94 | No Code AI Automation Framework from Concepts to Daily Efficiency with Nancy Bain

0:00Hello, everyone. Welcome to another live episode of the Leveraging AI podcast, the podcast that share practical, ethical ways to leverage AI to improve efficiency, grow your business, and advance your career. We have an amazing show for you today. And I'm going to start at the top of this by saying that many people understand that in theory, they can automate stuff in their business using AI. The people who went a little deeper understand they can build serious automations using GPTs within ChatGPT in order to achieve that. The vast majority of people don't have a clue how to do that. And from those of the people who actually are doing this and creating their own GPTs, most of us, myself included, are winging it.

0:46Meaning I know what I need to do. I've had experience and I'm walking through the process trying to make it work. However, like anything else in business, the right approach to this is to start with strategy and then develop a process and then go and do the thing that you're trying to do. In this particular case, creating automations that really move the needle and are efficient and driving value in your business, which is exactly what we're going to learn how to do today, step by step, by no other than Nancy Bain, who is an absolute guru and AI and business automation ninja. I'm not going to tell you how many years she's been running her business, helping businesses build efficiencies around different things.

1:31But she has been running her consultancy company, Supernova Media HQ for a while, despite the fact that she's still 25. So this is going to be a very detailed thought after step-by-step process on how you should approach building automations together with GPTs to really move the needle and get stuff done in your business from a strategic approach process perspective. And we're going to dive at the end into the actual implementation and the outcomes that you can gain. So if you're looking to do anything like this, and if you're in anywhere in business, you should, this is going to be a fascinating conversation for you.

2:12And hence, I'm really honored and excited to welcome Nancy to the show. Nancy, welcome to Leveraging AI. In the next few years, AI technology will change our world dramatically. Whether you are a business executive trying to catapult your business forward or just somebody who refuses to be left behind and want to advance your career, this is the show for you. I'm your host Isar Maitis, a serial entrepreneur and an AI enthusiast. You'll hear invaluable practical tips from innovative business leaders, AI practitioners, and some of the brightest AI minds in our world today on how you can leverage AI in ethical ways to advance your career and grow your business.

3:02It's good to be here, Isar, and I meet you at the intersection of that person that you, that user that you just described. Absolutely. Today, I'm going to show you that the shortest, most efficient distance between A and B sometimes involves a methodical framework and strategy to make happen. And so think short-term pain for long-term gain. And I don't mind saying, yeah, I don't mind saying how long I've been doing this. I've been a bootstrapped solopreneur for over 20 years. So finding the shortest, most efficient distance between A and B is my absolute jam. And so today I'm excited to share with you one of the tools in my toolbox, the custom GPTs and how I manage my builds and the finished product and including the limitations thereof, which we'll talk about.

4:19So I've got a bunch of screens to share with you. We're going to start off in my presentation and then we'll move over into my Notion build, the back end of, and the back end of Chant TPT. So how does that sound? Sounds amazing. For those of you who are with us live, either on LinkedIn or on the Zoom call. So first of all, thank you so much for joining us and bringing stuff time on your busy days to actually be with us here. I'm monitoring both the comments on LinkedIn and the chat on the Zoom. So if you have any questions that you want to ask, if there's anything you want to contribute to the conversation, please do that.

5:01I will add that into the conversation and try to find a time to ask Nancy whatever questions that you guys have. For those of you who are listening to this only at the recording as the podcast or the YouTube channel or on LinkedIn or Instagram or wherever you're consuming this, come join us. We do this almost every single Thursday at noon Eastern, where you can come and participate in these really amazing conversations with people like Nancy. Nancy, stage is yours. Take us through your amazing process. Awesome. I'm just going to share my screen here. So as I mentioned today, we're going to talk about custom GPTs.

