top of page

How this agency founder builds custom GPTs—and how you can too

Building your own GPTs for creative tasks can increase your productivity and give you more quality control over the output.

Profile picture of Aaron Gelbman


7 min read

When OpenAI unveiled its GPT Store, it had high hopes. The store would allow anyone to discover and share custom chatbots (called “GPTs”) built with OpenAI’s generative AI tech.

But just two months after its launch, TechCrunch raised a flag: the GPT Store was littered with spam. The store “is flooded with bizarre, potentially copyright-infringing GPTs that imply a light touch where it concerns OpenAI’s moderation efforts…[and that] serve as little more than funnels to third-party paid services,” TechCrunch reported.

This begged the question—can you trust what’s offered in the GPT Store? More importantly, instead of borrowing someone else’s invention, should you just create your own?

Jacob Cass, founder of branding and design consultancy JUST Creative, is a proponent of building your own GPT. “I know that it’s not for everyone and a lot of people are intimidated by it,” says Cass about how some creative professionals perceive AI tools for web design. “But once people are shown how it's done, they're like, ‘Oh, I get it now!’ Something clicks.”

To date, Cass has authored various GPTs—for his own use and available to all on ChatGPT-4—to help with everything from creating visual mood boards to blog posts. In his own words, Cass shares more about his experience with GPTs and how to go about creating your own.

Banner to create a Wix Studio account that says, "AI that transforms how agencies work"

How did you first get started with GPTs?

Jacob Cass: I'm a pretty early adopter of new technology: crypto, Web3 and NFTs. I have a bit of a “shiny object syndrome.” I love exploring new things and experimenting to see how they can influence or impact my career. AI has just blown my mind. Now, it's much more accessible and it's moving at such an incredible pace. It's hard to keep up.

At heart, I'm a designer and am classically trained in visual communication. And as someone specializing in branding, marketing and content creation, I wanted to create a bot that specialized in those areas. So, I created a custom GPT and trained it on those subjects and on some of my own materials—ebooks, articles and so forth—so it could understand my point of view.

How do you use custom GPTs in your workflow?

Jacob Cass: I've created a number of different GPTs (or I prefer the word “bots”) to serve my creative practice:

  • A brand bot: An ideation tool for branding, marketing, design and strategy, like customer personas.

  • A mood board maker: It outputs mood boards or stylescapes and has things like fonts, colors, photography, art direction, color palettes and so forth in an image. In the past, you'd have to scour books, Pinterest, the web, blogs and so forth, and it could take a long time. But [using this bot,] you input some of the characteristics you're looking for and it can generate it on the fly.

  • A minimal logo bot: For simple and versatile logos that send a clear message don’t combine too many things into one. I trained it to just do black and white.

  • A packaging mockup bot: When we're sending things to a client or we're trying to communicate an idea how things integrate together, like fonts, colors and a 3d tangible item, you want to be able to see how it looks and feels, so you can literally type in. You create a 3D, very realistic photo of it in seconds.

A screenshot of the opening window of a custom GPT called the Brand Bot with pre-filled prompts.
The opening panel of Jacob Cass' Brand Bot on ChatGPT-4 for branding, marketing, design and strategy tasks.

How do these custom GPTs help you in your creative work?

Jacob Cass: Productivity has been supercharged. They’ve sped up my workflow incredibly. You write your inputs and instantly get what you need so you don't have to search for it.

Why a custom GPT and not an existing, specialized GPT from the store?

Jacob Cass: Well, first off, is [a custom GPT] needed? Is there something else out there that could serve this purpose already? Look in the GPT Store and see if [an existing GPT] works for your particular scenario. I start by using their suggested default prompts. If it doesn't work, then experiment a bit further and decide to move on or not.

[GPTs are] built on an amazing large language model. So you're standing on the shoulders of giants, really. They've made it extremely easy to create your own GPT, which is both a blessing and a curse, because there's a lot of trash out there and you don't know what's good or not. 

As a designer, I felt [public GPTs] were limited and they weren't trained to my style, my aesthetic and what I wanted. So that's why I went and created my own. 

What is the trade-off between creating on your own from the ground up and using a GPT?

Jacob Cass: In general with GPTs, there are some downfalls because you don't have as much control. But for ideation, this is where I see the biggest benefit. You can input your thoughts and you get an output very quickly. 

Once you know how, it's a matter of putting in the right prompts and the right variables.

For anyone creating GPTs (both first-timers or even repeat creators), what guidelines should they follow for the best results?

