Content with Humans

Custom Chat GPTs: A Conversation with Adam Walsh, Bizualize

July 29, 2024 Leonie Seysan Season 1 Episode 5

Chat GPT presents a huge opportunity for smaller businesses, particularly around the creation of custom GPTs – but many aren’t sure how to harness it. 

In this episode, host Leonie Seysan speaks with Adam Walsh from Bizualize. Adam has been creating custom GPT applications and showing others how to do it from the time it became an option.

 

Adam talks about how custom GPTS are being used, and how quick and easy it can be to create one – some of his 1-hour DIY workshop clients have created a working tool within the hour! 

 

From creating tools to attract potential clients to drastically reducing the time taken for regular admin tasks, the possibilities are plenty. Listen in to get your own ideas flowing!

Creating a Custom GPT:  A conversation with GPT builder Adam Walsh


LEONIE: Welcome to episode five, Content with Humans. I'm Leonie from Article Writers Australia and host of the Content with Humans podcast. In episodes so far, I've spoken with videography agency owner, Dave Power, Sydney SEO consultant, Pulkit Agrawal, European content translation and localisation expert, Anja Jones, and Johnna Ehmke, a marketing strategist who began her career in America before moving to Australia.

They've all generously shared their expertise and insights. So, if you haven't checked out those early episodes, please do. 

If you've been wondering how to make the most of generative AI technology, I think you'll find today's guest and topic fascinating. I'm speaking with Adam Walsh from Bizualize. Adam develops custom GPT applications for clients and also teaches people how to do it themselves. He's currently developing a custom GPT for us at Article Writers Australia.

I'm not going to reveal much about that today, but I will say that it's a time-saving tool for content marketing managers to use. And we're looking for some marketing managers to test drive it for us and provide some feedback before we launch it. And I'll put something on my LinkedIn about that.

But for now, let's find out more about Adam Walsh and custom GPTs. Welcome to Content with Humans. 


ADAM: Thanks very much for having me, Leonie.

What is a custom GPT?

LEONIE: So, Adam, ChatGPT is still relatively new, and many people are still exploring its uses. It's possible that a lot of people don't understand what a custom GPT is. So, I thought I'd start by asking you what exactly is a custom GPT? 

ADAM: So, in a nutshell, a custom GPT is a very simple tool that you can build to leverage AI like ChatGPT and train it on your knowledge, your data, your behaviour. So very, very highly customisable, very easy to build. 

LEONIE: And so, what are some of the popular custom GPTs that people can access and their uses? 

ADAM: Look, Canva's got a nice little custom GPT that they've built. So that allows you to describe a presentation for instance, and then it'll use the Canva platform to actually help you design the presentation. But I mean, there's a whole range of different styles out there. There's GPTs to help you learn languages. There's GPTs to help you understand, you know, scholarly articles, for instance. So, there's a GPT out there called Scholarly GPT which actually allows you to ask questions and then it goes and finds the answers based on peer reviewed articles, for instance. 


Get to Know Adam

LEONIE: Okay, excellent. So about now, I like to throw in a random personal question that's got nothing to do with what we're talking about. So random personal question, if you had to leave Australia and live in a non-English speaking country, which one would you choose and why? 

ADAM: Very easy question for me because it's exactly what I'm doing. So that's Japan. And the reason for that is because Japan has a, there's a wonderful sort of balance in the culture over there between technology and nature which I really appreciate.

LEONIE: Oh, okay, so when are you off? Are you going shortly or? 

ADAM: We'll be going at the beginning of 2026. 

LEONIE: Oh, how exciting. 

ADAM: Yeah, we'll be renovating an old farmhouse in rural Japan. So, looking forward to it. 

LEONIE: Oh, excellent. We'll ask for photos.

ADAM: Yeah, absolutely. 

Popular GPTs

LEONIE: So, you create custom GPTs for businesses and I’m really curious and I know that the listeners will be too. I’d love to hear some examples of what people are doing with custom GPTs for their businesses.

ADAM: Yeah, look, I mean, the possibilities are endless. Hence the requests I'm getting are really broad. In a business sense, a lot of my customers are looking for simple stuff like they've got somebody doing social media posts for them, but they want those posts to sound a certain way or feel a certain way.

