[TMA] Thoughts on AI/UX: Quality, Bespokeness, and Adaptation

First: Quality

Last week, I discussed how, when working with AI, discernment is emerging as an even more important aspect of design work. I'm reminded of what Jesse said in his "The Elements of UX in the Age of AI" presentation (watch for only $14.99), where he shared that because LLMs are trained on averages of vast human output, they are inherently biased toward the median, which means:

“What it will always be best at is mediocrity.”

Without human direction, the output tends toward just okay—not great. And this confirms a conversation that I find myself having with folks about AI, design, and product development: while teams can produce more stuff quicker, is what they're producing better? And, generally, the response is, "No."

And I believe this is because UX/Design organizations aren't doing the necessary work to define standards for design and experience quality, believing that quality should just be felt, in a kind of "I know it when I see it" way. That may work in high design maturity environments, but in low-maturity environments, your concerns will be dismissed, in favor of other functions' crisp definitions of quality: product management's focus on adoption, usage, Net Promoter Score, and engineering's standards on code maintainability, functional quality (doesn't crash, performs quickly and as expected.

How might a UX/Design org define quality? Over 4 years ago, I built a framework for developing quality standards. With every passing year, I've found it to be only more important, because how often I see that not having clear standards inhibits UX/Design orgs' ability to advocate for their agenda.

So, if you're wondering about the role of UX/Design in a world of AI tools, consider doing the work to define those standards in your organization.

Second: Bespokeness

An exciting opportunity generative AI enables is, by lowering the barrier to building software, we may see more, 'smaller,' software tailored to specific audiences. I nearly called it "artisanal," but I don't know if you can claim that anything produced by AI is the work of an artisan, so I'm settling for 'bespoke." Readers of this newsletter probably know more and better examples than I do. One I'm aware of is ChatPRD, a purpose-built tool for crafting better product management documents.

In talking about this with an old friend, it reminded me of the late '80s-into-'90s era of shareware, with lone or very small developers serving niches. And of Mark Bernstein's 2007 talk on NeoVictorian Computing (he's whom I first heard the phrase "artisanal software."

Third: UIs that Adapt To Use

Something that comes up in the discussion of AI and UI is how, with AI, software UIs can evolve to suit the specific needs of the user. This is something that's been discussed for at least 25 years (I gave a talk at IA2000 on "How Websites Learn: Information Architecture That Adapts to Use", and you can see the hinkiest PPT ever of it).

We've had algorithmically-driven content delivery for decades—any Amazon product page, your social media feeds. So, when it comes to the content within the interface, this is old hat.

For decades, we've had the possibility of truly adaptable UIs, where not just the content, but the specific interface elements, are personalized (say, to meet you at your level of expertise, so power users get a different interface than neophytes). But this has never happened, and I suspect it never will. Users crave a kind of conceptual and cognitive stability in their tools. While an adaptive UI may be considered helpful by its creators, it will be met with frustration by its users.

How about some more... me?

If you feel like you don't have quite enough of me in your life, NN/g published my conversation with Therese Fessenden, about "what it really takes to build and scale effective design teams. We explore the systems, structures, and leadership mindsets that shape successful organizations, the relationships between product and design teams, and how design leaders can foster clarity, consistency, and impact as their teams grow."

Witness me in my orange shirt splendor on YouTube:

or, if you prefer audio-only experiences: