Build Your Business Brain GPT in 60 Minutes

Full video transcript

Hey everybody, Darby here with GenAI University tuning in for today’s live show of GenAI Live, the AI Implementation Power Hour, where we’re going to be diving into hands-on real-time buildouts to leverage AI in our businesses. Starting with our premier episode today of building out your AI right-hand brain, your own custom GPT in ChatGPT that is trained on all your business information, your knowledge, your context, your frameworks — so that you can leverage ChatGPT to its fullest potential, unlock more time, and make more money in our businesses by leveraging these AI tools inside of our own workflows.

We’re going to be joined today by members of our community who are tuning in live, and we’re going to be walking through an actual demonstration of building out a new GPT with all the relevant business context and information for one of my own businesses, showing you step by step how you can go ahead and implement this yourself. I was organizing a lot of the content here before we went live, which is really the first fundamental thing you need to do when going through this process — actually have your content organized, or else you’re going to be lacking in the context you’re giving ChatGPT (or any large language model, for that matter), and the outputs will not be as ideal as you may hope.

We got an hour for this show, and those that join live get a chance to ask questions and engage; those who watch the replay can ask questions inside the community. Make sure you’re joined inside our new Skool community at genaiu.com/skool.

Custom GPT inside ChatGPT means we’re training it with our own business information, our own context, and our own specific topic-related knowledge — what we want that version of ChatGPT to do — so we can tap into it again and again. There are definitely levels to this that you can get more advanced with over time, but there’s also a level of simplicity here that I think a lot of people overlook, thinking they need to be more technical. I’m going to show you the core pieces I’d recommend just to get started inside this one-hour sprint, so you can start leveraging a custom GPT trained on your own business context and information (or for a client you’re working with) to get more on-brand, consistent, quality outputs — instead of retraining a new chat thread every single time.

I’m calling this your own right-hand brain, your right-hand GPT. You may hear people call it a business brain or a bunch of other fancy names — that’s what marketers do. At the end of the day, what I’m going to show you is how to build a simple GPT that understands your business, your brand voice, and your goals, so you can use it again and again and build on it from there. By the end of the session, you’ll have a fully trained GPT inside ChatGPT (you’ll need a Plus account). It’s going to act like your daily AI teammate, helping you brainstorm, write, plan, create, and automate certain aspects of your business, using the knowledge we’ve trained it on as a source of truth.

We’re going to talk about setting up a custom GPT inside ChatGPT, training it on your business info, frameworks, and tone, and turning it into a reliable right-hand assistant that thinks with you. AI is a tool — it’s a large language model — and the more focused, specific information you give it about you, your business, and your goals, the better it will be at helping you achieve those goals. If you don’t give it any context or direction, you’re going to get generic AI content out. That’s just how these tools work.

There was prerequisite work for community members getting access to these calls live, and pre-work I’d recommend anybody do before building a custom GPT: gathering all your business information and context. This might take an hour, a day, or a week. You might have teammates with access to all this information, but it’s spread across 15 different Google Drive folders — I’m just telling you what I’ve seen. If you don’t have this information organized in a clear, easy-to-access way, why would you expect AI to be a magic bullet that suddenly understands where all this information is coming from? It’s not going to. It’s up to you as the human in the loop, as the operator behind these machines, to gather this information, identify the most relevant context, organize it into one simple, clear folder, and use that as your starting point to build from.

Let me illustrate this quickly: ChatGPT is the central knowledge point, and you’re right here operating it — like a right hand. But by itself, ChatGPT knows a lot of general information, which is super broad and not focused. When we’re talking about your business, your goals, and the outcomes you’re looking to leverage AI for, we need it to be more specific. So it’s your job as the human operator to gather information — think of it like going to the library to find everything you want to learn yourself, but also teach your AI assistant. Maybe you follow a book like Traction and the EOS operating system — ChatGPT has general knowledge of that, but did you know you can also give a custom GPT book manuscripts so it follows those frameworks more exactly? There are tons of different books, processes, frameworks, and instructional workflows you’ve probably already built in your business, or that your clients have already built, and it’s your job to gather that information — wherever it lives, in Google, in Notion, in a library, or something you need to create. If you follow what we teach here, building your own buyer’s brief is a great starting point, which is what we’re going to walk through for the foundational content and knowledge.

