Useful AI for Experience Design
With the boom in AI investment, I think it could be directed better when it comes to UX design tools. It could focus on saving us time instead of replacing us.
When I see most announcements about new AI features for new or existing tools aimed at designers—I feel like they’re missing the mark. Too often the reaction I see from fellow designers is repulsion and fear for their future. Too often the tools are being created to “make a whole design” from a prompt. In this situation, they’re expecting designers to just throw out the terrible generations and keep the one(s) they deem good.
Maybe for a PM or Executive this sounds ideal but for designers, it’s not what we want to spend our time doing. It’s more like being made to manage a random-but-fast colleague. Business owners are most likely to see it as a reason to lay off designers and replace them with AI Design subscriptions. They won’t realize this is a bad idea until jobs have already been lost and their products are already suffering.
So what WOULD be useful for designers?
Well, let’s look at what we spend most of our time doing. The following is a long list and I wouldn’t consider any of these things to be “design activity” even though they’ve been our responsibility for years. Are these tasks creative? With the exception of some naming tasks, no. Are they things we want to be doing? I certainly don’t enjoy them. For these reasons, I think they’d be prime candidates for AI to do for us.
Flat design
- Copy pasting text
- Copy pasting objects
- Switching my clipboard manager between Plain Text and not
- Direct selecting
- Reformatting text, images, video for import
- Importing text clips
- Importing photos
- Finding and re-creating logos
- Importing logos
- Scaling — with or without ratio lock, trying to snap to pixel
- Setting up and modifying constraints
- Aligning and nudging
- Rounding corners
- Grouping and ungrouping
- Returning to computer to see the full context of a comment,
reply to a comment, or check the design - Squinting to pick icon replacements from list of tiny thumbs
- Setting export presets
- Exporting
- Detaching components and making custom variations
- Breaking text styles to apply underline, etc. or making another variation
Special for design systems
- Naming objects
- Naming components
- Modifying constraints
- Creating variants to deal with responsive changes
- Struggling against component and variant limitations
- Replacing generic info, then replacing the generic with example
info in the design files - Renaming everything from a mess to an organized scheme
- Taking messy designs and normalizing into components
- Replacing custom colors with style equivalents
- Replacing custom fonts with style equivalents
- Replacing non-components in files with library components, re-
applying all of the overrides manually - Spotting differences between page designs, library
components, and implementations - Creating examples of color styles
- Creating examples of text styles
- Creating text styles for what should be standard variations
(underline, leading, etc.)
Prototyping
- Connecting screen flows
- Connecting loops
- Creating and connecting overlays
- Connecting nonlinear flows
- Swapping links to and from shared components vs individual
instances
Animations
- Setting up keyframes
- Modifying keyframes
- Applying global changes by modifying too many individual items
- Applying changes within animation tool to reflect new decisions
from a static design
How could this work?
It could work very well and make a lot of designers happy. But it would also require more work to get right. Instead of making an app to spit out whole designs, you’d have to make one or more mini-apps to perform exact little actions across many points of context.
For example, the “Copy pasting text” item. We spend many cumulative hours copying bits of text from object to object within a design and from external documents to our design documents. It could be coming from word processors, web pages, spreadsheets, other design apps, etc. in many different formats. We could be pasting them into new text objects, existing components with several text objects, different objects in a screen, multiple screens or pages, or even different designs. When I put it this way it sounds complex, because it is.
But it’s also something that I could tell a fellow designer and they’d be able to complete such a task from my description. This is what makes it something we could tell a Large Language Model to get a similar result. It’s a task that requires context of what is in documents and visible on a screen but doesn’t require worldly or self aware knowledge. This is something that AI could do for me and should do for me.
I think the same thing is true for all of these tasks.
So app creators, give us AI that does these things in our tools.
We don’t want AI that speaks for us, creates for us, or wastes our time generating a bunch of throw away material. We want AI that speed up these time-consuming non-creative tasks that our current tools make us do.
It’s the speaking, creating, learning, and researching that are design activities. We want to keep doing them because that is what’s going to produce the highest quality work and the highest capacity teams.