And they want to make sure that the message is consistent without having to sort of stand, without having to look over people's shoulders the whole time. So custom GPTs help them to train the AI so that their social media manager has a nice tool that they can use but it keeps them within the bounds of what they're meant to be doing brand wise. 

Then I've got customers that are asking me to help them build and train AI on their internal policies, for instance. So they might have new employees coming onboard and instead of the employee needing to chase somebody who is a mentor in the business and waiting for that person to respond to an email or waiting for that person to respond to a phone call, they can build them a tool that has all of the knowledge built into it. And they can ask those questions in a contextual manner that then can give them an answer that's actually specific to the situation that they might be in. And they can get that information very quickly which is really good.

Other applications, I had a gentleman recently ask me to build a GPT for him that will help him extract fruit prices because that's one of the services that he offers for some of his agricultural customers in the financial sector. So, it's really, really broad. 


Reasons for a Custom GPT

LEONIE: Okay, so would I be right in saying that the two main reasons to develop a custom GPT would be to either automate a particular business task or to save time on it or as a tool for your customers or prospective customers to use? 

ADAM: Yeah, I think that first point is, and you actually, you made a distinction there which I normally make. It's not so much around automation. So, using AI to automate things will be, that will be software driven stuff. The Microsofts of the world, the Googles of the world will take care of that. 

When we talk about AI in your business on a daily basis, we're talking about optimising processes. We're talking about you still being in control. You need to be there. You need to be driving the process, but the AI is making it a lot more efficient, a lot more effective and a lot faster. So yeah, you could say that one of those reasons is to optimise those processes. 

And yes, without a doubt, particularly in the future, being able to build little tools that your customers can use. And that might be question and answer sessions on your website, but it could also be just adding a little bit of a value to your customers as you know. I've done GPTs for people, for instance, that they actually wanna use them as brochures. So, instead of just having a PDF that somebody looks at an image, looks at some text, they can actually ask questions of the brochure. Well, what about my particular situation? You haven't actually spoken about exactly what I need. How does what you do translate to what I need? Is there an example of how what you do actually applies to my problem? So, brochures are actually, I think that's gonna be really interesting. 

LEONIE: That is interesting, isn't it? You could, I imagine, embed that on your website or something, or how would you approach that? 

ADAM: Yeah, you could. Because the thing with GPTs is that ultimately what they are is they're a platform agnostic tool. So, what I mean is it doesn't really matter. They're not tied to a platform. They're just tied to basically a web address or a URL, which means that you can put them anywhere. So, you can put them on your website, you can put them on, you can link to them in brochures, you can link to them in emails, email signatures, wherever you want them. 

LEONIE: Yeah, or you can have a completely different domain name for whatever tool it is and link to that.

ADAM: You could, yeah, yeah, you could. I mean, it really depends on the way that you look at it. The way that I do it is I try and keep it within the, for example, a custom GPT on a ChatGPT platform. I try and keep it within ChatGPT. And the reason I do that is to make it nice and simple, so the user doesn't need any third-party integrations or tools or anything like that. But yes, you absolutely could create a domain. You could create a little application that uses ChatGPT and create a separate domain for that. Yes, 100%, you could do that. 


Custom GPT Opportunities

LEONIE: So, what do you see as the main areas of opportunity, particularly for small businesses that a custom GPT could offer? 

ADAM: Yeah, look, ultimately what it comes down to is there's three areas really, and that's either creating value for your customers.

So that's around getting those little tools that can help you do things for your customers that you're not necessarily gonna charge for, but you also don't wanna absorb a whole heap of costs and time investment. The second thing is just really sort of improving that efficiency within your business. And that's just doing things faster, just doing things quicker, doing them better on a daily basis.

And then the third thing is, of course, communicating within your business, whether it's internally or externally, a lot quicker, a lot more effectively, a lot more concisely. So, when you're talking about, I think administration's a big one with ChatGPT. So those little administrative tasks that are taking 20 minutes, 30 minutes for people to do data entry or data collection type of tasks, for instance.

Something that you, perfect example, how many times a day do people go into a spreadsheet, copy and paste a whole heap of data, or you've got a PDF with a table on it and you need to take that data and put it into a CSV file or a Excel file. You know, that's potentially a 10-to-20-minute job, or you can use AI to do that in about 30 seconds. So, it's those little tasks that you're doing over and over again each day that will really, really add up.

