• UX for AI
  • Posts
  • AI for UX: Figma and the Gods of Hammers

AI for UX: Figma and the Gods of Hammers

In case you missed the news last week, Figma launched its new “Make Designs” AI and then had to pull it out almost immediately because everything it made looked exactly like an… (GASP!) Apple app. I think that is, at best, a distraction. Here are 7 use cases Figma should focus on instead and what UX Designers should focus on in this critical transition for the UX industry to avoid becoming “the gods of hammers.”

Are you Thor, the god of hammers? Source: Midjourney.

“Are you the god of hammers?”

In the movie Thor: Ragnarok (Marvel Studios, 2017), as Thor is about to die at the hands of his sister, Hela, he has a vision of talking with his dead father Odin:

Thor: She's too strong. Without my hammer, I can't...
Odin: Are you Thor, the god of hammers?

Figma is our hammer. 

Granted, a pretty nice hammer… And many a designer has a “special relationship” with it: https://youtu.be/xG_g2b4WTc4?si=K4czSf0Spj60q2TT  …But at the end of the day, that is all Figma is: a tool to channel your design power. 

Figma is not the source of your power, no more than Mjölnir is the source of Thor’s power.

It took Thor being at death’s door to have this revelation and to be reborn as the God of Thunder. The UX industry, too, is about to die in its current form.

And it is up to us to wake up to our deeper truths: 

Our hammers, no matter how good, are just tools. 

At our core, we are self-starters, scrappy Macgyvers, true creatives bursting with original ideas, driven by empathy and a powerful desire to serve. 

Deeper than the surface of the Figma pixel-pushing and auto-layout nesting lies a solid core of rapid experimentation, an unquenchable thirst for creativity and originality, a firm grounding in research, respect for the power of technology, and a love for humanity.

At our very core, we live to empower others. With hammers or without.

You can watch the spectacular scene from the movie here: https://youtu.be/KQvhX-3CkJM?si=fL5YmAXVueVdHM11

Now, let me switch gears to the following incredibly pompous and shamelessly self-aggrandizing critique of Figma’s latest AI-driven feature fiasco. Naturally, I wrote it as an open letter to Figma, the company that creates and maintains our favorite design hammer. It is not a critique of any one individual or team but more of the overall approach to AI projects that currently permeates not just Figma but the entire UX industry. Please don’t take anything I say personally – I am genuinely trying to help us all get better at AI. Here goes. 

Dear Figma

Some might say that your ill-fated play for the “text prompt to complete App Design” (aka “Make Designs” feature) is a sloppy, lazy, ill-defined waste of resources born from a lack of understanding of both – what your customers need and what the current state of AI is capable of. 

However, personally, I’d like to chalk it up to a well-intentioned experiment. 

After all, the alternative would be that you are trying to compete with such Johnny-come-lately’s as Uizard, the company targeting a completely different use case: 

Attempted Replacement of UX Designers. 

As far as I can tell, Uizard is targeting cheap startups that lack the vision and resources to hire a designer. (That happens more often than you’d think—in my 20+ years of experience, I have told quite a few startups that they were not yet ready to hire a professional designer and instead needed to code a POC first, raise money, and only then hire the best UXer they could afford.)

Dear Figma, that is surely not you! 

You know which way your bread is buttered. Teams of designers pay you to make their jobs easier, not to replace them.

That is why you should focus not on replacing designers but on providing them with the help they need to deal with tedious and repetitive tasks so that they can focus on—ahem—design. 

In other words, you should focus squarely on Augmented Intelligence

How to avoid these kinds of bloopers in the future

Dear Figma, to avoid future ill-fated attempts to release sloppy AI features that run contrary to your mission, I recommend:

  1. Augment, not Replace: Replacing experts with AI is a red flag. Take the time to research good use cases (I’ve done some of the work for you below). You may also wish to refer to my previous article for more on researching AI-driven use cases: How to Pick an AI Use Case

  2. Keep it scrappy: AI projects are like salami: the thinner the slice, the tastier the outcome. This means you need to have a sharp knife and a ruthless approach to cutting the fat and avoiding scope creep. It is very tempting to listen to Data Scientists who often over-promise the performance. Don’t let the tail wag the pig. Here’s how to stay lean: The Importance of Staying Lean

