Your UI is
not my UI

AI & accessibility today

Chatbots and agents are a revolution for accessibility

User
research

Interviews

We interviewed six people with disabilities about their setups and their AI use.

Illustration of a video call between Matt and Simon, laid over a calendar with an interview booked on each of six days, Rose's AI accessibility interview highlighted

Melody

Uses AI as an assistant, not a replacement.

“I do use AI, but only as an assistant for what I'm looking for. Not to do the searching from scratch.”

Illustration of a woman in dark glasses at a desk, holding a letter in one hand and a phone in the other, its camera aimed at the page

Matthew

Uses AI as training wheels, not autopilot.

“[…] you have to put your name to things. If it went ahead, did a bunch of tasks, emailed people, and I couldn't really take it back — that feels … uncomfortable, giving it that much autonomy.”

Illustration of a man in headphones typing on a laptop beside a stack of books

Sinead

With context about you, AI can help you in ways you didn't anticipate.

“I was really struggling one day … but then, interestingly, it came up with a really insightful reason: it was the drop in barometric pressure. The change had been so fast. And it said that's really common with people with brain injuries.”

Illustration of a woman at a laptop with her hand to her head, a storm cloud overhead and numbered overlays stacked across her screen

[Matt puts notes from prague here as they were better]

Sinead had a brain injury five years ago — came to disability later in life. Screen time is hard; reading a lot in one day is hard.

Her setup is a stack of fighting extensions: a plugin that bolds half of each word breaks the one that overlays numbers for voice control; the number-grid alternative is invisible under her dark-mode plugin. It doesn't work in Safari, and it collapses the moment she also wants to watch a lecture.

She has strong preferences (she likes Gwyneth Paltrow's voice in her screen reader) — but can't get these tools to work together.

AI providing information in an interface that is highly compatible and tailored for the user — not just adapting a layout, but making sense of what's happening to them.

foreshadowing: We found her quote interesting, as the AI helped her in unanticipated ways by being empathetic to her condition and environment.

No revolution

Chatbots & agents: helpful tools
The simple interface is very accessible
Trust & reliability are the problem
We want assistants, not replacements

Allana

AI should remove the disability tax on everyday tasks.

“Part of the enjoyable thing in life is struggling through things. Having everything taken care of for you takes away a bit of the fun.”

Illustration from behind of a woman on a video call, holding an enormous laptop, a coffee beside her

Rose

Everyone's settings are different and incompatible.

“I live in a disabled house […] all our settings are different, even across our iPads. It's actually an in-house joke: I can't look at my husband's iPad, he can't look at mine, and the kids hate using either.”

Illustration of a woman in dark glasses knitting on a sofa, curtains drawn, with laptops and a stack of tablets beside her

Myron

An interface where you're already a power user

“If I had an AI box, I could tell it to line up all my headers from cell A1 to A20 and just send it. That would save so much time. That would be a life changer for me.”

Illustration of a man in a wheelchair wearing a headset microphone, working on a spreadsheet at his desk

Myron is a college professor in Louisiana who has written a book. No perception problems — it's purely about movement, from a spine injury 23 years ago.

He's a virtuoso of voice control — steers his wheelchair by head movement, and the same controls drive his mouse pointer. In the screenshot he's reading out grid numbers to click precisely.

[somehow underline more how impressive it was to view him operate his setup]

Takeaway: he's an expert user of his setup — but it took him 23 years to get there. Can AI shorten that path for others?

Segway to next chapter: All our interviees had adapted their setups according to their personal needs. And all of them ran into issues. So now we want to talk about …

The limits of
adaptability

Today's model

One interface for the average, with tools and settings on top

…and they start to conflict [let's make this present already in the beginning]

Sinead's setup: her browser showing a lecture page, with four extensions laid on it, each a separate layer. A dark mode plugin darkens the page, bionic reading bolds the first half of every word, voice control's number badges land off their targets, and its number grid is lost dark on dark. The layers start to conflict.

Today's model: the vendor ships one interface, built for the statistical average.

Users have to either live with it or stack adaptation layers on top. We saw in the interviews assistive technologies (screen reader, high contrast, magnification) — but also the configuration everyone does for themselves: dark mode, themes, shortcuts, notification settings, extensions.

Those adaptations require that the app was built in a compatible way with standards in mind.

That approach brought us quite far. It's amazing to see in the interviews how low the barriers can get. But we also noticed its limitations.

[Next] We have stacked adapations, none of them knowing about the others, all of them conflicting with each. That's Sinead's setup in one picture.