So yeah, you're talking about administration, obviously content is a big one. What I say to people though, is that you're not gonna be creating videos with AI anytime soon. I know there's things out there like Sora that look really impressive, but right now it's really not there.

But there's things like transcribing, videos, taking those transcriptions and getting concepts out of them and turning that into content, turning that into articles very, very quickly, getting action points out of Zoom meetings, things like that, very, very quickly. Yes, there's tools that will do that for you, but they can't understand the inflexions, they can't understand the reasoning and the nuances in the same way that a LLM like ChatGPT or Gemini or Claude. 

LEONIE: Yeah, we're still a way away from that, aren't we? We still need, even with the content creation, it's quite clunky and definitely needs human oversight. It's really only quite simple things that you would wanna be doing. 

ADAM: Look, I think most people have come to the point, the realisation now that they've all tried a couple of the content generation websites, and ultimately you have to do so much work editing it that you might as well do it yourself.

So, it's, yeah, when you're talking about coming up with an idea for content, absolutely. When you're talking about building a framework of those ideas and how you can expand those ideas and how you can reuse those ideas and come up with other ways to express those ideas, absolutely, that's where it really excels. And to be honest, that's 50/60% of the work if you think about it.

LEONIE: Yeah, it is. It's a good time-saver for that sort of thing, isn't it? 

ADAM: Absolutely. 


Clever GPT Ideas

LEONIE: What's the cleverest custom GPT idea that you have come across? 

ADAM: Ooh, the cleverest. That's a hard one. Look, I think probably the cleverest one that I've seen right now would probably be a language GPT that I saw, which is around taking lessons. So, you're learning French, for instance, you're learning Japanese, you're learning Indonesian, whatever language. So what the GPT will actually do is it will capture the vocabulary that was used in a particular lesson, take that, turn that into a chat, and then you can practice the vocabulary that you learnt in each particular lesson on a weekly basis with ChatGPT as if your tutor was actually right there with you. I think that's a really, really clever application. 

LEONIE: That is a clever, that is a clever application. I wonder if it does Farsi. My in-laws speak Farsi, something, a language that I've never picked up. So, I'll have to have a look now.

Developing a Custom GPT

LEONIE: Something that surprised me when I first discussed custom GPTs with you a few months ago was how inexpensive they can be to develop. Could you talk a little bit about what's involved in developing a custom GPT? 

ADAM: Yeah, so developing a custom GPT can range from, you know, I'm doing a job, for instance, I've done some work with you, for instance, Leonie, you know, and I've got one of the, there's four processes that we're building. One of those processes has 32 prompts chained together

So that's quite a complex custom GPT, and that involves, you know, that involves a project approach, scoping out the GPT, understanding how it'll work, understanding all the different decisional matrices that will occur, you know, along that process. But it can be as simple as getting a good response that you like from a question that you asked, ChatGPT, and building a custom GPT that describes that response. So that GPT will give you that same style response every time.

And that can take as little as a minute to build. You know, I've got some, I've got GPTs that I've built that were quite literally two lines of text. It's just that I don't wanna have to copy and paste those two lines of text every single time I use them. So, you know, it really depends on what it is that you're trying to do. 

So that's the first thing. And the second thing is that because of the open nature of custom GPTs and the accessibility, you have this flexibility to build them bit by bit. So, you can try it out, see how it works. If you're happy with it, great. You then start using it and you're getting some benefit out of it. And then you can decide to yourself, I'd like to add another feature. So, you can go in and build that yourself. And because you don't need code, because you don't need a highly technical approach, you just need a strategic approach and a logical approach. You just need to understand exactly how to, you know, how to formulate the ideas, how to take what it is you wanna do and articulate that. 

LEONIE: So, it's a lot less time consuming and costly than if you were trying to create your own software tool. 

ADAM: Absolutely, absolutely, absolutely.


Options for Prospective Clients

LEONIE: What options do you have for clients who want to develop a custom GPT?

ADAM: So, I have, there's really four options that I have. So, I run a program called the AI Primer, which is a one-hour session. And my clients come to me with an idea, just a little task in their business that they wanna improve.

