Leveraging AI

175 | Stop Wasting Time! Automate Repetitive Tasks with Custom GPTs with Isar Meitis

• Isar Meitis • Season 1 • Episode 175

Share episode

Are you still manually repeating the same tasks in your business?

Repetitive work slows you down, drains your team, and steals time from high-value strategy. But what if you could automate those tasks with AI—without hiring a developer?

In this episode of Leveraging AI, Isar Meitis breaks down how business leaders can build powerful, no-code AI automations using Custom GPTs inside ChatGPT. Plus, you’ll discover free alternatives on platforms like Google Gemini, Perplexity, and Anthropic Claude.

From streamlining internal processes to boosting productivity, this episode is packed with practical insights you can apply today.

In this session, you’ll discover:

✅ What Custom GPTs are and how they work
✅ How to create AI-driven automations to save hours of work
✅ A breakdown of free alternatives like Gems, Spaces, and Projects
✅ Step-by-step guidance on building your own Custom GPT
✅ Pro tips for structuring GPT instructions for maximum efficiency
✅ How to integrate AI automation seamlessly into your business

If you’re not using Custom GPTs yet, you’re leaving massive opportunities on the table. Tune in now to take your productivity to the next level!

About Leveraging AI

If you’ve enjoyed or benefited from some of the insights of this episode, leave us a five-star review on your favorite podcast platform, and let us know what you learned, found helpful, or liked most about this show!

GMT20250317-145116_Recording_as_2560x1440:

