Custom GPTs – What are they, what I’ve done, and lessons learned

Getting ChatGPT to follow directions and stay focused consistently can be a bit challenging.

One approach, as I’ve outlined previously,  is to outline your instructions at the start of each conversation clearly. Providing detailed background information and specifying desired outcomes can also guide ChatGPT effectively.

However, creating a custom GPT might be your best bet if you often deal with repetitive tasks. It allows you to tailor ChatGPT specifically to your needs, ensuring better consistency and precision.

Over the past few months, I’ve set up a few and refined their configuration.  It’s not an understatement to say that working with custom GPTs has saved me hundreds of hours this year alone.

In this article, I’ll share with you a few examples of the types of custom GPT’s I’ve been working on and share some lessons learned in hopes that it helps you create something of value, too.

First – What Are Custom GPTs?

Think of Custom GPTs as personal assistants who are trained to excel in specific tasks you assign them. They focus only on what you need, making them incredibly efficient

Essentially, by giving it directions, details, and parameters, you can create a tool with a narrowed scope, allowing you to be precise and consistent in your work with it. 

At this time, creating custom GPTs is only available through the paid version of ChatGPT. 

To start creating a custom GPT, simply log into your ChatGPT account, click ‘Explore GPTs’ on the left menu, and hit the green ‘Create’ button in the upper right corner. 

Once created, your custom GPT will be listed in your menu when you open your account.

Some Examples Of What I’ve Been Creating

So far, I have created four custom GPTs. Each of them are designed to address a repetitive task and help me be more efficient and consistent in my daily work.

They are:

Recap Memo

What is it?

Each time I have a sales call with a prospective client for my business, I create a document that I call a Recap Memo. This document outlines areas such as business overview, their challenges, how to address them, and what success would look like if we do.

The memo’s goal is to help me organize my thoughts and ensure that I have the necessary information to understand the business, the goals, and the desired outcomes best. 

How does it work?

Because I have a physical document and multiple examples of it, I could upload these documents to the custom GPT to provide it guidance of what I was looking for. I then explained what the sections were for, the reason behind it, and the style I’d like the information to be presented in.

If I have a recorded sales call, I will upload the transcription from the sales call to the custom GPT and ask it to extract the necessary information from it and configure the document based on the instructions in its memory.

The outcome can be exported as a document or copied/pasted into a document I can edit and finalize.

Podcast Show Notes

What Is It?

For each episode of my podcast, The ChatGPT Experiment, I must produce show notes to help listeners understand the content.  This includes a summary overview of the episode, key insights, an episode timestamp, information about the guest, and ways to contact me.

How does it work?

Because I have a standard way of organizing the information for the show notes, I can teach the custom GPT what information I need and where to put it. 

Once the episode is finished, I extract a transcript in text form, upload it to the custom GPT and it configures the summary based on the format and instructions I’ve provided to it.  I can then paste the summary to my podcast software and make any edits necessary.

Content Outline Creator

What Is It?

The hardest part of writing for me is staring at a blank piece of paper or screen. I usually know what I want to write about and getting my thoughts out of my head and organized in a logical format can take some time. So I create a custom GPT to help me do just that.

How does it work?

I have instructed the custom GPT to follow a very specific step-by-step process. It asks me one question at a time, and then uses the answers I provide to move on to the next step in the sequence.   Because I have provided it attributes based on how I like to write checklists to help guide it through its questions, the GPT is able to extract meaningful information from me, understand the audience I’m writing for, and organize my thoughts into a helpful outline that I can then use to write from.

Content Editor

What is it?

Whether I’ve reviewing a draft of my work or work done by a client, the ability to remain objective and consistent is key to good editing. This custom GPT has been instructed to review draft articles and compare the work to a very specific set of requirements and writing goals.

How does it work?

Through a combination of configuration instructions and uploaded documents, I’ve shaped the perspective to be a thorough and consistent editor. I use checklists and sample responses from previous editing work to guide it through its process.  The end result is a list of recommendations that I can use to enhance for areas such as voice, tone, and style as well as audience perspective, SEO value, and overall clarity.

Lessons I’ve Learned So Far

As with anything to do with ChatGPT, this is a work-in-progress for sure. However, I do have a few lessons and tips to share from what I’ve experience that may help you fine-tune your customer GPT for better results.

  1. Detailed and Specific Instructions: Custom GPTs rely on specificity. Start by literally telling it its role and job is then work through all the details you need to have it be successful for you. The clearer and more detailed your instructions, the better your custom GPT will perform. Describe its tasks and expectations as if you’re training a new team member.
  2. Use of Documents in Configuration: Custom GPTs have a knowledge base that accepts documents for additional context and detail. This is a great place to upload previous examples output you’re looking for as well as checklists, rules, and lists that it should follow. Uploading example documents helps your GPT understand exactly what you’re aiming for. Always explain how these documents are to be used to refine its output.
  3. Using Step-by-Step Guidance: Using structured instructions in a step-by-step format (e.g., Step 1, Step 2, etc.), helps the GPT follow processes more accurately and consistently. The more detailed the steps, the more consistent your experience with the custom GPT will be.
  4. Feedback Loop: While editing your custom GPT, provide detailed and specific feedback on what you like and don’t like about its work. I actually copy and paste the output directly so that the editor can see what I’m referring to for my feedback. The custom GPT editor cannot see the work or read the conversations that are being created so providing the actual text that you’re pleased with or concerned about is a great way to manage the GPT effectively.
  5. Using Absolute Language for Clarity: I recommend using absolute terms (e.g., “always”, “never”) in instructions to define boundaries and expectations for the GPT’s behavior clearly. The more direct the terms, the more consistent your outcome will be. .
  6. Defining ‘Do Not Use’ Lists: One of the documents I’ve found helpful, especially in the writing and editing GPTs, is including a document of words and phrases that I never want it to use in our work together. I call this the ‘banned words’ document and instruct it to refer to the document when reviewing or creating text.
  7. Preservation and Saving Work: I lost some configuration time due to this so I’m sharing to help keep you from doing the same. Unlike tools like Google Docs where autosaving is the norm, you’re initial work will not be saved until you push the green button to create. Once the custom GPT has been created, it does a nice job of auto updating and you’re providing it feedback. You can always manually update as well after providing feedback if you don’t believe it has done it for you. But simply know that your first set of instructions are not saved until you manually do it yourself.

Relevant ChatGPT Experiment Podcast Episdoes 

The following podcast episodes provide additional detail on topics mentioned in this article. These can be found in your favorite podcast platform or on the podcast website www.ChatGPTExperiment.com

They are:

  • Episode 31: Gets published on Wed May 8 and goes into detail on the topics I’ve outlined here regarding using and refining custom GPTs
  • Episode 10: Focuses on prompting and configuring GPTs, providing guidance on how to direct and manage interactions with GPTs effectively.
  • Mini Episode on Voice to Text: A brief guide on setting up voice-to-text features on PCs and Macs, highlighting the advantages of converting spoken content directly into text.
  • Episode 21: Discusses how to make ChatGPT write in a style that mimics your own, emphasizing the importance of voice, tone, style, and audience consideration in content creation.

Conclusion 

The key to setting up, using, and refining anything you do with ChatGPT is to remain curious and keep giving it feedback. 

I’d love to hear how you’re using custom GPTs and what kind of results you’re seeing from it.

Stay curious!

Cary