fast. All these tools are coming out. There's no one actually using to educate.
00:20:55
Speaker
And then how to apply it from a real business sense. Like there's a lot of hype. There's like, here's a new tool, blah, blah, blah. Like, Hey, how can businesses actually use it in a reliable, scalable manner? And I think that's like the missing piece. And that's where like there's.
00:21:13
Speaker
That's where we can focus on. And I feel like that's a place that with your podcast and other people can really dive into.
Practical AI Implementations for Business
00:21:21
Speaker
It's like, hey, how can you help people, businesses use it today? Not just prompts in a bunch of content creation, but there's actually really cool, tangible things you can use it with.
00:21:32
Speaker
That's what I'm hoping to do. It's why this podcast now exists because I didn't find what I was looking for. So now I'm like, well, I'm going to have to go and figure it out. I'm going to have the conversations myself. Cause there are a bunch of, I finally found a bunch of YouTubers that talk about AI, but too often they're talking about theoretical, even LinkedIn people. There's very practical stuff. You're posting a lot of practical stuff. There's a lot of practical stuff out there.
00:21:55
Speaker
But oftentimes a lot of the gurus are talking about theoretical applications of like, Oh look, it can do this really cool, fancy thing. That's not even close to practical, but it's kind of cool to think about what this 1.0 version could look like two years from now. I don't know. Movie making. Whoa. It's not even close. It's not practical right now. It's stupid.
00:22:16
Speaker
I feel like a lot of the bigger creators like AI or AI could do this with this new feature or I created this tool that does all this amazing stuff. But if you think about it, I'm like.
00:22:29
Speaker
No business is going to use that because it's not reliable. It kind of does it half the time. And then my favorite ones are, oh, it's going to kill this, replace this, kill them. Either you're really, really misleading or you're really ignorant of how businesses actually operate because no business is going to use that and replace all their people. It doesn't happen like that. So I think that's where the hype comes in. A lot of people get like, ugh.
00:22:58
Speaker
But hey, what are some actual things that you can do? Again, beyond just prompting and using chat GPT. Right, right. Tell me more about what you're doing with your, what you're calling a cope, making a copilot. Like now let's dive into that. How are you? What, what kind of, what are you using it for? What kind of documents are you uploading and what instructions are you loading into this thing? So.
00:23:22
Speaker
Let's go. Actually, there's one marketing use case that I think is amazing is I created an ABM co-pilot. And what this does is it's a little advanced in terms of what you can use a knowledge base with, but
00:23:42
Speaker
It's initial tasks like, so we'll break it down and what it does. First thing it asks you about, you know, I think seven questions on what your like your target prospect, just like any ABM account based marketing, what the information about the account, right? It lasts you six, seven questions about it. Then at the end, it will say, Hey, can you upload any supporting documents?
00:24:08
Speaker
whether, whatever that is. There's actually, now custom chat GPT can't upload website content, like it can't train on it, but there's actually a site that you can use where you can actually be, you can, it scrapes the pages, turns it into a JSON that you can then upload into a custom GPT, hence creating profiles and stuff.
00:24:34
Speaker
Um, I don't know. I've just been using it for websites. I'm not sure about LinkedIn. You're going to have to test that out. That'll be something to test out. But if you had structured data with all your, your account list is all you really need to be in a CSV. Oh, for sure. Now, after it takes all that information, it provides a couple of things.
00:24:59
Speaker
It provides the 10 potential marketing strategies, you know, tailored on the information from the project. I give very specific instructions, right? Like provide strategies on how buyers buy in 2023, go beyond the typical marketing strategies, take unique approaches to this particular account, you know, tailor the strategies to be deeply personalized,
00:25:25
Speaker
based on the information provided. And then I'd say refer to products and solutions in your knowledge base. And that's where the website scraped is coming in because then it's referring to the products and information there. So it got that. Then I also have it do 10 potential messaging tailored to the prospect. And this one's
00:25:51
Speaker
The neat thing about a GPT versus a custom AI chatbot is you can interact and engage directly with the knowledge source. So for example, here I said, make sure to use the writing guide in the knowledge source when writing any content. So whenever I type in for this GPT, write me an email or whatever, it would actually refer back to the writing guide.
00:26:21
Speaker
to ensure that it matches the brand tone or whatever that is, the voice and style. Yeah. So, and I think like in addition to this custom GPT, I actually had hotkeys and I got this from another GPT called grimoire where you add hotkeys and then it executes instructions.
