Summary: Live Internet event with Jakob this Thursday | Jakob answered a bunch of questions about the state of UX | Perplexity lets users down with two terrible design choices | Don’t settle: build a side business | 1,000 Instagram posts
UX Roundup for March 24, 2025. (Midjourney)
Live Internet Event with Jakob This Thursday
I am doing a live Internet event with Dovetail this Thursday, March 27 at 11:00 AM Pacific Time USA.
Equivalent times:
USA Eastern Time: 2:00 pm
London: 18:00
Berlin: 19:00
Dubai: 22:00
New Delhi: 23:30
The event is free, but advance registration is required.
I’m speaking at a live event this Thursday. Contrary to this image, all participants, including myself and the moderator, will be virtual. (Midjourney)
My Take on UX Issues of the Day
I did a podcast interview with VUX World (YouTube, 1 hour). Here are some of the points I made:
AI-generated research now surpasses many humans
One of the most provocative realizations for me was recognizing that recent AI developments, particularly in "deep research" tools from OpenAI and Google, now routinely deliver insights equal to or even superior to many human experts. These AI-generated reports surpass typical consultant reports, approaching the caliber of skilled researchers. Although not yet at a Nobel Prize-winning genius level, the quality improvement over the past two years alone is astonishing. This directly challenges the assumption of human supremacy in professional research, making many consultants' roles increasingly questionable and potentially obsolete.
Why notable: Directly threatens human-centered professional prestige, stirring significant controversy among traditional experts.
AI providers have miserably failed at UX design
My greatest frustration with AI is that the providers have almost entirely ignored user-centered design principles, despite decades of established best practices in UX. Companies like OpenAI and Midjourney initially produced interfaces plagued by usability barriers—unintuitive naming conventions, poor error handling, unclear affordances, and frustrating interactions. Such disregard for usability is inexcusable given the availability of well-documented UX heuristics. Only recently have these providers started making marginal improvements, but the core issue remains unaddressed: a consistent neglect of fundamental usability practices. Such failures significantly undermine AI’s real-world effectiveness.
Why notable: Strongly criticizes prominent technology leaders, directly calling out their professional negligence in UX, which may provoke defensive reactions.
Many AI vendors have tossed out the book of UX design principles. (Ideogram)
AI stigma is directly caused by poor UX design, not inherent human distrust
The prevalent stigma surrounding AI, particularly evident in medical diagnostics, is not due to inherent human distrust but rather poor user experience design. Physicians reject AI-generated diagnoses primarily because AI tools fail to communicate their reasoning, confidence levels, and limitations clearly. The blame rests squarely on the AI developers for neglecting UX, not on user biases or sensationalist media alone. I am convinced that with proper UX—clear explanations, intuitive interaction, and transparent confidence indicators—AI acceptance would dramatically improve. AI’s current lack of trustworthiness is a design flaw, not an inevitable limitation of the technology.
Why notable: Directly assigns blame for societal distrust of AI on designers and providers, a provocative stance likely to be controversial among AI developers.
Many users are still not willing to trust what AI is selling, even when independent studies show that AI’s results are rated superior if people don’t know that those results came from an AI. (Ideogram)
The Articulation Barrier is a major UX issue in conversational AI interfaces
A subtle yet significant usability problem emerging with conversational AI is the "articulation barrier." Users, even though communicating in their native language, frequently struggle to clearly articulate their intentions and requirements to AI systems. Contrary to assumptions that natural language eliminates complexity, many people fail to precisely express their true needs. This barrier disproportionately affects the 95% of users who lack advanced abstract thinking skills. Despite theoretical ease, conversational interfaces thus still require substantial UX research to understand and assist user articulation, highlighting an underestimated complexity in designing natural language interactions.
Why notable: Illuminates a critical yet overlooked usability challenge specific to conversational AI interfaces, an insightful point for UX professionals.
The articulation barrier is real: 95% of humans struggle to specify abstract or complex requests in prose language. (Ideogram)
AI usability is greatly enhanced by proactively asking clarifying questions
An important usability breakthrough in recent AI interfaces is the proactive clarification of user requests through targeted questioning. AI tools like OpenAI’s deep research significantly improve usability by prompting users with intelligent clarifying questions before finalizing their outputs. This UX technique dramatically reduces misunderstandings and user frustration while simultaneously addressing slow AI response times. Clarifying questions offer users a sense of control, significantly improving satisfaction, efficiency, and overall user experience. This UX innovation is one of the most promising recent developments, likely setting new standards for conversational AI interfaces.
Why notable: Provides practical guidance for UX professionals on significantly enhancing conversational interface effectiveness through proactive clarification strategies.
It’s best if an AI agent asks any clarifying questions first, before is initiates a project that will require substantial time to complete. (Leonardo)
Classic usability principles remain fully applicable to new AI technologies
Despite dramatic technological evolution from command-line interfaces to graphical UIs to conversational AI, my "10 usability heuristics," established in the early 1990s, remain remarkably relevant and applicable. Fundamental principles such as visibility of system status, user control, and consistency have proven timeless, reflecting persistent human cognitive needs rather than transient technological fads. The challenge today is not creating new heuristics but intelligently applying these existing principles to emerging AI scenarios. UX designers must reinterpret and reapply these foundational insights, rather than assuming new paradigms completely invalidate traditional UX wisdom.
