Summary: Generative UI benefits from design systems | Autoformat credit card numbers as chunks | Systems thinking beyond features | The top companies in AI are mostly located in the same area as the top dot-com companies | Dovetail AI scores 2-1 in review for user research | Free keynote by Jakob Nielsen | Free Llama 3 may receive more use than its quality warrants
UX Roundup for April 22, 2024. (Midjourney)
Generative UI Feeding on Design Systems
Brian Armstrong from Coinbase posted, “We've been ramping up usage of AI tools on our design team” and that the designs comply with the company’s design system because their Generative UI tool was trained on Coinbase’s existing library of UI designs. (Watch the 2 min. demo video in this post.)
Coinbase is the largest cryptocurrency exchange in the United States, and Brian Armstrong is its CEO. I find it very interesting that the CEO of a fintech company with an AUM (assets under management) estimated at $80 Billion writes about GenUI and makes correct use of very nerdy UX concepts such as “design systems.” What other CEO even knows whether his or her company has a design system? In any case, at Coinbase, UX and GenUI seem to have that much-vaunted “presence in the C-suite.”
The second interesting conclusion from the brief demo of Coinbase’s GenUI is that it seems very promising to integrate GenUI and Design Systems: the two seem made for each other. Also, GenUI will generate better UX when trained on a large set of high-usability UI that has already been designed for the specific vertical of the company in question.
Of course, I also have hopes for general GenUI, à la UIzard (one of the top companies in my recent survey of AI tools used by UX professionals). However, general GenUI tools will probably need more detailed prompts with more context than what’s required for a finetuned GenAI.
One of the points made in the Coinbase demo, is that their use of GenUI automatically enhances the accessibility of the resulting designs. (This, of course, will only work if the AI’s training set was already designed with strong usability for disabled users.) This case study shows the promise of my vision of replacing the current poorly functioning approach to accessibility with full individualization driven by future versions of GenUI.
Mixing a Generative UI AI tool (GenUI) with a design system may be the recipe for increased productivity in the UI design stage of the UX process, resulting in increased screen quality and improved accessibility. (Ideogram)
Autoformat Credit Card Numbers into Chunks
Baymard Institute (the world leaders in ecommerce usability) reminds us in a short carousel to autoformat credit card numbers by inserting spaces that break up the long string of 16 numerals into chunks.
1234567890123456 vs. 1234 5678 9012 3456
Which of these two formats lowers data-entry errors the most? Obviously, the second. Follow this guideline, and you’ll be a step closer to complying with usability heuristic number 5, error prevention.
Credit cards have terrible usability in so many ways. At least help users reduce errors when they enter their number into a form on your website. (Midjourney)
Systems Thinking Beyond Features
I’ve recently seen people recommend Micro1.ai as a place to hire developers overseas. (I have no personal experience with the service, so I’ll stick to commenting on their business model, which I find promising.)
As you notice, the website domain ends in “ai,” which at first made me wonder, what AI is there in hiring offshore engineers. That’s an old-school HR problem. Silly me. All “old-school” problems are subject to AI disruption.
Micro1’s selling point is that they prescreen for 1% of the world’s engineers, using a combination of AI screening and human interviews. So far, so good. But the systems thinking comes into place through the other services they integrate: they also handle the profusion of confusing local employment laws and run global payroll for companies that take on overseas talent from their service. Finally, they offer continued AI upskilling for the developers placed through their service. (The latter is especially important when hiring developers: current AI slightly more than doubles the productivity of software developers, but the estimate is that this may increase substantially until we might reach a 5x productivity for AI-enabled programmers in 5-10 years. Staying on top of these changes could be hard for the average developer-employing company without a specialized AI team.)
The systems-wide solution approach follows the traditional saying that customers don’t want a drill; they want a hole in the wall. Similarly, you don’t really want to identify a person to hire in some obscure country. You want to work with that talent despite the stupid local laws, and you want him or her to remain a top talent. Solving a more complete picture is an example of the systems-level thinking we need for all areas of design. Giving me a list of top-rated developers is a feature. Offering payroll in hundreds of countries is a feature. Offering continued education in AI is a feature. Combining the lot into “talent-as-a-service” is systems-level solutions design.
Unfortunately for my readers, Micro1 is currently limited to software developers — they don’t offer UX talent. The market for recruiting software developers is about 10x the market for recruiting UX staff: there are 3 M UX pros in the world and 30 M software developers. My recommendation for average development projects has long been for 10% of the staff to be UX and 90% to be developers. These two estimates align fairly well with each other, given that many development projects still lack UX.
I can’t blame Micro1 for going after a 10x market first. But I hope that they (or similar companies) soon expand to also cover UX recruiting. There’s so much high-talent / low-cost UX staff accumulating all over the world, and we need an easier way to bring this talent onto our teams.
The world is your oyster with new global talent-matching services like Micro1. (Midjourney)
Geographic Location of Top Companies Shows Strong Bias
The following map shows the geographic location of the headquarters for the 22 AI services that were used the most in my recent survey of the AI tools used the most by UX professionals.
Headquarters location for the top 22 AI tools used by UX professionals in 2024.
Compare with this map of the location of the top 22 websites in 1998, at the beginning of the dot-com bubble:
Headquarters location for the top 22 websites in 1998.