Advanced AI for Business Strategies
00:26:44
Speaker
So one of the hotkeys is implement, you know,
00:26:49
Speaker
report task number one from the knowledge base. And so report task number one in the knowledge base is essentially a Word doc that is a set of prompts. So when I say execute that, it looks at the knowledge base, looks at that name, and then executes that task in there. And then it does it. So that knowledge task was online search, look for reports,
00:27:13
Speaker
filter reports gave me the top content of the latest report from this company. That was the research task number one. And then another one is implement competitor analysis. And so it would execute a competitor analysis using Bing and any other offerings that you've uploaded as well. So that's a full co-pilot that
00:27:42
Speaker
a marketer could ask against, continue to ask, gets messaging, gets base strategies, and can upload more information or just get strategies, recommendations, advice on one particular account. Imagine the potential for that.
00:28:04
Speaker
Yeah, it's huge. I, it's funny, I find using cheap, uh, custom GPTs is a lot like training interns on how to do things. Right. Like I used to run in part of a different organization. I worked for two organizations. I ran an internship program and I'd have 10 new people come in and every year and they'd be working 15 hours a week, uh, for a whole year. Uh, which means I was retraining 10 to 12.
00:28:29
Speaker
interns every single year which when you're doing that and they were doing Important important jobs any jobs I could figure out how to train them quickly enough to get value out of them And then then learn some stuff launched a lot of marketers in their career this way But I would train them how to write good blog posts and that's the only thing they would do or how to manage social media It's the only thing they would do or how did how to just do certain styles of video but when you're doing that you have to have a certain amount of training on What's your job description?
00:28:59
Speaker
who are you and what do you do? What's the mission of the organization we're aligning this to? What is the reference docs for how to do this that you're going to get trained on, you're going to read through it all one time, and then you'll probably have to go back to refresh every once in a while. And then what are the standard operating procedures for each thing you have to accomplish within your role?
00:29:19
Speaker
I had to get crystal clear on it because they don't know anything. Like they're coming right out of high school under this stuff. I was getting them like not even post-college. I was getting them pre-college. Yeah. And I'd vet them a little bit, right? But I had to train them on how to execute step by step. I'm like, okay, when writing a blog post, you're going to get, this is the information coming to you. You're going to have a keyword. You already have the title pre-written. Here's an, I'm going to give you, you're already going to have an outline already pre-written for you. Your task is to go and learn everything you can about this within two hours.
00:29:50
Speaker
Then, based on the knowledge and training you've gotten, you're going to then write this blog post in this specific way, starting with an intro that opens it up broad and has a story, then moving on to the snippet optimization, then moving on to the transition to the next most likely question they're going to ask, and then to this section, then to this section. And you're just going to repeat this step over and over again.
00:30:09
Speaker
You train cheap each ease the same way You have reference material and training, but then you actually break down the instructions As far as you can go because it will execute those instructions step by step if you let it But you have to it's so funny how smart it is, but how stupid it can be at the same time you notice that Oh
00:30:31
Speaker
I try to get it to just ask me questions.
AI in Creative Processes
00:30:33
Speaker
I'm like, only ask me one question at a time. It asks like five questions. Yeah. It has a hard time counting for some reason. Yeah. I don't know what it is. Counting it is not good at like counting words, counting characters, counting questions. I don't know. GPT is troubles. The one thing that I've always, so there's a couple of things there too. Like, for example, like, you know, you developed a super long prompt, but
00:31:00
Speaker
The models are always changing. They're always updating them. So sometimes your prompts don't work. Like it's a term called prompt drift. So it's like, Oh, it doesn't work anymore. So you have to, you always have to constantly update it. The one piece I love putting in though, is it's called few shot learning. Put it to me. That's just a fancy word for putting examples. So you just, Hey, if there's a specific way you want to write something or create something or advice something,
00:31:27
Speaker
Put that in as an example, one or two examples. It fixes a lot of the stuff. The moment I put an example, it's exactly what I want.
00:31:38
Speaker
I've done all that a little bit. I started making a, I made a custom GPT to automate a process I was already using that was so step-by-step, but it was, you couldn't program at it. You couldn't program it because it was a little bit too intuitive, but GPT could handle it. It was a fantastic book on how to name things. It's probably, I've read a couple of books on the topic when I was naming a lot of podcasts working for Sweetfish. And one book in particular was fantastic called
00:32:05
Speaker
Hello, my name is awesome. Fantastic book. It had one naming method that was just so good because it was so easy to execute where you'd go and come up with a bunch of words that were related to the thing you were trying to name. Find rhymes with that word, then take the rhymes of the words, find movie titles with it in, or find the rhymes, find movie titles, and then, or idioms, and then swap out the original word. So like I came up with,
00:32:36
Speaker
Or was it Mike club was a community we were building for sweet fish. Yeah. Cause you can hear a fight club, but it's cause I would took Mike rhymes with fight.