Hello and welcome to the Leveraging AI Podcast, a podcast that shares practical, ethical ways to leverage AI to improve efficiency, grow your business, and advance your career. This is Isar Metis, your host, and I've got a really powerful and really exciting episode for you. Today, we are going to talk about how to create small automations using custom GPT within chat GPT. We're also going to learn of other options that are free to do similar things on the other platforms. I'm going to take you step by step on how to build extremely powerful automations that you can use in your company starting today. And if you haven't done it so far, or if you've done it a little bit and you wanna be able to take it to the next step, this is going to be maybe the most powerful and most effective episode of any AI podcast you ever listen to, because if you're not using GPTs, it's absolute magic that you can use across every single aspect of your company. So let's get started. So what is a custom GPT. A Custom GPT is just a set of instructions that are prebuilt and packaged into a conversation within the chat GPT environment, meaning instead of typing and retyping the same prompt again and again and again when you wanna do a repetitive task, you can package it all together, together with reference information and instructions and some other capabilities, and then reuse it again and again and again. Every company, every department in every company has multiple repetitive tasks that they're doing, whether it's happening every day or once a week or once a month, it doesn't matter. You can create these custom GPT to do these things for you. Now, this capability was first introduced in OpenAI a long time ago, but since then, more or less, everybody else has caught up. So now you can do something very similar on all the different platforms. So let's take a quick look on the different platforms. The Open AI universe, it's called Custom gpt. In the anthropic world, it's called projects. In the Google Gemini world, it's called gems. And in perplexity it's called spaces. There's many differences between all these different tools, but the key ones are there's different types and quantities and sizes of files you can attach to each and every one of them as reference information. There's a different context window, meaning how much information you can have in each chat. In these things, there's the limit. To the amount of characters you can put in the actual instructions themselves. There is connectivity to the internet that may or may not exist. In some of them. There are conversation starters, which are like buttons that you can add in order to help users understand what the automation does. This is not a big deal, but it still helps image generation. API access and using Canvas. As you can see from this image still the most powerful tool out there is custom GPT because it has all the different functionality, which some of the other tools have, some of these functionalities. I must say one important thing that gems and spaces are probably relatively close Two Cha GPTs Custom GPT and capabilities, and they are free. So spaces were free more or less since the beginning. Gems became free this past week, and that's one of the things that drove me to record this episode to show you what are the possibilities. Now, the fact we're going to focus today on custom GPT within the Cha Gpt universe. You can use everything that I'm going to teach you in this episode in each and every one of the other tools as well. At the end of this episode, I'm going to give you an exact checklist on how I approach any new GPT project that I'm approaching. But let's start very, very simple with a very high level that there's two ways I usually create GPTs. I either create them for myself or for one of my clients, which means they're very specific to something very specific, or I create them as a tool for other people to use. The most major difference is when it's a tool for other people to use. I start with a bunch of. Questions in order to help the GPT or the large language model. In that case too, in order to help the GPT understand the specific user and the outcome that they're looking for, so these could be questions such as enter your company URL, or who is your target audience? Or what is the goal of whatever the thing that this does, whether it's a post, a proposal, an email, et cetera, and the way I usually do this in order to make it easier for the GPT and most importantly, easier for the user, I actually ask it to ask the questions one by one. Now, even if you do that in many questions, you will ask all the questions together. So what I do is I start with something like. I would like you to ask the user the following questions, and then I write question number one. Let's say the example is what I said before. Who is your target audience? Please define it in as much detail as possible. There's cipher, graphics, geographics, whether from what they're interested in, pain points and so on. And then the next row is do not do anything until you receive an answer from the user. Once you receive an answer from the user, take it into consideration in your process and then ask. And only then I write the second question, and then I add the same line again, and then the third question, and so on. If I create it for myself or for one of my clients, I already provide the answers to all these questions as part of the knowledge base. So I would upload a document that clearly defines the target audience or the tone of the company, or the why of the company, or a template with. How I wanted to create the output or good or bad examples and so on and so forth. I can add all of those things because I know them, because they're for a specific company and a specific use case. Now, when you write the instructions, the instructions can be anything. You just want to explain the GPT what it is that they need to do in order to help the user, and I actually refer to the user when I write the instructions, ask the user to do this, provide the user. This kind of information and so on just works well for me, maybe other people doing it differently. Now, when you write the instruction, there's a length limit, as I mentioned, and it's different in each and every one of the tools. In custom GPT, it is 8,000 characters. Now that sounds a lot, but if you're trying to write really complex, detailed