5:40And first question, what are they? In a nutshell, they are personalized versions of ChatGPT. They are designed to streamline our workflow. I have 21 of them, everything from a 15-minute meal helper to a workout buddy to a schedule optimizer, long-form content. They're designed for a single, singular task, and that's important to keep in mind with regard to their limitations. So let's talk about some of the use cases. I've mentioned some of the ways that I'm utilizing custom GPTs. Today I'm going to be showing you my Gumroad Guide Manager. The this it helps me manage my Gumroad tutorials. I think I've got close to 40 tutorials over there now.

6:45And so this saves me my Gumroad Guide Manager GPT that's enough. Certainly helps me save a enormous amount of time in managing that. So what are some of the pros? We've already talked about time saving, and they boost performance and ROI in our campaigns, and they improve my decision making. I love to screenshot or download my analytics from Google Analytics, from Gumroad, from Beehive into ChatGPT and have them help me drive data-driven decisions. With Custom GPTs, managing expectations is key. Okay, they're not silver bullets, and they're not a generalist. And with that in mind, I polled my LinkedIn community and asked them their experience with custom GPTs.

7:50And you can see that only 10 % of people are finding that they're responding perfectly. Please keep that in mind when you're using them. I've talked a lot about prompting frameworks and the recipe for the perfect prompt, but really nothing replaces iterative prompting in my experience. And the cons, I mean, no technology is flawless and GBT is going to produce content that isn't accurate. One of the things that I wish was different is that they're designed to solely exist within the chat GPT interface. So you can't download them or add them to your site or host them anywhere. And also, when you're measuring the success of your GPTs, some of those KPIs aren't quantifiable.

8:45So you need to keep that in mind as well. I want to pause you for just one second about this particular topic, both the pros and the cons, and give my two cents on this. It's a lot about learning how to do it right, and that's what we're going to spend the session doing. Once you finesse the process, it's actually following instructions, I would say, 95 % of the time. When in the beginning, I probably would have answered most of your audience saying, sometimes it follows instructions and sometimes it doesn't. And so a lot of it is figuring out how to do this. But now I reuse the same structures again and again that have proven to work.

9:25So that's how I go around that part of the problem. The other part of the problem, and we mentioned that briefly in our call before this interview, which is there is a way to actually build a similar process, meaning they don't call them GPTs, they call them assistants, but it's the same thing, into the API. and if you do that, then you can have access to them anywhere because they're connected to the API. So now you just need some kind of a front-end user interface for them, but then they could be on your website, they could be on a chat, they could be a back-end for a Zapier, like whatever it is that you want to do with this, you can use them in a different way.

10:05I don't want to open that topic because we can do another whole episode about that. I would love to do. yes please to another whole episode because well done i'm gonna i'm gonna hold you accountable for what you just said i was hoping you're gonna say that but all i'm saying is people need to know that there is a way to get those gpt's out of the chat gpt universe and into the rest of the world it's a little more technical a little more complicated we're not going to cover this today but there's a way to do that exactly and that's what we offer at supernova media is those custom solutions. And then you can involve your clients and their use.

10:46So they can start to service your clients and you can streamline all of that. And so that's where it's at. And I said earlier, I've got 21 custom GPTs and one hell of a personal assistant that was created using the APIs. It's not for singular use case either. So let's see. Absolutely. Thank you for that. So I'll look forward to another show on that. So custom GPTs, there are AI actions within those. You can connect to Zapier, which further connects to 6 ,000 other apps. And you can use them for things like lead generation, post-purchase follow-up. It can automatically update your Google Sheets with sales data.

11:42So lots of automations are possible. Not all of my custom GPTs use automations. In fact, the ones that I have that I'm going to show you today do, but not via the custom GPT interface. This represents my custom GPT framework, and it will make more sense once we go visit Notion. And let's talk about Notion because Notion is one of my top three loves, and it is where I begin all my builds. I honestly don't know how I ever functioned without notion. I love it for a structured development process and centralized documentation. It's great for collaboration. But when I built it, I didn't realize the last bonus there that I love the most of all, it's portable now.

