Build a Buyer's Blueprint GPT with ChatGPT

Full video transcript

How’s it going everyone? Darby here with Gen AI University. In this video, I’m breaking down how to create a custom GPT inside the ChatGPT interface that can follow a series of instructions and commands you give it to produce faster, desired outputs and results. If you have access to a lot of our different recipes, workflows, prompts, and commands inside our community, you’ll notice with the demonstration I’m sharing today that you can replicate what I’m doing here with any number of our existing prompts and recipes, and also how you can build your own.

The example I’m demonstrating is our Buyer’s Blueprint, which is by far top three, if not number one, of our most utilized workflows. This workflow originally came from a collaboration with John Benson and his copy pro marketing blueprint, using Jasper at the time around October/November 2022, right before ChatGPT hit the scene. We were using Jasper to generate this entire blueprint, executing each command one at a time, but stacking them on top of each other ultimately created a full, robust buyer blueprint quickly.

I’m going to show you how I’m building that recipe — which we spent countless hours crafting and developing prompt by prompt — into a custom GPT. I’m in my ChatGPT4 interface. If you have a ChatGPT4 account, you’ll most likely have access to “My GPTs.” I’ll click My GPTs, which opens the section to search other GPTs I’ve created or create my own. Clicking “Create GPT” opens the GPT builder, where I can interact using natural language to give instructions, and it starts configuring the icon, name, description, and instructions. It comes up with conversation starters, and I have the ability to upload additional content to a knowledge base — files, specific data from my company, or something like an article or PDF I want the GPT to reference. You can also select whether you want web browsing, DALL-E images, or a code generator, and go further with actions, which I won’t get into today but which let you build on top of a baseline GPT.

To start, I want to note that any number of our workflows, recipes, and prompts you see out there can be turned into instructions for a GPT you build yourself. The benefit of a GPT you just run with a basic input to get an output is that you don’t need to copy and paste a bunch of prompts every time into a new ChatGPT thread, Claude, or another tool. It’s all hosted right here — which could also be a con in case ChatGPT and OpenAI ever have issues, since we’ve seen a lot of change happen fast in this industry. So keep in mind that anything you’re creating in terms of natural language enhanced prompts and command stacks should be saved somewhere else. I prefer to save mine into a platform called Notion, which is what I’m showing here — a board we’re developing to host more advanced trainings and resources. I’ve created a database in Notion hosting the development of this workflow, which might give you ideas for storing your own content. Many of you who purchased our NFT library card have access to our bronze edition library card, which already has many of these workflows and recipes built for you.

I’ve already gone through this process a couple of times to confirm it functions the way I want before demonstrating it. There are two parts: the first half of the prompt is the market segment context we want to give the AI, and then there’s a “start” section — the buyer’s brief, primary complaint, secondary complaints, topics, goals, descriptors, solutions, and all the elements of the buyer blueprint. I’m copying the buyer blueprint’s market segment section first, since this is what I created initially when building this GPT — and this is actually what ChatGPT ended up giving me. I’m showing you how I’m taking what we’ve already created as a pre-built workflow, uploading it, and showing how the GPT builder interacts with it, and then where I add specific parameters.

I paste this into the GPT builder and hit enter. This kickstarts the process of building the GPT from start to finish. Then I fine-tune the configuration to make sure the instructions get the information they need first, and once given that information, it has the right prompt and command to kick off the next chain of actions. It takes some back-and-forth interaction rather than being instantaneous. The builder responds: it’s set up a GPT to act as a world-class marketer and researcher specializing in marketing psychology, human behavior for buying decisions, and detailed analysis. For the name, it suggests “Marketing” — I say no, use the name “Buyer Blueprints,” version 4.1. It updates the GPT and then asks what the logo should be. The most important part of starting this thread is that we’ve already given it the buyer blueprint — you could start from a completely blank slate and just describe what you want, and the AI will help get you there.