00:32:46
Speaker
And then you swap the word, you find, oh, fight club is a cool title. If I swap out the original word Mike, then I have Mike club. And I'm like, Oh, that's a good name. See how the process works. But it takes a while because for every single word you have, you have 10 different rhyme combinations. And for every rhyme you have to come up with, you can find 10 or 20 or 30 different idioms or movie titles. And then you're swapping out the words and then run it and analyzing it to see what sounds good. Chat GPT can run that effortlessly though.
00:33:15
Speaker
Have you made a custom GPT for it? I'm still engineering the prompts to give it exactly, but it runs. It does pretty good. Um, it's so many words to walk through though, that I had to break it into multiple steps. So I'm like, Hey, stop at this process and wait for me to say next.
00:33:35
Speaker
Yeah, because again, there's a certain amount of character there's a character limit to how much it can give to you in each Each output and because it's doing so much process in between each one I have to like break it up into steps, but it's getting pretty close. I'll send you an example I'll send you for
AI Prompting vs Traditional Search
00:33:49
Speaker
sure. Yeah, I'll let you play with it. You can see how it works Yeah Well, you just you just touched on one of like the the most basic Not the most basic like a really good technique is chain of thought right? So I think like
00:34:02
Speaker
If you notice, whenever you prompt anything, well, chatbot, chatgbt, whatever, it immediately starts answering. But in your prompt, you can always put, you know, take your time. There is no rush. Think through the step by step.
00:34:20
Speaker
And whenever I put that piece in, it really helps sometimes because it doesn't immediately go into the answer. It'll still put the output up pretty quick, but it gives it more options to actually break it down, like you just said, one step at a time rather than trying to answer it all at once. And so if you're finding
00:34:46
Speaker
Sometimes your prompts are like, Hey, this doesn't make sense. Or it's just too fast or it generates stuff that like is kinda right, but not really try to put that in think through step by step. And a lot of times it could definitely help with the output. It's so funny because sometimes you feel like you're dealing with a machine and therefore I can get exact stuff, right. But it feels, it feels more like you're coaching a human, a very,
00:35:12
Speaker
I don't know how to put this, but it's like a genius, an awkward, an awkward 19 year old genius who lacks common sense. That's chat GPT. Yeah. You know, if you're coaching something with like all the knowledge out there on the internet and yet still has a hard time making common sense decisions. Yeah. You're trying to get it to think through that. For sure. For sure. For sure. No, I, I found that too is just the way that, you know,
00:35:39
Speaker
I don't know if you've found this, but I think for all of us, we've been trained to Google. Just put a couple words and it comes out with something. But the thought process behind using a large language model is a little different because you're right, it's like, think about a teenager with
00:36:01
Speaker
this vast array of knowledge that you have to kind of help maneuver in a certain direction. And that's why you put like act as whatever, give it some context, give it some criterias on what kind of outputs it does versus just trying to put in one or two sentences and looking for the output.
AI in Team Communication and Marketing
00:36:21
Speaker
It's two very different
00:36:24
Speaker
There's a lot of talk about prompt engineering, but it's more of understanding there's a difference between one or two lines versus, hey, giving it actual context to what you want it to do. Have you gotten anywhere in the productivity space and business productivity marketing where you're starting to automate more stuff? Yeah. That to me is
00:36:49
Speaker
that is more applicable or accessible than GPTs. So for example, I love using these three examples. The first one is a lot of companies have some sort of tool. An example would be Gong, where they're collecting a lot of customer information, like collecting on calls or discovery calls or demos or whatever. And those are great.
00:37:18
Speaker
But who has time to actually listen to them and process specifically what you want out of it? So what I've developed is there's a tool called respell. And all I do is either drag and drop the audio. Sometimes the audio is like an hour long. Just drag and drop the audio.
00:37:42
Speaker
using a longer prompt, figure out like, hey, what specific things do you want to get out of that call? And then in about a minute or two, it'll spit it out. But what you can do with the respell is connect it to either Zapier or it has native integrations to Slack,
00:38:02
Speaker
and then pump that out into Slack because I think one of the challenges for customer insights is how do you get that insight to a broader range of people, not just yourself? You can put that on a Slack channel, you can put that in a Google Docs. People don't have to listen to those customer insights, they can just see
00:38:25
Speaker
in a couple paragraphs in bullet points, hey, these are the specific things that let's say product marketing needs, but you can also create one for demand generation. You can create one for creative, whatever that is, you can tailor those outputs based on your prompting and what do you want it to look for on that. So in one to two minutes, so like, for example, like every couple of weeks, one business would be, they would just
00:38:53
Speaker
load in, I don't know, five, six, eight calls, get the individual summaries, and you can take it even one step further and summarize the summaries, summarize those eight into here's the trends that you're looking for. And you can do this all through drag and drop copy paste,
00:39:13
Speaker
I would say 10 minutes to get not just individuals, but those summaries of summaries. That's just like one awesome use case that you can now automate with human, with a little bit of manual, but the little bit of manual is just control C, control V or drag and drop. And to me, that's like a super easy way for your whole team to see, to actually understand your customer.