The similarity is big: both maps are dominated by the west coast of the United States. Even so, there are some differences, as seen in this comparison table:
Area | 1998 dot-com | 2024 AI | Change |
San Francisco Bay Area | 41% | 68% | +27% |
Other USA West Coast | 27% | 9% | -18% |
Rest of USA | 27% | 0% | -27% |
Europe & Israel | 5% | 9% | +4% |
Australia | 0% | 14% | +14% |
The non-west-coast parts of the USA have been eliminated as headquarters for leading companies in the current tech revolution. At the same time, the rest of the world (outside the USA) now has 5 times as many leading companies as before. Still, having 18% of the top AI tools based outside the US is a tiny percentage.
Overall, the geographical bias has increased, with 2/3 of leading AI companies found in the tiny San Francisco Bay Area. It’s hard to imagine that the Rest of World will accept this state of affairs in the long run, with their intellectual future held hostage by a region taking up 0.003% of the world’s land area (5,322 square kilometers for the 3 Silicon Valley counties vs. 149 M km2 for the world).
This geographical centralization may change over time. You will notice that for AI, I am plotting headquarters as of Year 2. (Counting the release of GPT 4 in 2023 as Year 1 of the AI revolution.) In contrast, for the Web, I plotted headquarters as of Year 6. (Counting the release of the first GUI Web browser, Mosaic, in 1993 as Year 1.)
AI is growing much faster than the Web did, but even so, it’s still so early that we should expect major changes in the next few years. One of those changes is likely to be some degree of geographical diversification.
Dovetail’s AI Features for Analyzing User Interviews
The UX Design Institute has posted a 7-minute video where Rachael Joyce (Head of Research and Insights) demos and reviews Dovetail’s features for analyzing the recordings of user interview recordings. She scores two features as being useful, one feature as not being good enough yet, and one feature gets a mixed review.
Overall, this is a positive review, but Dovetail clearly needs more work to be considered a great AI tool for user research.
After analyzing Dovetail pecking at 4 ways of using AI in the user research process, two AI features get good scores, one gets a bad score, and one scores neutral. (Ideogram)
Live Keynote by Jakob Nielsen: 10 Foundational UX Insights
I will present a live keynote on ADPList on May 15, 2024 on the topic of “10 Foundational UX Insights” in support of their mentorship mission. I thank my good friend (and stellar UX researcher) Alita Joyce from Google for being the session chair and managing the Q&A.
The event is free, but advance registration is required.
My speech is live on the Internet at these times:
San Francisco: 10 AM USA Pacific Time
New York: 1 PM USA Eastern Time
São Paulo: 2 PM Brazil Time
London: 6 PM British Summer Time
Paris/Berlin: 7 PM Central European Time
Dubai: 9 PM Gulf Standard Time
New Delhi: 10:30 PM India Standard Time
Singapore/Beijing: 1 AM the next day (May 16) Singapore Time/China Time
Here’s the summary of the talk:
UX has come a long way since its early beginnings at Bell Labs in the 1940s. As we enter Year 2 of the AI revolution, it’s clear that the core principles of UX design are not being replaced but are evolving with AI integration. This journey through the 10 foundational insights of UX reminds us that while “there's nothing new under the sun,” the way we apply these insights is constantly changing. By embracing AI as a tool for enhancing these time-tested principles, we can create more personal user experiences than ever before. Let’s honor the past as we design for the future, ensuring that the essence of UX — empowering humans while making technology subservient — remains at the heart of everything we do.
Jakob Nielsen presents a live keynote on ADPList on May 15. (Midjourney)
Llama 3: Being Free May Drive More Use than Quality Warrants
Meta has released an upgraded version of its generative AI product, Llama 3. Tests show that it’s almost as good as GPT-4. Usually, I would say that “almost as good” should equate to a failed product. Who wants to use something second-best?
(In contrast, Claude 3 Opus is about as good as ChatGPT, and may be slightly better at language tasks such as rewrites. I use both, especially now that Claude seems to have scaled back on its offensive censorship.)
However, Rowan Cheung points out in a post that Meta is making Llama 3 available for free. This is in contrast to GPT-4, which is only available to OpenAI “Pro” users who pay $20 per month. This fee is laughably small, and any company should gladly pay for ChatGPT Pro subscriptions for all knowledge workers since the productivity gains will be about $1K per employee per month. However, evidence seems to suggest that most people who have experience with GPT have been using the substandard free version, ChatGPT 3.5. (This is also why many people are not that impressed with AI: using a miserable product from 2022 will do that to you!)
“Free” is a powerful proposition, despite the logic of immense ROI from paying up. Being free may rapidly gain Llama 3 hundreds of millions of users. This may be Mark Zuckerberg’s plan to regain relevance in the tech industry, after his disastrous foray into the “metaverse” (which I predicted would fail almost a year ago).
Even though ChatGPT 4 is probably still better than Llama 3 (especially when counting added features like the Dall-E image generation tool), the difference between the two may not be enough to persuade many people to pay. OpenAI clearly needs to launch GPT 5 soon.
Even though GPT-4 in itself may be worth $1K/month for a knowledge worker, the upgrade from Llama 3 to ChatGPT 4 may only bring incremental gains corresponding to about $100/month. This is still a 5x ROI on the $20 subscription. But clearly, many people don’t think in such accounting terms and will be persuaded by the siren call of “FREE!”
Llama 3 faces off with ChatGPT 4. Will people pay the extra $20 for better AI, when the free product is almost as good and has a cuddlier name? (Midjourney)