Once you gather this key info, you take your ChatGPT instance and say, “You’re nice, but you’re super general for me. I want to build a custom GPT that references all of this key information.” That becomes a right-hand teammate — not a first-time employee who showed up with no information, no context, no know-how, and no frameworks, who you just told “you’re hired, go.” What would a human say in that situation? “You can’t expect me to do a good job if you’re not going to give me the tools, resources, information, and direction I need.” AI is no different — it will act like it knows what it’s doing and go fast in a direction you didn’t want. So we’re going to take a custom GPT and build it out with all this information feeding into it, and then you’ll be able to update that information over time.

Now it’s time for the rubber to meet the road. On the left-hand side of my ChatGPT account, I select “Explore” under GPTs. This shows a bunch of custom GPTs I’ve made — some for different clients, some for internal purposes. Marcus, one of our more popular free GPTs, was trained on our buyer brief book. But the thing we want is “Create” in the upper right corner, which opens the interface we’ll spend the most time in. Right now we have a blank Configure interface: you can give it a name, a description explaining what it does, and instructions. Instructions come in the form of prompts — simple, direct instructions to an AI system. We also call these “recipes” when they’re more comprehensive, with multiple layers of prompts strung together. You can create your own frameworks and place them in as instructions so the GPT executes them — for example, Alex Hormozi’s $100 Million Offer framework. We’ve created a recipe in our library based on that book, essentially all the instructions from the book that help you create your $100 million offer. You can do the same with almost any business book that lays out “my framework for doing X, Y, or Z.”

Starting from a blank slate can be overwhelming, so from here, turn your attention to the “Create” section — this is where you work with ChatGPT to uncover what you actually want to make, and where you provide information that helps it understand exactly what to build. I’m going to share the simple knowledge base I put together leading up to this workshop, for the next evolution of a custom GPT for my side hustle company, since we’re expanding our live events, opening a Skool community, and hosting new things we weren’t doing two years ago.

The key pieces of information I’m bringing in: first, a brand style guide — colors, tone, images. You don’t need a super pretty PDF, but having hex codes and colors is helpful so ChatGPT follows that style guide when generating images. Second, a framework snapshot of what a successful event page looks like, so when I say “create an event for this and this at this time,” it references that structure and rewrites the copy appropriately. Third, our game instructions, since the pitch about how the game works varies by audience but the fundamentals stay the same. And last, the most important piece: our buyer’s brief — also called a strategic brief or creative brief. This is the source of truth outlining what your product or service is, what market segments you serve, your niche, and the surrounding context. You could use ChatGPT or Claude to help build one of these top to bottom using the buyer’s brief book and the Market Sauce 9000 recipe — I was doing this with Claude earlier today, and it’s insane how much better the outputs are now compared to when we built that recipe three years ago.

The point of a creative brief is having one document you can send to anybody so everyone gets on the same page about who you are, who you serve, and how you do it differently. Without one, you’ll either repeat yourself constantly or lack clarity on your marketing goals because it’s not all on paper. At minimum, a creative brief covers your brand’s promise, the problem, the solution, and your audience segments. My own version ended up being 96 pages after AI helped me build it out — I was actively reading it as it produced outputs to confirm the information and context were actually good. Sections included a branded style guide, imagery style, and brand tone, so the GPT understands the personality you want reflected across social media, emails, and everything else. All of this was originally created in Claude before going into the document I’m now uploading to ChatGPT. Your document could be as simple as five pages covering your brand tone, style, and vibe — it’s a living document you continue to update.

