Since recording this video, OpenAI has made uploading documents a free feature with ChatGPT 4o. So while this video is already a bit outdated, the core concepts around leveraging structured content with AI still apply. I didn't feel like re-recording the whole thing 😆. Enjoy!
I've spent a lot of this semester playing around with what I call AI writing systems - integrating generative AI into various writing technologies throughout the content creation process. I've been using my classes as a kind of playground, testing out different approaches.
And honestly, a lot of them went wrong, didn't work, or were really difficult to handle. That's why I haven't written about it much. But I wanted to share a few very simple ways to start thinking about your own AI writing system.
A key component is connecting generative AI technologies to your own personal structured content and/or curated knowledge base. That's the first step. After that, you need to be able to connect that to various other apps, not just ChatGPT.
To really start thinking about AI within a more systematic context, first consider where your content is - your notes, instructional materials, manuals, company data. How can you make that more organized and then connect it to gen AI in useful ways?
The key for me is not to upload everything into AI, but to focus on just the specific content or knowledge needed for a particular task. Create an AI model, bot or chat that focuses solely on that content. Here are three ways I've done that so far that are super simple:
Provide structured content like a syllabus to a chatbot as a prompt. By including course info, outcomes, grading policies, etc. in a consistent format, you can essentially create a customized AI assistant just for that class.
Upload structured documents directly to your AI tool. Tools like ChatGPT now allow this for free. I also showcase Triplo AI which lets you paste in content to train a custom "mind" that you can then engage with.
Use OpenAI's Playground to create an AI assistant by uploading structured data and instructions. You can then connect the assistant to other apps via Zapier or Make to automate workflows.
The common thread is leveraging content that is already organized into predictable components, like a syllabus with units and policies, or documentation with consistent sections. This structured data helps AI understand the context better and enables you to create reusable prompts or knowledge bases.
This is just a sneak peek of how I'm using structured knowledge bases to improve AI performance and automate parts of my writing process. I plan to share more details in future videos and in my upcoming prompt ops course for paid subscribers.
Have questions about AI writing systems, knowledge bases, or structured content? Leave a comment and let me know what you'd like to learn more about! And stay tuned for a deeper dive on what exactly structured content is and what types work best with AI.
3 Easy Ways to Create AI Writing Systems Using Structured Content