00:39:41
Speaker
Well, I mean, it's automating a lot of steps. You still have to kick it off manually, but when you upload it, it's transcribing it and then using AI to interpret it in the right way and then auto posting it through Zapier to Slack. So like you've automated four steps in the process there. It'd be great if it was also hooked up to the right call system and then tagged appropriately that it would just kick over there automatically, but we're getting there.
00:40:02
Speaker
It's getting getting we're not far away. Yeah, I feel like most AI automations Are you usually still have a manual process somewhere? Mm-hmm, and they're not fully automated which even with this podcast I'm trying to get it so it's fully automated because I don't like it This post this will be turned into a blog post
00:40:19
Speaker
I don't like that I have to manually drag it over to cast magic and then turn it into a blog post and then copy and paste that over to the blog. I'm slowly automating more and more of this whole process because I'm lazy and I, if I have to do something the same thing twice, I'd rather machine do it. Well, like the thing that, you know, there's another automation that I love doing is, um, you know, like there's a show that I do and it's about 30 minutes, 40 minutes long.
00:40:46
Speaker
And the outputs are always the same thing. A blog post, LinkedIn description, YouTube description. So right after, same tool, respell, take the audio, drag and drop, and then it outputs one, two, three, all the things. Now.
00:41:04
Speaker
It takes about a minute or two, but to set this all up takes about an hour and a half. For me it did because each, every single output I had to actually individually create, like test and create. So the output is exactly what I want. And there's a little trick too, where you can kind of train the bot. So let's say it outputs a YouTube description. One, you know,
00:41:32
Speaker
There's no way you just put the YouTube description as is. There's a little bit of human review. After you review it, let's say you made a couple of changes, but this is exactly how I want it to look. Just copy and paste that into the prompt as an example. And then every single time you can technically train the outputs to be more closely to what you're
00:41:55
Speaker
what you what you want to eventually put out. And so like the blog post was like that. The LinkedIn post was like that. The YouTube description was like that. So what used to take me like two, three hours plus whatever time it took to create the blog post now takes a minute or two. And even with chat GBT, it still took me like 20, 20 minutes because I'd have to find the thread, copy the transcript, all that stuff. I don't need to do it anymore. Just drag drop and it's there.
00:42:26
Speaker
And now I have the content ready, and it took two minutes total to do that. So I think these are the kind of things that as a marketer, hey, could you think about the automation of this?
00:42:41
Speaker
Yeah, like you can almost just do an audit of your existing processes, manual processes, automated processes, and what steps can be enhanced, improved, or were missing because of a lack of ability to tap into something like AI that you weren't doing before.
00:43:00
Speaker
There's a lot of different things going on that can now be improved because there was a limit. There was just limits to marketing automation before for sure. He's going to fill a lot of the gaps of what were possible because you didn't have the data and constructually structure it without great, great pains and a massive dev team to be able to do it before.
AI Agents and Automation
00:43:16
Speaker
Yeah, you can. Yeah. Yeah. Curious. Where do you think?
00:43:22
Speaker
Like I don't want to get too far out in the future, but what do you think's like on the near horizon for AI? Like it's, you're pretty confident this is going to happen over the next year, max, like within the next two years, but generally within the next year that you're most looking forward to. Uh, the big one that I feel that like a lot of companies are solving for is agents. Where if you look at what AI can do now, it can do tasks, even with automation, it does tasks. It can do this. It can do this. It can do this.
00:43:53
Speaker
So I like Dharmesh Shah, he has the founder of HubSpot. He had a great definition of AI, what is the definition of an agent? A definition of an agent is it can accomplish goals, not tasks. So if I put in, the goal is I want a marketing report every quarter on this.
00:44:17
Speaker
It will then go out and create all the tasks, all the prompts, everything you require for a quarterly report. So you don't have to create the reports. You don't have to upload anything. It'll do it all on its own and then ask you, hey, I need these type of documents.