So this is all inside a Google document, and now I need to tell ChatGPT what I actually want to make. This is work you should do before you start building a GPT — spend an hour, a day, or a week gathering your brand information, style, and tone of voice (or have your team do it), so you have everything you’d give anyone new coming into your business who needs to know what you do and who you serve. I drag and drop the files in as context and give it a starting prompt: “I want to create my very own AI right hand, a custom GPT that understands my side hustle entertainment business with my brand and my goals.” It took that information and added it to the knowledge base directly.

I give it more context by voice-to-text: “I’m creating an AI right-hand assistant that understands all of my company’s information and context and can work with me as a collaborative content planning, writing, brainstorming, and assistant to help me with creating material for our marketing, sales, and operations.” From here, the most important part of building the GPT is that it collaboratively works with you on what it needs from you. If this is your first time through this process, just follow along with its prompts — it’s designed to help you build it using its own frameworks and instructions.

I name it “Side Hustle Sally” — a friendly, collaborative, cooperative, creative personality that understands our brand information to help with creating content. I’ve given it the knowledge base, my brand voice, company info, game instructions, and event types, with a primary focus on marketing research, sales outreach, and event creation and communications. I already have a brand voice section written out in the buyer’s brief — “If Side Hustle were a person, they’d be a charismatic friend who shows up to the party with a game no one’s heard of and somehow it becomes the highlight of the night” — so I copy that in directly rather than rewriting it. I throw this back to the GPT builder: friendly, collaborative, creative, a charismatic connector who mixes entrepreneur smarts and playful curiosity, confidence, and warmth, never stiff or corporate. This tone influences the content it creates around knowledge, brand voice, events, internal systems, marketing, and sales ops — warm, welcoming, playful, inventive, confident, visionary. As you go through this yourself, think about how you want your content communicated to the world, whether social media or email. If it came back too corporate, direct, and abrasive, I’d say that personality doesn’t belong on this team and adjust it with additional prompts.

The builder then asks whether it should also generate and refine marketing research and trend insights — analyzing audience psychographics, spotting event trends, summarizing competitor activity. I say yes, and lock in the full configuration. While it updates, I create a profile picture — pulling a brand mascot image (a unicorn entrepreneur character) into Canva and uploading it as the icon, on a brand-color background. I can easily rename the GPT at any point too.

Now it has a full section of instructions describing personality, archetype, and mission, plus auto-generated conversation starters relevant to planning the next event. In the sidebar I can see the knowledge base: game instructions, brand guide, event framework, and the creative brief. I could start with a simple five-or ten-page brief covering just the basics — a valuable document to have regardless of whether you’re building a GPT. Compared to a regular ChatGPT thread, where I’d have to copy and paste all of this knowledge in every time, this is all hosted and referenced automatically.

There’s also a “Recommended model” setting I hadn’t seen before — you can leave it on “no model recommended” so users pick their own, or select a specific one and test it. This matters: when ChatGPT-4o was the main model and OpenAI released GPT-5, some people missed how 4o interacted with them, while others preferred that GPT-5 took longer to think. Under the GPT’s name you can choose auto, instant, thinking, or pro (“research grade intelligence”), plus legacy models — I haven’t experimented much with 5.1 yet, so I’d recommend testing which one actually produces the outputs you want for your use case, since I can’t answer that generically.

For the demo test, I prompt it: “We’re looking at booking a few events in Q1 for comedy shows related to South by Southwest, leading up to that event at some venues downtown… create an event for our Luma page for an upcoming show on January 31st in downtown Austin, live comedy show from 7 to 10, doors at 7:00, show starts at 8:00, networking after — go ahead and create all the assets for this event and all the social media content we’ll need.” Without me re-supplying any context, it searches the knowledge base and even proactively suggests something for the Skool community, since I never mentioned it but it’s referenced in the uploaded files. It’s referencing the information I gave it for event details, though I do go delete some emojis it added.