instructions, it's not. And I actually hit that limit multiple times, writing GPTs. Now, when it happened to me the first time, I was really surprised and I did not know what to do, and I started manually trimming what I wrote before. So shortening sentences. Eliminating different things that I found that may not be exactly written the way they need to be, and so on and so forth, trying to make it shorter. Then the next step of my evolution was, oh, I can actually use Chachi PT or a different AI tool to shorten the instructions while keeping the essence of them and making them shorter. So I literally. Copied my entire instructions to our regular chat and said, help me write these instructions in a shorter and more concise way. Still maintaining all the instructions as is not losing any step along the way. And then we wrote a much shorter version, but this may still not be enough. And then I've learned something even better that I can actually write some of the instructions. Into the knowledge base. The knowledge base is the files that you can upload into the custom GPT. Now, before we dive into what else you can upload in these files, let's talk about how to do the instructions. So if some of the instructions live in files that you upload, how do you manage that? So the way I like to look at this is as if the main instructions, the one that are in the custom gpt, is written in the instruction section. And I'm gonna show you in a minute the user interface and how it looks like are the project manager, they're the general flow of the work and the specific instructions that I upload in files to the knowledge base are the workers. They do the actual specific work of specific steps in the process. When I use it this way, I can add technically a lot more instructions than I can than in the 8,000 characters and still having it very well structured So they just told you that you can upload information as files to GPTs and gems and spaces, but what would you want to upload? Well, this could be a document that explains your company's tone of voice, so it will write like you. It could be good and bad examples. Of what the output should be like. This could be a proposal, this could be an email, this could be a newsletter, this could be anything else that you wanna say. Okay, here are good examples that have worked well for me. Here are five posts that get the most amount of engagement for me on this and that platform as a reference and so on. It could be a template or templates if you're doing several different things with that GPT. So I want you to write a proposal following this template that is attached as something. As I mentioned, you can use it as part of the instructions as well. Now, something very important to know when you attach files to the GPT is naming convention works like magic, meaning name the file, exactly what it is. Let's say it's an example for my company. I would call it Multiplai's tone of voice, and that will be a tone of voice document explaining how my company multiply uses in writing different types of outputs. Another file can be a detailed description of my ICP, my ideal client persona. So I will call the file Multiplai's ICP definition, and in there I will define everything that I can about my clients and so on and so forth. Now you want to reference the exact file names. In the instructions in the GPT. So before we go any further, and I confuse you more with all the different options, let's look on how the user interface looks like. So if you are in chat, GPT, the regular user interface, then on the left side there is on the top you have the menu, and then it says Chat, GPT, which starts in your chat. And then if you haven't built any GPTs, you don't have any of that, I. You just have a button called Explore gpt. I have multiple gpt that are already created, that I'm using regularly, but if you click on Explore gpt, it will show you all the different gpt that other people have created and shared with the universe. You can decide to share yours with the universe as well, and we'll see in a minute how to do that. But over there There are millions of GPTs that you can pick from to do different things for you. But if you, on the top right there is a plus create button. If you click on the create button, it allows you to create a new GPT what is available in the user interface. So there's two modes. There's create and Configure. In the create mode, which is just this, these tuggle buttons on the top of the left side, you can write a prompt that explains to Chachi PTI what you wanted to do, and then you will. Create one for you. I highly recommend not doing it. The configure allows you to be a lot more structured and will give you better results from the beginning. And so the first thing there is, is this little circle with a plus button that just allows you to prompt it to create a icon for the GPT. I always do that last when I kind of know how this all works out. And you don't even have to do it then you give it a name. The name can be anything you want. Just make it simple for you to know what it is and also If you're deciding to share it with the world, then make sure that everybody will understand what it is. The second thing is description. Again, this is just a longer version of the name, explaining to other people what this GPT does. And then the third component is instructions. Instructions is where you actually put the instructions. As I mentioned, in most cases, this will be your entire instructions and everything that needs to go in them, but in. Some cases, this will be part of the instructions and some of it will be in the knowledge base. Then there's this cool thing called conversation starters that basically adds buttons on the actual GPT to help people understand what they can do. I will show you an actual example in a second. Below that, there is knowledge. Which there's a button to set, upload files, and over there you can upload the files that become your knowledge base. And we discuss about what these can do. And then there's two more components to the gpt. One is capabilities. There are four capabilities that are optional on all GPT web search that will allow your GPT to access the web canvas, which will allow your GPT to use Canvas, which is a side by side solution that allows you to see the. Output as a document that you can edit on the right side of the screen