I don’t like the first logo, so I ask for something more simple and minimalistic, neon, with a microscope. If you want to create your own buyer blueprint, all you need is the source code, which community members have access to through our resources — you can modify it however fits best. Once I like a logo, it updates and moves on to conversational style and tone. When it asks which type of products to focus on, I say it should focus on the user’s inputs — this matters because if you don’t clearly direct the GPT on what inputs to expect, it may drift. You could specialize the GPT for only e-commerce products or one brand, for example women’s health supplement products with a specific product catalog, by customizing it here. That’s why the source code for the Buyer’s Blueprint is a useful starting point — you can modify it to work for you in many different ways.

Moving to the Configure tab, I see the name “Buyer Blueprints 4.1,” described as an expert marketer and customer researcher focused on audience analysis and actionable insights. The role and goal section, human behavior analysis of market segments including demographics, and everything I gave it initially are reflected there. Constraints: avoid generalizing or vague output, focus on specific actionable insights, no fluff. Guidelines: follow the structured approach starting with the buyer blueprint, use formats provided in the instructions, use visual and emotional language to drill into the core offer, and provide no-nonsense marketing advice.

From here I go back to my Notion doc, which holds the “start” text from the first time I built this GPT — get clear on the product or service, then once the user provides that, access the internet along with the buyer blueprint focus research. I copy this start section and the rest of the workflow so it focuses top to bottom, paste it into the builder, and save. Saving opens a new thread ready to interact with. I enter “my product is Side Hustle, the party game for entrepreneurs” and hit enter. The GPT should analyze the product, identify the segment and niche, and any additional context, then execute the buyer’s blueprint from top to bottom — the entire series of prompts in this workflow — in one pass. Now I don’t need to copy and paste every time; I just open Buyer Blueprint 4.1, enter product and market segment info, and it generates the elements of the blueprint.

This could be used in many ways depending on your business, brand, clients, products, and services, and how much existing research or data — audience surveys, product data — you could feed into your own version to make the output more geared to what you’re creating. In one shot, it goes all the way down through expensive alternatives, profit, and offer benefits — without needing to copy and paste like you’d do in Claude or ChatGPT previously. It’s now a simple chat interface you can interact with, ask questions, and prompt to continue. Once you reach the end of the buyer blueprint, the next step is deciding what to do with the information — the next series of prompts or instructions, whether creating content, ads, or emails around the blueprint. You can copy and paste those prompts in, or build a variation of this GPT with those frameworks and templates included after the blueprint.

Going back to the GPT, I click edit and add “the digital online version, play for free” as an example of interacting with the builder further. The more advanced you get with actions, the more capable GPTs become — this was just an example of going from start to finish for what a GPT looks like. You can save and copy the link to share with anyone who has access, though they’d need a ChatGPT Pro account, which has pros and cons. You’ve also got the source code for something like this, so I can save the role, goal, guidelines, clarification, and personalization back to my Notion board — so if my GPT ever gets wiped, I can copy and paste it all back in, and it continues down through desires, behaviors, and pain points.

That’s how you turn a prompt workflow into a GPT — how you can use the GPT interface to interact with the instructions you’re giving it, refine the approach, and take an active role in making sure it starts in the right place and executes everything correctly. It’s coming up on 8 o’clock central time, which means the world is waking up and writing emails with ChatGPT, which might be why it’s running a little slower right now — or maybe it’s just my computer. It’s a matter of testing, tweaking, iterating, refining, putting it into action, and getting feedback from the community and from my own use to modify and enhance the Buyer Blueprint GPT.

Throughout this video, we’ve covered how to take an existing workflow prompt recipe and create it with the GPT builder, configure it, and add your own knowledge where it makes sense. In a future video, we’re going to explore APIs that the GPT can use and how to add those as actions to do even more. Hope this is helpful in getting you out there creating your own GPT or seeing where these come into play — there are a number of different directions to expand on this as it’s still new technology we’re able to utilize. Hope you find it useful, keep calm, and keep exploring out there. I’ll see you on the other side.