12:44I've got all my files, they're beautiful, all in one place. And I recently purchased a subscription to Gemini. and their equivalent to the playground and the API is the studio, their AI studio. Now my files, they're beautiful. I'm going to show you here right now and I can take them there and I can work with another LLF. So I highly, I love it for that. And it's funny because this quote right here is how you began your talk today. And that is, did you know that the most successful digital marketers are often the ones who master the art of deployment? And with that in mind, I'm going to go show you my notion build for my Gumroad Guide custom GPT.

13:39Now, I'm going to pause you just for one second. For those of you who don't know Notion, it's a universe that kind of combines task management and documentation with all their history and so on. But you can do this either with parallel tools like Monday or ClickUp or Jira or Asana, or you can just keep a Google Doc history that can achieve similar things. I highly recommend using one of the tools that is like Notion. And I personally use ClickUp, but it's very similar in the things that it can do. A lot of people use Monday. It's very similar in the things that it can do. But having something like this when you're planning a project, and I think that's where Nancy's process shines.

14:24She literally created a very well-documented process that she's going to share with us right now on how to create these DPGPTs versus, as I mentioned, just winging them. So with that, let's dive into what you have in Notion. Great. Right. And just Notion has a great free version. I started in the free version and until I brought the team in, the free version worked great for me. And if there's all kinds of free resources, in fact, if you were to go to Gumroad and type in Notion template, free Notion template, it. Those are great ways to start using Notion by copying a template to your own Notion and using that as a framework for success.

15:12And if I told you that I didn't do that here, I'd be lying.

15:21One last thing before you dive in, for those of you who don't know Gumroad, Gumroad is like a, you can create your own storefront or marketplace of digital products that you can sell. These could be courses, could be training, could be basically whatever it is that you want to sell. So it creates you this storefront marketplace of digital stuff that you can sell. And it's wonderful because, and I could do another show just on that, quite frankly, because... Let's just make it the Nancy Bain series.

15:55okay so back to notion and one of my you know top three loves i here we let's have a look and we're just going to have an overview of it today so we start off with our strategy and use case and you'll see that we've got a problem statement a project overview and gpt requirements um every one of these is linked to in my Google Drive so that I can easily open them, refer to them, and iterate them. And so this is what my problem statement looks like, the issue, the solution, and the impact. And ChatGPT helped me with each one of these documents. I didn't invent this on my own. I went and discussed my goals with ChatGPT and they helped me to define these documents.

16:54Why is it important to have these documents? Because having a custom GPT is an, it's like prompting, it's an iterative process because you're going to launch your custom GPT today and you're going to change it three times in the next 30 days. And so I love another reason why I love linking to these files and having these files is for version control. And I recommend that you develop a good system for naming your files. Here's my Google Drive over here and how I've organized my files. In fact, Google Drive is where I began my build. in organizing my files. And so name your files well, keep version control in mind.

17:48Down here. Again, I just want to add something about what you just said. People are like, why do I need all of this? I'm going to do two GPTs. Like, why do I need to start with all of that? And I think the mindset needs to be that you can start developing these automations for hundreds of things in your business. So if you have two GPTs, yeah, no problem. Like you can be messy and whatever. But if you're planning to build one for every small use case in your business, which over time we will, and it may not be GPTs, it may be something else. But having it documented as far as what was the goal, what we're trying to achieve, what data we needed to use, all these kind of things will allow you to continue with new tools that come out, new capabilities that show up, new platforms, new improvements, because everything is going to be consistent and everything is going to be aligned.

18:36with these processes that you put in place. Add that to the fact that most companies that are going to do this are not solopreneurs. So whatever in your head is awesome, but it may not be known by the next person to you or the next department or the next division. And so having this all documented and sitting in a place that makes sense, in a structure that makes sense, with naming conventions that make sense, will allow this to scale in an effective way. So I think what you're doing is brilliant as far as looking at the bigger picture in the long term versus, yeah, I can do a GPT like right now?