and have your chat on the left side of the screen. And every time I'm doing any creative work, like creating blog posts or posts for social media or writing proposals and so on, I will ask it to be in Canvas and I will check that box because it will be easier for me to work with it afterwards. The third option is DE image generation. If you want this GPT to generate any images, you want that click, and then the bottom one is code interpreter, which allows it to write Python code and analyze data. if that's a part of your process, then you want that. Uncheck all the stuff that you don't need because it just confuses the GPT is what it can and cannot do. As an example, if you want your GPT to just use the data that the user uploads to analyze the monthly financials, then uncheck the web search so it doesn't have access to any additional information. And then on the bottom there's a button that most people don't use, but actually provide very powerful advanced capabilities, and it's called ADD actions. And if you click on that, it opens another user interface. In that interface. What you can do is you can connect this GPT to a third party tool, API that API could be anything from weather to stock, to navigation guidance to geographical information to your eRP or CRM, anything that has an API and the way you do this is it allows you to add authentication if there is one, and then it allows you to add a schema, which is how you define the rules of how an API works. If you know nothing about it like me and you still wanna be able to use it, cha GPT actually created something very, very cool. There is another GPT that they created that's called Get Help from Action GPT, and there's a button that links to that on the bottom. If you click on that, it will open a new window for you that will give you different options on. How to write the schema for your API. This could be because you have access to the documentation of the API or an example of the API or a URL of the company, and it will go and research it by itself and many different options. Like I said, this is an advanced option, which we're not going to dive into right now. So now let's look at a quick example. this GPT that I created as one of my demos to small businesses, when I did a executive ed lecture here in town, Created this for them as a demo, and it's called invoice summary, GPT, and it's called invoice summary. And the description says create a table summarizing all uploaded invoice. Now all uploaded invoices. So if I click now to edit. You'll see that the name has invoice summary. The description is this, the instructions are not very long, and it just tells the GPT what to do. You're an expert accountant with 20 Rivers experience. Your task is to collect information from invoices and arrange the information in a table. Read every invoice carefully and look for the following information in each invoice. The company. Issuing the invoice issue date, payment due date, invoice number or invoice id, total amount in native currency formatted as, and then I give it the format like dollars and euros and so on. Total amount in USD formatted in the same way, and then create a table with a column for each and every one of the items above. Add a row with the information From each invoice and add a total row for the invoice amount column to calculate the total amount, and then I created a button. In the conversation starters called, upload all the invoices you would like for your summary, and you can see that the, that the button shows up. Now, I didn't upload any files in this particular case because I didn't need any, but let's say I had a format I wanted it to follow as far as the columns and how they are. I could have done this as well and I've checked code interpreter because I needed to be able to do math and calculate the number and so on. And now what you can see is that on the right side, I have the GPT for testing. So your test environment is live, and every time I make a change on the left, it will immediately impact. On the right, there's an update button on the top. When you create the GPT the first time, this will say create instead of update. So if you're just listening to this, that's the button on the top right corner of the custom GPT universe. And if you click. On it the first time, it will create the GPT and every other time it will be an update. Once you click on that, it will pop up a window that will ask you how would you like to share it? Would you like to share it with the world? Would you like to share it with people with the link or would you like to keep it private just for you? And once you click on that, it will create the GPT. Once you click on view GPT, it will take you to the real regular world, and then you can use the GPT in here time and time again. I'm not gonna show you how this works, but this is actually pretty cool. You can upload multiple invoices and it creates a CSV file with all the information from all the different invoices, perfectly, accurately, every single time with multiple invoices, and it can save you a hell of a lot of time if you're just needing to manually key that information into your ERP or your accounting software. Now I want to take you to the next level, and I want to give you a really cool example that I'm working on with one of my clients. So one of my clients is Sonance. They're an amazing company. They're a leading manufacturer of speakers, and they are a company that has very clear why and very strong core values and culture, and it's very obvious with everything that they do. So on their About Us page, they're saying At Sonance we envision a world where music is woven like a thread through the fabric of everyday life, Our hidden audio solutions subtly elevate each moment turning the ordinary into magical shared experiences because life is better with music. Now one of their core values is helping everybody, the environment themselves, their ecosystem, and everybody in the world. They do a lot of volunteer and philanthropic work, but in addition, one of the things that they're trying to do is to obviously help their distributors and they're trying to help their distributors on many different aspects. One of them is helping them write better proposals, better documents, and better information for their distributors clients so they can sell more of their product. They're also trying to help educate