  3. Ethics matter: Take the time to evaluate and label various prediction outcomes – understand the benefit of correct predictions vs. the cost of false predictions. The thinner the use case and the more you try for Augmented Intelligence rather than replacing the experts, the less damaging the mistakes will be. Here’s a practical value matrix exercise approach to ethics that anyone can do: AI Accuracy is Bullsh*t. Here's what UX must do about it. (Part 1)

  4. Understand the capabilities and limitations of modern AI: Your designers should spend at least 1 hour a day interacting with various models so that they understand firsthand what is possible and what is Data Science hype. AI is too important to leave it to Data Scientists. UX Designers should get involved ASAP and stay involved by asking good questions about use cases and outcomes, helping you create the long-term vision, and ensuring the ethics of your business decisions. Because at the root, ALL AIs are biased:  Transforming AI Bias into "Augmented Intelligence" – a Powerful Tool for a Better World

7 AI for UX Use Cases to Focus on Instead

Dear Figma, instead of ill-fated, over-ambitious experiments of replacing designers, focus on solving their current pains using augmented intelligence. Here are 7 use cases to get you started:

1. Stop trying to replace the design of full applications. Augment the design of individual pages instead.

Pages are much easier to design than full applications. It’s much easier to write a prompt and describe accurately what you need for one page. You can easily make 4-8 versions of a single page. You can rely on existing DSM patterns and components to create a page. You can even have the AI listen in on the user session and change the design automatically while on a customer research call! We wrote about research techniques for AI products here: AI and UX Research.

2. Create realistic content. 

This should be a no-brainer. I spend at least 30% of my day or more looking for realistic data for my vision prototypes, and I use ChatGPT to come up with realistic content. Then, I am forced to painstakingly copy it into my prototype, one table cell at a time. This is exactly the kind of thing you can easily automate! 

Learn from the customer’s own collection of prototypes and create custom content that fits the company’s needs to a T. Understand how to create custom IDs and time stamps in the right format, as well as industry-accepted content like IP addresses, URLs, e-commerce thumbnails, fake people portraits, etc. Fill out the prototype content by default using the table header to create the right sort order without waiting for a prompt.

Also, provide the means to change the selected batch of content based on a text prompt. For example, select a column in the table that is labeled “Date Time” and make it responsive to a head attribute called “sort.” Also, allow me to prompt: “Make all these dates occur in the last 10 min, LIFO sort.”

OK, I know there are Figma plugins that do that. The point is it’s all added friction — it’s past time it should work like that by default, and it should be based on previous designs my team has done. Otherwise, it’s just more crap I have to deal with. 

3. Create visual mood boards and skin pages.

Don’t rely only on text for visual mood boards! That’s just silly. You are a visual design tool – act like one! Allow me to use a bunch of images and text together to create an AI-generated mood board. Then give me 6-8 versions of a few pages to reskin using what you came up with. Allow me to add more images and simple text prompts like (“Warmer,” “more purple,” “more gradient,” etc.) to change the outcomes. Allow me to select visual things I like in your output and point out things I want you to iterate on. Yes, that might require some new interaction design, and you might need to get (GASP!) CREATIVE! The most powerful combination is text + visuals. You have a unique chance to do both well.

4. Go directly to React code. 

You should provide a direct and tangible connection between design and React code. As I wrote recently in Open Letter to Junior Designers: “Code talks and bullshit walks.” Pictures are just bullshit until you put them to code. With AI, this picture-to-code connection should be a no-brainer.

My advice?

Stop playing with the auto-layout and, instead, go directly into the React Flexbox layout. UXPin does a better job of this than you do. That is something you should probably pay attention to. Or not.

5. Fix the DSM mess.

Speaking of DSM, Figma – please fix the mess! DSM components that do not match reality. Reality that does not match components. Designers detach and create their own components with no regard to previous design patterns, using random font and color choices. Cats and dogs, sharks and hamsters, all living together… You can fix all that. 