And we build a custom GPT on the spot to get that task done for them. That's a very small task, very simple task. I then have the SoulCraft AI Community, which is an online subscription model. And that's around, you know, tips, tricks, courses, bi-weekly Zoom sessions. And that's around just learning the skillset required and how to think about AI, how to think about how these LLMs actually work and how to think about how you approach them. And that's around essentially building your skills to the point where you can use these tools really effectively in your business and then actually learn, okay, when is it that I need to go and build myself a custom GPT? Because I don't always build a custom GPT.

You know, sometimes it is quicker to just simply type the prompt over and over again. So, there's, you know, part of the skillset is understanding, okay, well, when is it better to go and build a GPT? And then the third option I do is team training. And that's around sort of how to bring those skillsets into your business and how to identify the opportunities within your business and then apply the correct skillset.

And then the fourth option is of course, which is one of the, the one that I'm working with you on the moment, Leonie, and that is, I will build a custom GPT for you. And that's ultimately just a project-based approach. 


Challenges with Custom GPTs

LEONIE: So, I know that with the custom GPT project you're working on for us, one of the challenges has been that open AI are still constantly tweaking and changing things. That must really keep you on your toes. What are some of the challenges that that creates? 

ADAM: Yeah, it does. So probably one of the biggest challenges is just trying to keep up because a lot of the changes, a lot of the tweaks are kind of, they just happen in the background. There's really not a lot of information as to what's being done on a daily basis and what those changes are. So that can be quite difficult to keep up with. But yeah, look, I mean, at the end of the day, it's like any technology. You know, it changes on a daily basis almost. And, you know, it's constantly being updated. You just gotta keep at it. And I suppose I'm lucky in the sense that I can spend, you know, a large portion of my day just ensuring that I'm across all of these changes. 


Issues when Developing Custom GPTs

LEONIE: There's something else that I was wondering about. As professional content writers, while we found GPT quite helpful for generating ideas, but not all that useful for the type of complex long-form content that we do, for example, for B2B marketing teams, because it isn't always accurate in its responses, and as they say, it hallucinates and sometimes makes things up in an effort to please you, does that cause problems with the development of custom GPTs as well? 

ADAM: I wouldn't say it causes problems, but it is something that you have to take into account. So, the idea of hallucination with LLMs is that the hallucinations are kind of that they're a by-product of the creative engine within these products. So, what will happen is it tends to opt towards giving you something rather than giving you nothing. So yes, so it'll make stuff up, which is completely wrong in order to just give you something. So, when you're talking about developing, you know, GPTs, there's plenty of ways that you can combat that. And one of the simplest ones is to just say, if you can't think of anything, tell me. 

LEONIE: Yeah, I do, I tell it not to make things up. I tell it to make sure it's factually accurate and not make something up. 

ADAM: Yeah, well, see, I'll give you a little tip on that because the best way to work with LLMs is active direction. So, if you tell it not to make things up, you're not really giving it an out. You're just saying, don't make things up. What you're probably better off saying is instead of don't make things up, give it a specific thing to do. Give it a specific response that it needs to give you. So, for instance, if you can't find an answer, please tell me no answer available. 

LEONIE: Okay. 

ADAM: So, you’re giving it a task to actually perform in the event that there's no answer.

Whereas if you say, just don't make things up, you'll find that… One of the weird things about LLMs, I've come across this quite often, is that if you say, don't do something, quite often the word don't seems to disappear and it'll go ahead and it'll actually do it. It's a very strange thing. So, you've actually got to really focus that language on when you want it to not do something, rather than say, don't do it, say, avoid doing this because avoiding is the active language. 

LEONIE: Ah, okay. And then give it something else to do instead.

ADAM: Yes. 


The Funniest Incorrect Response

LEONIE: I have to ask, what's the funniest incorrect response you've ever had after asking GPT or another Gemini or any of the others, a question? 

ADAM: Yeah, so probably one of the funnier responses, I think was, I showed a photo of my wife. We were sitting out there, we were sitting out there, this is very early on in the piece I might add, but we were sitting out the back there, sort of camping and my son was mucking around, and he took a photo of myself and my wife.

And we had our cups of coffee and he asked ChatGPT to describe the photo and it said, oh, there's two men sitting at a table. And anybody that knows my wife knows that she's a very girly girl. So, she didn't like that very much at all.