them on how to better use AI in their businesses across everything that they're doing and their CEO Ari Suran, Who is an amazing example of how A CEO should drive AI change in a business. He's fully committed, they've established an AI committee. I've done multiple sessions of training to their team, both in person and online. They have been investing heavily in AI learning and AI implementation, and ARI himself is vested in taking the time and developing solutions and using them and showing them both in the company and to other people in their ecosystem. So he wanted to create a GPT that will help their distributors write better cover letters for their proposals that align with each company's core values and why? Because he believes, and I agree with him a hundred percent, that doing that creates a better emotional connection and explains to the users why they should buy from you and not what they should buy from you. And that makes a very big difference. He wrote a really good GPT that actually does that, and the only issue that he had was how to collect each company's information each time. And the process, which is not that hard, was to go to each company's website to collect information from their about us page, their homepage, and so on, then save it as documents and then upload these documents as the knowledge base for the GPT and reference it in the instructions. And he asked me if I can help him develop a better, more efficient solution for that part of the process. Now, before we dive into exactly what I did, I wanna explain something about GPTs. GPT are incredible and again, broaden that that is true for gems and spaces and all the other ones as well. We're just talking in this episode about gpt specifically. Gpt are amazing, but if you give them too much instructions, they start getting confused just like a great intern would. So you always have to consider them as such. So if you try them to do two, three different complex things, the better solution is actually to break it into several gpt that each and every one of them are doing a very clear step of the process. And this is what I've done in this particular case, so what I've done is the first manual step of the process I created as a separate GPT. So let's see how this looks like. I created a GPT that's called Define Company's Y Core values and tone. And you can see that there's a button that I created. Please enter your company's website, your URL. You don't even have to click the button, it just says that over there so you know what to do. So let's take another company from their industry. I'm just gonna copy their URL. That's all I need to do, and I'm going to drop it in this GPTI don't need to do anything else because all the instructions are built into it. When I click go, what the GPT will do, it will go and research their website across multiple pages and will try to identify all the information that I told you to identify. So you can see the output now becomes core values, beliefs, and why, and it gives me a detailed description of this company, so I will read it to you just to show you how great this is without doing any work on my end, Denizen mission centers on redefining the concept of home by integrating technology and design to create personalized, enriching environments. They believe that a home should be more than just a place to live. It should be a space that enhances one's lifestyle and wellbeing. Their approach emphasizes customization, collaboration, and authenticity. Then he talks about their core values, authenticity, customization, and client-centric approach. It talks about their why, and then it gives a style guide, tone of voice. Tone of characteristics, warm and welcoming. Each and every one of those things that I'm reading to you has a full sentence or a few bullet points after it. I just want to go through it quickly. So warm and welcoming, expert, yet approachable, inspirational, and visionary. Then it adds do's and don'ts on how to use their tone properly and what to avoid. So like, avoid jargon, steer clear of overly technical language That could, clients. Don't generalize. Refrain from over promising. So I got all these things and all these definitions just by dropping in the URL of the company, and I can do this for every single company. And now all the company needs to do is to copy this into a document and create a version of the second GPT. The second GPT is called brand voice generator. I added two different options. So here helped me rewrite a draft, and the other is write a proposal. But in the back of it, each company can add more buttons to do other things and relate to these different options in the instructions. And then all you have to do is connect in the backend of that GPT. So if you go and edit the GPT, you can add file in the bottom, in the knowledge base with the output of the first GPT that clearly defines your why. Clearly defines your tone of voice, clearly defines how to write on your behalf and what to avoid when writing on your behalf. And then this thing works like magic. So let's look at the output for Sonance as an example. So I created a now specific version with the Sonance stone of voice in there. And again, this is the same exact GPT and all it says is, okay, write a proposal cover letter. So let's, let's try that. So I clicked on the bottom and it tells me, okay, is your client an individual or a business? And if it's a business, give me their URL. So let's use the same company just as an example. So I'm gonna say it's a company and their uRL is, and I'm gonna give the URL by the way. I don't have to type all of that. I could have just placed the URL. And the other thing that it's asking that I need to do is to obviously upload the proposal because I'm writing a proposal cover letter, so I will do that as well. So all I have to do is drop the proposal file in there and then hit go. So all I have is the proposal and the URL and I click go and I will read the whole thing and what it does actually in the background, in the instructions. In addition to knowing our to Novos, it actually goes and searches the client's company's core values, and it tries to create an emotional connection between our core values and why to their core values and why in the cover later while referencing stuff from the proposal. So the output here is our mission is to seamlessly integrate technology