At least, be the blind-spot indicator: warn designers and notify the design leadership when rank-and-file designers are about to make stupid decisions. Or better yet, guide the Junior UXers on how to fix their designs with React-based components coded by AI. If the DSM components are “real” (by which I mean “code-based”), then designers will be less likely to create bullshit designs, and developers will be less likely to recode the whole thing AGAIN to match the bad design, thus perpetuating the mess. 

6. Create a viable accessibility solution.

As Jakob Nielsen wrote so eloquently here, accessibility failed, and AI is our best hope to fix it. Make an automated AI to test components and pages for accessibility. Better yet, create a function whereby the current DSM components can be converted to fit a particular accessibility challenge: vision impaired, large text, improved contrast, cognitive decline, and the like. With AI writing the code, we no longer need to be tied to a single UI. All UIs can now be individualized to customer needs. You can usher in this new era of love and harmony… Don’t miss your chance.

7. Solve the tables already!

Tables suck. Everyone makes their own, and usually badly. Ditto for simple forms of all kinds. There is simply no need for that. You can solve it easily as I described here: Getting Ready for AI-pocalypse: Short-Hand UX Design Notation as AI Prompt. 

ChatGPT already does a great job – see below. Now, all you need to add is my custom DSM components (headers with sort, cells with renderers, etc.) and go directly to a React table in code. We love you, Figma, but with capabilities like this already on the market, no designer should ever again be forced to assemble another table cell by cell by cell by cell by cell by cell by cell:

Source: ChatGPT, Collected July 17, 2024

Finally, Dear Figma 

All those improvements will directly help designers return to being designers. Instead of glorified pixel-pushers, designers will once again be able to make the time to talk to users and stakeholders, understand the use cases, and tie technology advances to user satisfaction, long-term loyalty, and the business bottom line.

Stop trying to replace designers like Uizard.

Instead,

Use Augmented Intelligence to help us be more efficient and productive and deepen our relationship with UI developers through robust, code-based React DSM. Give tedious, repetitive tasks like table creation to a machine, and leave the creative tasks that require human connection and empathy to designers. 

Or don’t. 

You will not be the last design tool to disappear in the last five years. Invision, too, seemed utterly unassailable. Until it wasn’t.

85% of AI projects fail

According to Forbes magazine, 85% of AI projects fail. Figma’s “text prompt to complete App Design” is just the latest sacrifice on the altar of humans learning how to work with our new “robot overlords.” 

To make sure your next AI project does not fail, start with these three things: 

  1. Stop looking to tools like Figma to define what you can and cannot do as a designer. The current state of UX tools reflects the lack of clarity and thought leadership, mass confusion, and abject terror that currently permeates our entire UX community. Remember, tools are not you, and you are not your tools. The sooner you go back to understanding who you are at your core, the better. If you have a tough butt in dire need of kicking, here’s an article I wrote just for you: Open Letter to Junior Designers.

  2. Sign up for this newsletter. We are dedicated to writing about practical techniques for your UX for AI success. Also, Daria and I have a book coming shortly specifically written to help you succeed in the new “AI Normal.” Sign up here to get notified when it’s published and get your discount code. You can sign up here.

  3. Attend our spectacular full-day hands-on UX for AI workshop At UXSRAT on September 9th. We will teach you lean UX for AI techniques you need to succeed: how to pick the right use case, draw an AI product storyboard, create a digital twin and a vision prototype, apply a value matrix to train AI to think in human values and iterate your design rapidly while keeping users front and center while balancing the AI model requirements with the available training data. This workshop WILL sell out like the dozen of our previous workshops in Boulder, Rosenfeld Media NY, Copenhagen, and Lisbon, so get your ticket now: https://strat.events/usa/ 

NOTE: This article is sponsor-free. No one has paid me for this content. All opinions are my own. If you found it useful, please share it, along with your thoughts.

Peace,
Greg

P.S. Our full-day hands-on UX for AI workshop at UXSRAT on September 9th and will sell out like our previous workshops at UXSTRAT 2023, UX Copenhagen, Rosenfeld Media online, and UXLx in Lisbon. To ensure your access to practical techniques you will need to succeed in your next AI-driven project, get your ticket now: https://strat.events/usa/. See you there!

Reply

or to participate.