LEONIE: I hope it's gotten a bit better since then. 

ADAM: Much better, much better. 


Before Custom GPT Creation

LEONIE: What were you doing before ChatGPT and what made you pivot into custom GPT creation? 

ADAM: Yeah, I actually started in business consulting, specialising in values, core values. So, I had a process that I was using to actually extract or what I'll call discover, what I call the business soul, which was ultimately the values that are the things that actually drive the actions within a business to fulfil the function of that business. So, this was a 90 minute, a four session, 90 minutes per session process that I would take my customers through. And ultimately it was high value, and it was relatively low cost, but it was still being out of your business four times for 90 minutes each. And it was still a $2000 upfront cost. 

And at the end of the day, and unless you can do something with those values, if you're about to embark on a large cultural overhaul, if you're about to embark on something where those core values are very, very important to unblock you and move you to the next stage, then it was a good process. But it was hard to sell into the market that I was after, which was that small to medium business.

LEONIE: So, is custom GPT building is a more practical way to help?

ADAM: Well, so what happened is I was looking for an online version of this business soul discovery process. I wanted to cut it down, but the problem was that there was no real way to maintain that discussional, that conversational element to the process, because I can do a video course on how to do my process, but you're not talking back and forth. You're not discussing. So yeah, I was really looking for an online way to an online version of this program that my customers could self-serve. And as soon as I saw ChatGPT, I thought this could be it. I mean, it's able to have those conversations. Maybe I can get it to have those conversations instead of me or with me or to augment what I do. 

So over about eight months, I ended up training ChatGPT OT on every stage of this process, how I worked with the answers that my customers gave me, the logic I use to actually go and do something with those conversations.

And we ended up cutting it down to 90 minutes, one 90-minute session at $350. So that was an outcome straight away. And I thought that's a business outcome by using a tool like ChatGPT.

And the second thing was that I would run this program with my customers, and I would show them the AI. I've always been very transparent. It's always been very important that my customers knew that I was using AI to do this.

And I just kept getting people coming to me and saying, hey, look, love the session, that was great. But tell me more about that tool that you were using. Because I really liked the way that that sort of led the conversation and was really able to assist in what we were doing.

So, I had a couple of people sort of request me, request help building their own tools. And that just kind of really, really took off. And after a while, I was actually doing more of that work than the business soul discovery.

So yeah, look, I got into AI by doing exactly what I trained people to do ultimately.   


AI Impacts and Changes

LEONIE: So ChatGPT and other generative AI are having quite an impact, particularly when it comes to content marketing. So, for example, a lot of SEO companies and small businesses now use tools to create content and save money. And as a result of that, I see a lot of previously freelance copywriters who were doing quite well are now going back to paid employment, or in some cases, even pivoting to train people in using AI tools because they've lost their writing income. What other impacts or changes have you noticed whether they're good or bad? 

ADAM: Yeah, there's going to be little impacts like that across ultimately all industries, really. You know, I don't see many places that it's not going to touch in some way.

But the second part of that point there is that, you know, like you mentioned, they're pivoting to train people in using those tools. So, what it's done is, to an extent, yes, it's affected some of that really, I suppose that lower end work. You know, I can't imagine that there's copywriters out there and correct me if I'm wrong, but I can't imagine that there's copywriters out there that are losing that really advanced work. You know, that, yeah. So, it's more that sort of simple stuff. 

But, you know, it's opening up other opportunities. So if I'm a copywriter and I'm seeing that part of that, of the industry sort of retract a little bit, if I know how to use AI and I can go and I can show other people how to get that stuff done themselves, well, then I'm sort of, you know, I'm recovering that sort of, that market. And I think that's gonna occur right across the board. 

I think, you know, there's going to be a process of, and this happens with all technology, the technology comes in, it starts to sort of take up some of those simpler tasks. It starts to take up the mental load, if you like, of some of those more admin-y type of tasks. And then we start to see that disruption. But then what it does is that, of course, it boosts productivity, it boosts, you know, efficiency, which then means more work needs to be done. And people still need to do that work. 

