The recent launch of the ChatGPT AI agent builder represents the biggest update yet, enabling anyone—even a complete beginner—to construct AI agents and automations.
These agents can automate complex or tedious tasks, potentially saving you time or even creating a new revenue stream. A highly recommended use case for this new tool is turning your existing, valuable ChatGPT projects (like a custom script writer) into standalone AI agents that you can sell or share with your team, friends, or family.
Step 1: Access and Prepare Your Existing Project
1. Open the Project
Click into the specific ChatGPT project that you wish to convert into its own standalone agent.
2. Copy the Instructions
Locate and click on “edit instructions.” Copy the instructions that you have previously provided for this project.
Step 2: Begin the Agent Creation Process
1. Navigate to the Agent Builder
Return to the agent builder interface.
2. Start a New Agent
Click “start.”
3. Name the Agent
Add the type of agent you want to create (for example, a “YouTube script writer”).
4. Input System Instructions
Replace the default instructions with the instructions you copied from your existing project. For example, the instructions might be: “I want you to be a YouTube expert script writer that builds me scripts based on the scripts from my past successful videos. Every single response should make a video that is 10 to 12 minutes long.”
Step 3: Configure Settings and Add Context
1. Refine Agent Settings (Optional)
Within the agent builder, you have options to add context, include the chat history, change the model if other models yield better results, or change the reasoning effort.
2. Upload Knowledge Files
If the value of your existing project is based on a file you uploaded, rather than just the system instructions, you need to add this file to the new agent. Click on “tools” and then “file search.”
3. Locate and Upload the File
Search your computer for the relevant document (e.g., “script outlines 1 through three”). Drag and drop the document into the agent builder.
4. Attach the File
Attach the file so the agent has access to this proprietary knowledge or data.
Step 4: Preview and Test the Agent
1. Preview the Agent
Once configuration is complete, click “preview” to see if the agent works.
2. Test with a Prompt
Use the agent the same way you would have used the original project. For instance, input a title and prompt the agent.
3. Review the Output
Observe the agent’s actions, which might include searching through script examples and files, reasoning, and then building out the script exactly as you want the video scripted.
By training the agent with your proprietary information, knowledge, and data, you create a powerful tool that you can then give away or sell.