To fix that, I go to the upper-left dropdown, click “Edit GPT,” and first rename it from Side Hustle Sally to Side Hustle Derby, since these outputs are coming from me. I add explicit instructions: for every single output, never ever use emojis, never ever use em dashes, and review outputs before finalizing to confirm those rules were followed. I run the exact same test prompt again on the renamed GPT — zero emojis this time. It’s a simple demonstration, but it shows how a few lines of explicit instruction can influence output. Think about any signature sign-off you always use in emails — if you wanted that included every time you reach out to a sponsor, you’d add that blurb to the instructions too, so it understands how you communicate.

Someone in the chat compares AI to a mirror, reflecting back what you give it — I agree, and note that’s worth an entire episode on its own: knowing when to step back if it’s reflecting too much unhelpful context back at you, or being overly agreeable about every idea. If you have a framework for approaching problem-solving, or training a new employee with an SOP, giving the GPT that same information is fundamental to it handling similar scenarios the way you would.

For a second test, I ask it to do deep research on potential sponsors to reach out to for the event — find sponsors, websites, contact info, LinkedIn, and specific pages to reach out via email or phone. I still need to direct it on what to do, but it goes and does that research. There are still limitations with a custom GPT built this way that could be unlocked with more advanced setup — for example, I didn’t find an easy way to directly connect this custom GPT to my HubSpot account via MCP (Model Context Protocol), a way for AI tools to communicate like APIs to pass information back and forth, though HubSpot has recently released its own way to integrate directly with ChatGPT. The foundations are set regardless: the knowledge base and instructions are doing the work. I can let this run in the background while doing something else, and instead of spending 30 minutes to two hours manually pulling this together for every side hustle event, I can pop this GPT open and get the event, assets, and content in a fraction of the time — freeing up time for the parts of running these events (which help more people discover Gen AI University) that I don’t otherwise have 20 hours a week to spend on.

The action items from today: gather all of your knowledge, information, and company context — brand voice, tone, style guide — and get a buyer’s brief together. Follow the Market Sauce and buyer’s brief book, and the Market Sauce 9000 recipe, available for free in our foundation training inside the Skool community at genaiu.com/skool. Use that as a foundation, put together a creative brief, and combine it with your other company information to build your own GPT. Even a super simple version built in one hour, like today’s, is a skill you can build on any time you want a more advanced GPT or assistant, because you’ve already gathered the information and created the baseline instructional prompt you can copy, paste, and upgrade as needed. Ultimately, this saves time so your right hand can do this work while you focus on higher-leverage work.

Answering a question about virtual teams: yes, you could have ten different GPTs each with hyper-specific outcomes — everybody should have at least one version of a GPT that understands this type of foundational business context. Another example useful for clients: take your top ten performing YouTube video scripts, upload the transcripts to ChatGPT or Claude, ask it to identify the framework that models that success, and create a prompt for writing an outline in that format. Then upload that outline and framework to a content-script-writing GPT you can use anytime you need a script in that format, following the same approach with extra context for guidance.

On the point about this being “so much better than building a project” — projects have their place for certain things, but there are differences here, and fundamentally with large language models it comes down to understanding context, or what people now call context engineering — we’ve been doing this for four-plus years. You just need the right information you’d teach a human about your business. Organizing that first is surprisingly the place a lot of people skip, then they come into AI and wonder why it’s so generic — you’re not giving it the information it needs from the start. Spend the time to get your knowledge base organized first, add in one solid framework you’re already using day to day, and get your GPT made by collaborating with the custom GPT builder. There are levels beyond this you can go down, but without a good context foundation, going two or three levels deeper is a waste of time, energy, and resources that could have been saved by putting the knowledge base together first.

Thank you all for tuning in live for today’s kickoff show of the AI Implementation Power Hour. Replays will be available inside our classroom in Skool — make sure you’re subscribed at least for free to get access to our foundation training and future shows. Keep coming AI on. I’ll see you on the other side. Cheers.