19:12You can build a custom GPT in five steps and a hook generator, and it can be a damn good one too. But this framework ensures that if I get hit by a bus tomorrow, someone can come in and take over right where I left off. And that's important. Back to the specifications and the instructions, these are actual files that are uploaded into the back end of the custom GPT. So we use my formula structure for any of your files that are uploaded to the database. And I've got a guide on that if you want to reach out. how to use, I'm a big Google Workspace user. So the guide is how to use Google Docs to manage your instruction specification files with Markdown.

20:13Here we have the training data. Everybody's training data is going to look a little bit differently. The important thing here is that I follow a consistent format for each of the documents in the training data. This is also using Markdown because these files will be uploaded into the knowledge base. So these, for me and my Gumroads, these represent the different bundles that I have, the tutorials over there. I alluded to this early in the game, but model training and updates, it's a thing. And one thing I haven't mentioned already, especially in model training and down here performance and analysis, I've got a lot of tasks and reminders set up.

21:04Again, I'm talking far more about Gemini these days because of Google Workspace, but I've been able to automate that whole process and connect it to my notion to make sure that I'm reminded it's time to go and do a thorough test and at least some iterating. Can you dive a little deeper about the model training segment and what exactly you mean by that? Because this means different things to different people. And some people think that only if you're Google, you can train models, which is not true. So can you explain a little more what you mean by that at the very end i've got the best practices for that and i and basic do you want them now it can wait through the end as long as you're going to answer that question yeah and so it's training can be as simple as going to the on the front end of your custom gpt and asking it to perform the functions that you've trained it to do and documenting any issues with that yeah makes sense but i'll give you that i will add to that uploading documents references for it to use is a part of training a model there's a lot of steps that would be considered training a model that will help you perform the tasks better.

22:34Yeah, and links, sharing links. Yeah. So this is specific to my build, and I'm constantly analyzing the performance of my Gumroad tutorial. So I've got a weekly content review, monthly competitor analysis, and what these link to are prompts, actually, that they're just ready to go. And again, I get a calendar reminder that it's time to do a review my Gumroad performance. And then so I'll go copy this prompt and put it into the GPT. I've got a couple of automations for the custom GPT, but you'll note when we get it to the back end, they're not facilitated by the actions. I've got two. So this, but they're facilitated by a Zapier, but not connected to the custom GPT.

23:29So now anytime a Gumroad, add Gumroad subscribers to my Beehive mailing list and then add them to my Gmail contacts. And here I've got a link to my Zaps. When that makes it convenient, those are the two Zaps. If I need to go edit them, they're right there. So this is... I'm going to pause you for just one second. For those of you who don't know what Zapier is, Zapier is the glue that connects any software to any software that you're using. And it's a no-code automation build platform that anyone with zero experience in coding or creating automations can connect and transfer data from one app to the other.

24:17In this particular case, somebody becomes a client on Gumroad. It knows how to grab the email, first name, last name, whatever information is captured over there and populate it in other places. In this particular case, Nancy's email marketing platform, as well as just her Google account. So she has them if she wants to send them emails and so on. So you don't have to do these things manually. You can literally connect almost, let's put it this way. If you haven't heard of Zapier before, you can connect any software to any other software because they have 6 ,000 different connected applications at this point.

24:50And so a lot of this stuff can happen. There's two other tools that are like that. One of them is called Make. There's more than two, but the two that I use, one of them is Make, which is a little easier to use as far as user-friendliness, but doesn't have as many connections and capabilities. And the other end of the scale is NA10, which is an open source, no-code automation tool that is amazing, but it's a lot more technical. So if you have the technical skills, it's a great path to go. So that gives you different options. Thanks for that. And I wanted to add that Zapier was really tough, but now they've got the prompt.

25:31So you just type in docs to calendar to Gmail, and it will design the Zap for you. So it certainly made it a lot more user-friendly. And I'm so happy to see that because it was tough. Now, let's go over to ChatGPT and take a look at it in action. Let's go into the back end here. Again, I'll pause it just for just one second. First of all, for those not watching this, we are opening the editor for the GPTs. For those who don't know even what it is or where it exists, you go into the regular ChatGPT. So chat.openai.com, I think. And then in there on the top left button, on the top left bar, there's a button called Explore GPTs.

