How ChatGPT can help enhance your design documentation process
ChatGPT creates amazing first drafts of design documentation
Photo by Tara Winstead: https://www.pexels.com/photo/person-reaching-out-to-a-robot-8386434/
Despite creating a weekly UX article for the past two years, creating design documentation isn’t my favorite task. It often involves flipping between a Figma file and a Word editor, deep research into formatting, and much more.
So when I needed to write design documentation to help designers understand when and how to use Data Visualization, I tried to see if ChatGPT could do the heavy lifting.
If you’re not aware, ChatGPT is a Generative AI tool sweeping the internet with how quickly it can generate code, write essays, and, in this case, write documentation.
However, was it suitable for Design Documentation? Sort of. Here’s what works and doesn’t work with using ChatGPT for design documentation.
ChatGPT is excellent for creating tables of data for documentation
One of the best parts about ChatGPT is the ability to create a data table that fills in different states, palettes, and more.
This sounds like a minor deal, but it’s a huge timesaver. For example, let’s say I needed to create a color palette that shows how a component looks in the following states:
Default
Hover
Clicked
Active
Focused
Disabled
I also don’t just want this for one color. I want this for three:
Background Color
Border Color
Text Color
This helps me also create styles in Figma once I’m further along and see how things work. However, that’s not the only thing it’s good at.
ChatGPT is excellent at giving relevant examples
Can you give me three examples of when you should use a dropdown menu?
That may be tough to think of off the top of your head. However, this is something that ChatGPT can help with and is one of the most common ways to use it.
Generative AI, like ChatGPT, is often best suited for writing copy, error messages, and other writing-based design tasks. As a result, it’s excellent at coming up with best practices, use cases, and examples for when to use something.
This ties into ChatGPT’s best major strength.
ChatGPT is a virtual assistant that can do some research for you
From Jarvis for Iron Man or Jamie for Joe Rogan, we all love the idea of personal assistants and how useful they can be.
The idea of somebody always online and looking for information or relevant research to answer your questions is revolutionary. This is one of the greatest strengths that ChatGPT provides.
It can scour its database, find relevant information, and pull it up as research or reference. Whether you need to talk about the POUR method of accessible design or best practices around line charts, ChatGPT can quickly reference data and spit it back at you.
Here’s why.
ChatGPT can quickly generate crap that sounds right
ChatGPT is not an omnipotent AI (yet!), which is why you can’t completely trust what it spits out.
The main problem is, ChatGPT can only regurgitate what it finds. If what’s in its database is 20 years old (or inaccurate), it will spit that out and make it seem like that’s the truth. So if you do not know about a subject and rely on it to generate knowledge, others will call you out.
The old saying, Garbage In, Garbage Out, applies more than you realize with ChatGPT.
You won't ask the right questions to get good documentation if you don’t know what you’re looking for or why something should be designed a certain way.
Even if you stumbled into something decent, you couldn't edit it if you don’t understand the topic well enough to know when something’s inaccurate.
Therefore, the words, Generative AI, are essential to remember. ChatGPT can generate tables, microcopy, and more things that can be useful, but it requires human judgment and design skills to deliver something polished that provides value.
However, that doesn’t mean that ChatGPT is not worth it.
ChatGPT skips past blank canvases with decent first drafts
The main thing that you need to keep in mind is that you create design documentation for other human beings. Whether Engineers need additional context or other Designers need rationale, design documentation is useless if no one is willing to readable.
ChatGPT made getting to a first draft a thousand percent quicker for me. That is already valuable, but it also got me mostly correct documentation and components (like tables) that I could test.
It wasn’t so reliable that I could leave most documentation to it. Instead, using it as a tool allowed me to skip past what was troubling me, the blank canvas, and spend more time analyzing the details with Figma (and similar programs).
As a result, I spent my time checking the following things:
Do I recommend the generated color palette (probably not)?
Do I agree with the best practices written there (usually ok)?
Is the wording clear and understandable (usually not)?
etc.
Whatever ChatGPT generates needs edits and refinement, so it’s not an alternative to writing documentation. But if you’ve ever procrastinated writing design documentation because it’s hard to start, consider generating a draft in ChatGPT.
It allows you to avoid second-guessing how to get started and focus on one of your more significant strengths as a designer: paying attention to the details and understanding if things make sense to your user.
Kai Wong is a Senior Product Designer, Data-Informed Design Author, and Data and Design newsletter author. His new free book, The Resilient UX Professional, provides real-world advice to get your first UX job and advance your UX career.