Jacob Cass:

  1. Upload training and onboarding material: See if you can create your own GPT with the variables that you want. Upload the data that you want it to use. [For my custom logo bot,] I gave it variables of how it should work and what the output should be: simple, minimal, clean, geometric. This is my style and my aesthetic, and what I think makes a really strong logo.For better outputs, you definitely should train it and say how it should act. I trained it to act in a conversational, inspiring and engaging way to make it a bit more fun. I trained it to ask a question back, for example, if it needs more variables to provide a better output. So this way it gets more information out of the user.

  2. Experiment: Try different prompts, and ask it to do different things. You see the results and then you know what’s working or not.

  3. Tell the AI to simulate a persona: [Give] a direction of who [the bot] should be acting like—[who] you’re pulling information for and you will get better results. [Tell it to] act, for example, as a marketing expert.

  4. Write your task and outline the task steps. If you give it more variables, it will give you a better output. Let's say, create ten taglines and each should be two to six words.

  5. Give it context, such as the product’s core selling point.

  6. Give it a goal, or describe what the task needs to achieve.

  7. Describe the output. What does the output look like? For example, the output should be a bullet point list.

  8. Be mindful when instructing the AI to “do” vs. “don’t” do something. I have found that bots don’t listen very well to things like “do not.” It's better if you train it on how to do things versus not doing things.

  9. Share your bot with others and get feedback. A lot of people have been using [my] Brand Bot and I've received output of how people are using it.

How should we determine where in our workflows to integrate GPTs?

Jacob Cass: What you really need to figure out is where ChatGPT or custom GPTs can make your processes more efficient or more effective—and then reverse engineer that. 

I’ve found so many uses for it. For me, it’s not just for design or strategy. There are so many other aspects; I’m not a developer but I’ve used it to generate code. I’ve used it for SEO, articles and content creation. It's about experimenting to see where it can be used, and then trying it because you learn and then you can improve it. 

How should professionals decide which AI tools to use?

Jacob Cass: It’s about understanding what you're trying to do and finding the right tool to do it. You cannot access or create custom GPTs without the paid version [of ChatGPT] and I’ve definitely noticed a difference in the output from [ChatGPT-] 3.5 to 4. The image creation feature is a paid feature, but you can get around that paid feature by using the same image creation model on Bing. 

There are other tools out there that give you more flexibility and customization versus ChatGPT or Midjourney. You should also consider where you draw the line in the sand regarding ethics.

For example, Adobe Firefly has been modeled on their own stock photography and not scraped from the web, so that could be a better choice for some creatives. But then Adobe has also used its whole library and the people that have contributed their images weren't paid for their contributions, so where do you personally draw this ethical line?

How are you communicating with clients about your use of GPTs for their projects?

Jacob Cass: When it comes to using AI-generated artwork, I find the output is kind of like the new stock photo so I haven't used it [as final imagery] in any client projects. It's more for the ideation process. It hasn't got to a point where I can get final images for a website that I'd be happy with.

[Recently,] I was in a [client] workshop and we had a document that they had worked on in the past, with their mission and vision, and it was all in Italian. So, I uploaded this [into ChatGPT] to translate and summarize, told it to act as a branding professional, and to provide core values and a description of each one. We got some great results, and the client and I were blown away. But before I did any of this, I did ask my client, are you okay if we experiment with AI for this process? 

What type of industry backlash have you seen about GPTs and AI tools?

Jacob Cass: Firstly, there are brands that try to get away with [using AI discreetly], and then there's backlash. BMW did a Christmas ad with reindeer, but one of them had a fifth leg.

Plus, I've lost a ton of followers and subscribers because I talk up AI, and there are illustrators, artists and photographers out there that just can't wrap their heads around it. They're in denial about AI, and [worry it’s] going to take their jobs. I think AI is like horses before cars—you have to embrace it and use it because it's not going anywhere. It's the way things are going.

What do you see next for AI tools in the creative industry?

Jacob Cass: AI video is crazy. What I find fascinating is the storytellers, the creative directors and the producers—and how they put together the angles, the shots and the art direction to tell a better story. They can control the AI better because they know what makes a good story and video. That’s really only possible if you have the knowledge and theory beforehand. 

It’s AI plus humans. That’s what it comes down to, not just AI. 

Subscribe to the FWD newsletter for more stories from industry experts in your inbox.

Find new ways FWD

Thanks for submitting!

By subscribing, you agree to receive the Wix Studio newsletter and other related content and acknowledge that Wix will treat your personal information in accordance with Wix's Privacy Policy.

Do brilliant work—together

Collaborate and share inspiration with other pros in the Wix Studio community.

Image showing a photo of a young group of professionals on the left and a photo highlighting one professional in a conference setting on the right
bottom of page