00:44:36
Speaker
and go off. I believe that there were some, the resemblances of that. There were tools like Cognisys, God Mode that were trying to do that, but they only worked 60% of the time, 50% of the time. But now I think that's where, you know, you could see inklings of that with custom GPTs, but I feel that is the big piece that would be coming out. And
00:45:04
Speaker
I think a lot of people have been talking about it. Bill Gates has been talking about it. Sam Altman has been talking about it. They're all talking about it. And that always whenever they talk, when there's a lot of talk, you know that they're trying to solve for that. And you can start seeing a little bits and pieces of that.
00:45:21
Speaker
Yeah, I have one of my clients that I work with right now is, is solving that on a sales negotiation coach level where it's like it's coaching, but it's way more in depth. It essentially taps into your Salesforce org. Ask prompts you with the right questions to give it the right data.
00:45:36
Speaker
Yeah, and then takes into account your business. So it's trained on your business, your deals, what you can and can't do the priorities of your executives. But then you're giving it like all the stuff that your buyers are interested in, what are their requirements for large enterprise deals? What are their their their key stakeholders, all that kind of stuff. And then it starts actually prompting you like this is what you need to do in this next call. And then starts actually like here, the here are the it actually creates for you like the
00:46:05
Speaker
I don't know, the offers that you're going to make to the company. Here's good, better, best, and taking everything into account that you've been talking to with these companies.
00:46:13
Speaker
to help actually create original value for it. So Close Strong, that's why I even got kicked off this podcast, because I saw what they were doing with AI. I'm like, this is literally like having a pro negotiation coaching your pocket. Yeah. It's beyond a consultant, because it's able to get, well, I guess it's like a consultant, as if an experienced 30-year plus negotiation consultant was always by your side, helping you get each step of the process right, so that you can close at a higher percentage.
00:46:42
Speaker
oh yeah your tools gonna make money because somebody other people money and it's gonna be like of course we all know that rapid feedback is the key but no one can afford having that person do more than a training session and we all know we don't let we don't we don't retain training materials correct yeah so having the
00:47:01
Speaker
You go ahead. No, I'm excited about that future. But also, it's a little scary, right? Because what does that mean? Are we ready for agents to actually start doing things for us? There was a demo I saw way back, I think it was in February, about this company called Adept, where they would
00:47:24
Speaker
do the research for your real estate for yourself.
Current AI Capabilities and Applications
00:47:27
Speaker
So you'd put like, hey, find a home that is 600,000 or 700,000, 1 million, whatever it is. And it would find it. It would go to websites and would just go find those things automatically without you having to go to the website, just goes to it. So it's like, that's coming and you can like, they release their, Adept just released their like beta version. And it's not,
00:47:50
Speaker
as that, but it's like it can go to websites and start searching. So, you know, think about like in a month, two months, where does that go?
00:48:02
Speaker
Yeah, I mean they're gonna be doing they're gonna be doing Not simple tasks, but more complex tasks for you like an actual personal assistant would for sure I mean we've all heard like the google call from like man. This is a while ago five six years ago. Maybe yeah where google's ai assistant was making a hair Appointment or something for a client and it sounded real it talked like a human. Yeah, it was just calling up and it was every it blew everybody's mind
00:48:24
Speaker
I just don't know if that was nobody nobody could use it I mean because it wasn't really available It's kind of like a nice little demo of like it was it might as well have been a sci-fi film because yeah We never actually saw it but now we're actually getting to the point where hey Like these are gonna be real soon and they're gonna be calling people for you They're gonna be doing a website research and actually going out and executing on your behalf. Mm-hmm. Is it scary? I don't know I
00:48:48
Speaker
Like when you think of like long-term implications and that's where most AI shows get like really like utopianism or apocalyptic real fast. I'm like, screw it. I just want to implement what's good now. What's good right now for sure. Yeah. And right now somebody doing research, basic research for me on the web sounds really nice. So I'm buying into that.
00:49:07
Speaker
Carl, thank you so much for joining me on this episode of AI Micro Skills. Where can people go to learn more about what you're doing with AI? For sure. So our website's called zero260.ai, and just what I talked about here, our goal is to really help businesses and people. You tangibly use AI today, help educate them. So we try to go beyond the hype, and hey, what can you do today with it?
00:49:32
Speaker
connected to your business goals and you know, increase productivity, enhance your efficiency and just help you do what you need to do for your business. There you go. Don't fall behind on AI, contact Carl, let them help you get ahead faster because we all know that having someone to do it with us can always accelerate the whole process rather than you have to hit every roadblock there is along the way.
00:49:56
Speaker
So reach out to him in his company and remember AI feels complicated at first, but it's like everything else. It can be broken down into tiny steps and pieces. So make sure to stay tuned to subscribe to this podcast and our various social channels at AI micro skills.com in order to start catching up with the future.