Why notable: Reinforces the long-term value and robustness of foundational UX heuristics, providing reassurance and direction for designers amid rapid technological changes.
AI will soon learn how to design new user interfaces based on 30+ years old usability principles since they remain equally relevant in the new era. (Ideogram)
Personalized AI interfaces will become a critical competitive advantage
A key emerging UX opportunity in AI is deeply personalized interfaces, becoming precisely tailored to each user's preferences, history, and workflow. Personalization creates strong inertia, making users reluctant to switch products once AI deeply understands their unique contexts. Currently, general-purpose AI lacks this advantage, resulting in low switching costs and weak user loyalty. Companies focusing on deep personalization will create lasting competitive differentiation, achieving sustained customer relationships. UX professionals should therefore prioritize personalization strategies, as future AI competition will likely hinge on deeply individualized user experiences rather than merely raw performance.
Why notable: Highlights personalization as a strategic UX imperative for long-term product differentiation and competitive advantage in the AI market.
Users will likely develop a preference and maybe even attachment to a personalized AI that knows and reflects all their past interaction history. (Ideogram)
We are still in the early formative days of AI usability maturity
Despite impressive recent advancements and widespread media attention, we remain firmly within the early, formative stages of AI usability maturity. Comparing today’s AI UX to the internet’s early days, I see similar patterns: enthusiasm, initial excitement, rapid change, and inconsistent usability practices. History teaches patience: true maturity takes years, not months. UX professionals should expect substantial improvements and evolution in best practices throughout this decade. Early frustrations and suboptimal UX design reflect initial immaturity rather than permanent limitations, promising significant professional opportunities as AI usability practices inevitably mature and stabilize.
Why notable: Provides a reassuring historical perspective, setting realistic expectations and strategic patience for UX professionals currently navigating AI’s turbulent early growth.
AI is still a baby and we’re early on the maturity curve for AI usability. It’ll take years before we truly know how to build good AI UX. (Midjourney)
Perplexity Design Fails
I am generally a big fan of Perplexity, but a recent update to its mobile app included two major usability fails, as shown in these screenshots:
On the home screen, a new icon in the upper right corner shows a person with a plus (left screenshot). What might that do? I thought maybe it was related to customizing the “new voice mode” they had announced, so I clicked it. Turns out that this icon is a promotion to “invite your friends to join Perplexity.” Much as I like Perplexity, I will not spam everybody in my contact list with emails from them. This type of social non-permission marketing is already odious, but it’s obnoxious beyond belief to promote it as if it were a useful UI feature with a permanent icon on the home screen.
The right screenshot shows the UI for choosing the voice for Perplexity’s new voice mode. What happens when you tap one of those names? You’d think that the software would play a sample of that sound. No such luck. You have to select a voice based solely on its name, not how it sounds!
It’s an obvious (and classic) usability guideline to provide a preview at the time of selection between preference settings. This is especially true when the choices are not obvious. Big UX fail by Perplexity. (I am pleased to report that the first of these two design flaws has been corrected in the most recent update to the Perplexity app.)
Perplexity failed to follow basic UX design principles in a recent release of its mobile app. (Ideogram)
Don’t Settle: Build a Side Business
ADPList has a very interesting concept for its latest conference, a one-day affair on April 23 with all-virtual talks so you can attend remotely from anywhere in the world. The conference is called “Don’t Settle” and is about how to quickly start your side business (without quitting your job).
In typical ADPList fashion, registration is dirt-cheap at only $69 (USD). I usually applaud this low-price strategy for facilitating professional growth for colleagues in poor countries, but given the event topic, the bargain-basement pricing may also be helpful for people in rich countries since this may not be a conference you can expense and ask your boss to cover.
There are several cases where I particularly recommend that you pursue a side business:
You’re stuck in a legacy design team that hasn’t pivoted to be AI-first, meaning you have zero future career prospects. Do anything you can to take on a few outside AI projects to gain the experience needed to land another job.
You’re a solo UXer, working alone. Gain some added perspective by working with other people for a change. This is a scenario where you can see substantial growth even from fairly low-level contracting jobs, as long as a skilled senior professional supervises you.
You’re a graduate student who doesn’t want to become a professor: you need to build a resume that hiring managers will relate to. Or you’re a professor who wants to include some relevant industry perspective when teaching your students.
You live in a poor country where local salaries are low and local UX maturity is equivalently low. Your side business should target international clients, which will make you good money for local conditions but, more importantly, expose you to higher-maturity design practices.
There are many reasons you might consider starting a side business. (Napkin)
1,000 Instagram Posts
I recently passed 1,000 posts to my Instagram channel. Most of these posts are images from this newsletter, but in the last few months, I’ve also started posting some videos. (My YouTube channel is the main home for my videos, though.)
It’s remarkable that I have created a thousand images in less than two years — particularly since I used to be known as the “no-graphics usability guy.” My advice to minimize graphics dates back to the era of dial-up modems, when response times suffered unbearably from any but the smallest of images. But as users’ bandwidths have increased, I abandoned this old usability guideline. As you can see, now that we have AI images, I’m in favor of illustrating stories with pictures that are either entertaining or supplement the text.
Celebrating 1,000 Instagram posts: mostly images, but some videos. (Leonardo)