into everyday life enhancing experiences without drawing attention to the technology itself. We recognize that tennis then shares similar philosophies, striving to integrate stunning technology with architecture and interior design to provide memorable. Experiences that enhance the way people live and relate in the space. And then it says, in closing, our proposal for the AV installation. And it gives a lot of information about what's in the proposal. And then it ends with believe that our shared commitment to innovation and design excellence makes us ideal partner for this project and so on and so forth. And it asks, gives you a space to put in your name, your position, your company in. To sign on the bottom basically. So the magic here is that with dropping a proposal draft and dropping in a URL of a company, it knows how to create an emotional connection between the why and the core values of the two companies to appeal to the emotional side of the person that's gonna read this. And I can do this with any kind of document, obviously, not just with proposals. So. Let's do a recap and a summary, and I will walk you through the full process on how I create gpt. The first thing that I do when I need to create A GPT is I clearly define the outcome. What is it that I'm actually trying to do? What is that repetitive task that me or my team or my clients are doing on regular basis? And what is the output that will be the most effective to the process? Then I usually start with a regular chat. I don't actually start with A GPT. I start with a regular chat and try to write instructions that will actually make it happen. I go and iterate back and forth and back and forth and back and forth until I get it to work. Okay? And then I convert it into a GPT. Then I test the GPT, and then I iterate in the GPT itself. The way I iterate in the GPT itself is. I will run it and we'll see if it works and if it doesn't work, I will try to finesse it. And in many cases when I don't know why it's not working or what should be the next step, I will ask the actual chat. So when you run the GPT, it's a regular chat, so when it's in that chat that runs the GPT. And it doesn't do what I wanted it to do. I will literally ask something like, how can I improve the instructions in the GPT in order to prevent this thing from happening? And they will give you an idea and then we'll ask it. Where in the instructions should I place that? And then I go and place it into the instructions. I must have been, in many cases, I decide myself where it needs to go in the instructions. Another way that I iterate, as I mentioned, is I would go and say, please write these instructions in a more concise and clear way. I will put in the entire instructions and they will write it more concise and then we'll put it back in and I go back and forth, uh, usually with two windows, uh, in order to get the GPT to work perfectly every single time. Now, a few other things that I mentioned, naming conventions of the files in the knowledge base are important and referencing them in the right place. So go and check specific file name for this type of information and then do 1, 2, 3, 4, et cetera. In the instructions in the GPT itself is helping a lot to the GPT to have structure and clear on what it's doing. And then for more advanced capabilities, as I mentioned, you can connect this to any API of any tool that you have access to, which is extremely powerful. You can use Canvas as a side by side scenario to do these kind of things. You can ask it to do web search and you can. Create multiple gpt that can work as a step-by-step process. Another thing that is very helpful when you're creating several gpt that work together is you can use the at symbol in a regular conversation to call any GPT that you have created. So you're in a regular conversation and I will show you an example. I have a GPT that helped me write drafts of LinkedIn posts based on the transcription of my episodes. So All I have to do is drop in the entire transcription of the podcast and it will write a draft for me. It will write the draft in canvas, as we mentioned before, and now you can see in the example on the right I have the actual, thing that it is suggesting. And on the left I have the regular chat, but what I can do right now once it's done, is to call a GPT that will write me better hooks. So what I can do right now is I can add sign. And then you can see a popup menu will show up. And in the popup menu I can see all my gpt and I can call the one that's called hook generator. And you can see there's one called Post Hook Generator. And because it's within the chat and it knows the context, I don't even have to tell it anything. I can just hit go and all I have to do is select post hook generator from the list and copy and paste. The post in there, and then it will suggest different hooks to me, what is actually happening here is that this GPT is now running within this new conversation and it's just calling it in the background, and you can call multiple GPTs in a single conversation. So again, you can create several GPTs that you want to use as a step-by-step process and then call them in a regular chat in order to do the sequence the way you want to do this. So final summary, gpt, Gems, Spaces and Projects in Claude are really powerful capabilities that you can use across multiple use cases. Cases in your company that generate automations that can save you and your team and other teams. in your company lots and lots of time while generating high quality results. If you know how to write the instructions properly and if you follow everything that I shared with you today, you should be able to do that If you found this episode valuable. Please share it with other people who can benefit from it. Click on the share button on your phone right now and share it with a few people who can benefit from it. Also, I would really appreciate it if you leave us a review and if you have any comments or any suggestions or other things that you're doing with GPT or you wanna learn how to do GPTs, please reach out to me on LinkedIn. Isar Metis, I-S-A-R-M-E-I-T-I-S. I'm the only Isar Mateis on LinkedIn, so it's, Really easy to find and tell me what you are doing with gpt. I would really appreciate that. And until next time, have an amazing rest of your week.

People on this episode

Head Shot

Isar Meitis

Host