26:23And if you click on that, you can see all the GPTs that people are creating. And also on top of that, you'll see all of your GPTs. So you can on the top right, create a new one, or you can click on any one of them and look at how it's actually written or if you're creating a new one. And within there, there's always two options. You can create one, which is basically writing prompts to explain what you're trying to do, which I don't know about Nancy, but that's not what I recommend doing. The other option is to configure, which allows you to actually build it in a more structured way. Either way, you're going to end up on the configure side.

27:00I just think the first step is not very helpful. So now that I've seen what we're looking at, this is what we're looking at. And now to you, Nancy, to go into the details. And I'm going to echo that because I do not recommend the create mode at all. I'm just going to leave it at that. We're on the same page. I think everybody now should know that's the right way to go is to start with configure. I go immediately to the configure mode. So I'm going to give you an overview here. And if you've been peeking at the screen, those who've been able to look at the screen, you're going to notice right away that in the instructions, I'm saying leverage the knowledge base file.

27:52And down here, check it against this file. And where's my scroller here? Refer to the knowledge base. Many times in my instructions, I am referring to these knowledge-based files. And these are the knowledge-based files that we looked at over in Gumroad. And these knowledge-based files all have a consistent format and they're all using Markdown. down. So that's my biggest complaint with the custom GPTs is that by the way, that they don't consistently refer to the knowledge based files. Okay. So if you hear in in the conversation starters, these are the first four prompts, I call them the leader prompts.

28:42And you've noticed, I've tailored these. I've custom tailored these. Refer to the writing style. So any time that I can save a prompt by saying, I remember, I do it. Even in my conversation starters, I'm reminding the bot to refer to the knowledge base files. You will get the best results by reminding it here. You will save time by including in your prompts to check the knowledge-based files here. So I want to again pause you for just one second for the people who are not watching this. There are four main components, five actually, but the fifth one, I don't know if we're going to touch today, of the instructions that you're going to give or what you're going to build into your GPT.

29:39One of them is instructions, and this is basically just a very long prompt of what the GPT needs to do. The second part is called conversation starters. And these are basically, you give it to somebody, if you don't have these conversation starters, people don't really know how to use the GPT. So all these are buttons that show up in your GPT, like a button that you can click. And that's going to be the first prompt that you can use, which kind of gives you an idea what you can do with this GPT. The next component is reference documents. It's called knowledge, and you can upload files to that.

30:10And these could be multiple types of files that you can upload. The interesting thing that Nancy said that I never thought about, but the more structured and consistent you make those, the easier it is for the GPT to understand the reference material that you're giving it. So those of you who don't know what Markdown is, and Nancy mentioned that several times, so I'm going to explain. Markdown is basically the code behind documents that tells it what's heading one, what's heading two, what's our numbering, what are bullets, what's the subject. Like all these things are called Markdown and you don't have to know anything about it other than create a well-structured document.

30:51Like tell it, this is the title of the document. This is heading one. This is heading 1.1, 1.2, 1.3, heading two. And if you use that, you are using Markdown. And if you do that, when the GPT wants to understand the content, it's a lot easier for it to understand because it knows how to read Markdown. And so it understands how the document is structured and when to find different kinds of information, which will increase the chances of you getting the correct information. So that's section three. And section four is basically what capabilities it needs to have. Does it need to connect to the internet?

31:28Does it need to be able to create graphics with Dali? And does it need to do any kind of sophisticated math or code writing? And you want to unselect all the ones you don't need because otherwise it's just overhead. And so these are the different components that Nancy's talking about. So far, you covered the main thing, which is create very well-structured knowledge and then refer to it every time you can to make sure that GPT actually does. absolutely and i will say you can create your instructions copy and paste them into chat gpt and tell it to to convert it to markdown if you have google workspace you can open up a doc and create your instructions and there's an extension that will change that to markdown formula so So there's some really easy ways to structure your data.