So, it always, there's always a level where the technology starts to impact on an industry. And then it kind of hits a point where it's like, okay, now we need the humans back in here. And that's when we start to see new jobs springing up. That's when we start to see new, you know, new type, new careers, new, you know, completely new industries starting to spring up. So, yeah, I hope that answers your question. It is a very, very difficult question. 

LEONIE: Yeah, I think you haven't, some of the things that are going to happen as a result of it, we probably haven't even thought of yet. 

ADAM: Oh, I guarantee it. In fact, I would guarantee you that in 10 years from now, the jobs that have been created by AI, I would say 95% of them, we would have no idea what they're going to be. And we have no way of anticipating. That's my thought. And on a business level, I'm talking, I mean, obviously there's medical breakthroughs that we're anticipating, and there's technology breakthroughs that we're anticipating, and those will happen. But insofar as how we as business owners use this technology in our daily lives and our business lives, yeah, we have no idea what's, we have no idea where that's going to go. And I think that's quite exciting, to be honest.

LEONIE: It's certainly going to keep us on our toes, isn't it? 

ADAM: Oh, absolutely. Absolutely. 

LEONIE: I'm not very familiar yet with Google's Gemini and other generative AI tools, but do they have similar opportunities to create custom applications? 

ADAM: Yeah. Yeah, so Google has recently released something called Gems, which are sort of similar to GPTs. You've also got Anthropic has a product called Claude, and they have a similar feature, which is either about to be released or has been released. So yes, in short, yeah, you can do similar.

There's also a little sort of bot making websites out there, I suppose you might call them, little platforms like Poe, P-O-E, and that actually allows you to make your own little GPTs or your own little bots based on most of the popular platforms out there. 

LEONIE: Oh, fantastic. Okay, I'll have to go and have a look at that.


AI Tools

LEONIE: So, when comparing the generative AI tools, which one would you rate most highly and why?

ADAM: Look, for me, it's ChatGPT without a doubt. I just think that it performs the best right across the board. There are certain elements within the other two that sort of, you know, the strength of Claude.

Claude, I feel Claude's probably, you know, a little bit stronger as a language tool, potentially. I do like some of the language that Claude comes out with when you're talking about content, but for me, it's ChatGPT. It just performs the best across the board.

LEONIE: And so, are there any other exciting GPT tools that you've seen out there that listeners might like to go and have a look at? 

ADAM: Yeah, yeah. Look, I'd probably recommend to people, definitely, definitely have a look at the Canva GPT. I think that one's, I think that's really, really interesting, actually.

That's quite useful. But if you're talking about productivity, yeah, probably have a look at Scholar GPT. That's quite an interesting one.

There's some interesting ones around programming. So, there's a GPT called CodeCopilot, which is quite an interesting one. When you're talking about education, there's one I've been playing with lately called Universal Primer, which is a really open-ended sort of way to learn pretty much anything. It's hard to describe, but basically, they've kind of developed a learning platform, I guess, a learning framework, and you can kind of apply whatever subject it is that you want to learn to that framework, which is quite interesting. 


Custom GPT Timeframes

LEONIE: Okay. So, if someone wants to have a custom GPT built for their business, what's the typical timeframe for a GPT project, if there is a typical timeframe? 

ADAM: Yeah, look, it's a, I mean, it's a very, it depends on the scope of the project.

Some of them, I mean, as I say, I have customers out there that are using JPTs that are built for them on an AI Primer session, which is one hour. So, they came, did a one-hour session with me and walked away with a tool, and they're using it. Some of them take months to build.

It really depends on the scope, but the sweet spot's probably roundabout, roundabout, roundabout a month, I'd say. That's probably the sweet spot, yeah. 


Contacting Adam

LEONIE: So, I'm going to put up your contact details so that anyone who is interested in creating a custom GPT can arrange a chat for you. Our video people will be able to see that. So, would you like to let people know what your website address or other sites that you'd like them to go and see are? 

ADAM: Yeah, certainly. So if you have a look at my website it’s www.bizualize.com.au, and that’s Bizualize is B-I-Z-U-A-L-I-Z-E.com.au. So, on that front page there, I've got all of my different types of training based on whether you want one-off training, team training, or ongoing training. 

LEONIE: Fantastic. Adam, thank you so much for sharing your insights with us. Look forward to talking to you later. 

ADAM: My pleasure. Thanks very much, Leonie.



People on this episode