One interface can't fit everyone

Setups are very personal
Adaptations conflict with each other
Configuration costs energy
Standards have limits

A new concept

Personalization over adaptability

A UI generated for your needs.

A UI should be the result of two things:

what you're trying to do,
and who you are.

[Here start clicking through the rotor] An app for the exact same task needs to look different depending on the condition, state, environment and preferences of the user. In one word, the interface needs to be empathetic to the person in front of it.

The empathetic UI meets you where you are, so the adaptation work disappears.

Where this goes: launching an app is creating an app, just for you, in that moment. Generated on the fly, instead of adapted with layered tools.

There is less linking your UI to my UI. Or even my UI versus my UI tomorrow.

Your UI is not my UI

One task, many UIs

Three stacks side by side, each starting from the same intent, I'd like to translate this text. Rose's: condition, hyperphotosensitivity; environment, on the couch, curtains drawn; situation, a medium energy budget today; and below it an iPad with a dark, dim translator. Myron's: a spine injury, steering by head; at his desk, headset on; focused on a spreadsheet; and below it a monitor with a translator that only listens. A third, unnamed person's, drawn fainter: needs reading glasses; in a crowded subway, one hand free; running late, heatwave, stressful; and below it a phone with big buttons in reach of the thumb. Past it, another empty stack runs off the edge.

In our previous model we started with the interface and the person had to adapt. In the new model, we start with the person.

[Next] Remember Rose from our interviews. She wants to translate this text.

[Next, three times] Then what we know about her, one box at a time:

her condition: hyperphotosensitivity. Light drains her energy and can make her sick, so nothing on screen may be bright;
her environment: on the couch, curtains drawn, so the room is dark and any glare stands out;
her situation: a medium energy budget today, so the UI should ask as little of her as it can.

[Next] Only then the UI. It's the result of the intent and the person: a dark, low-glare translator on her iPad.

[Simon takes over] [Next] Same intent, another person: Myron.

[Next, three times]

His condition: a spine injury, 23 years ago. He can't use his hands; he steers his wheelchair and his mouse pointer with his head;
his environment: at his desk, headset on, so his voice is always within reach;
his situation: focused on a spreadsheet, so nothing should pull his head, and his pointer, away from it.

[Next] His UI has nothing to point at: it listens. Same task, a different UI, and Rose's is still there to compare.

[Next] And it doesn't stop at two. Someone who needs reading glasses, nothing you'd call a disability, in a crowded subway with one hand free, running late in a heatwave. Or Rose herself, tomorrow, in a different situation.

[Next] Their UI: big, clear buttons where the thumb reaches, readable without the glasses they didn't bring.

[Next] And the list goes on: bespoke UIs, generated for everyone, in every situation.

The avatars jump to each person's part.

Empathy is
our point

Intent What you're trying to do

Empathy Who you are right now

The industry is already doing this

[here screenshots of those examples might fit]

Google Search AI Mode generates a custom layout for every query
Artifacts — AI tools generating working interfaces on demand

UIs with empathy

Remove the adaptation burden from the user
Don't rely on apps built with accessibility in mind
Remove compatibility friction between adaptations

Launching
is creating

Apps can't predict you

“No developer can predict all the ways a user will want to manipulate information.”
David Karger

Not just how you use them, but who you are.

The economics changed

Google Research: LLMs already turned a finite library of texts into an infinite one — generated on the spot for any need.

Bespoke apps were never economically viable. Therefore the app had to make a choice for you and take a stance.

Google Maps lets you avoid tolls, highways, and ferries, but not winding roads, if they make you carsick.

AI changes this. apps becomes these APIs. The UI is a wave of a magic wand.

AI changes the cost side. The functionality already exists as APIs; what AI makes cheap is the interface on top. For the first time, the made-for-one interface is affordable. Google Research frames the shift well: LLMs already turned a finite library of texts into an infinite one — generated on the spot for any need. Interfaces are next.

As it happens, when we talked with Rose again, to get an update to se how things progressed since our first interview. And she created an app to organizise or stream library according to her preferences. Not only how she wanted to organize games, but bringing to this interface her disability. (Only possible because economics have changed)

What if launching an app means creating an app?

We started with disability.

But everyone benefits from bespoke software.

Imagine everything is bespoke

Privacy
Stability
Isolating

Closing

Key takeaways

One interface can't fit everyone and the burden lands on the user
A generated UI should serve both the task and the human
With AI, bespoke interfaces become viable for the first time

Back to Sinead

No fighting plugins. No setup that collapses when she opens a lecture.

Just an interface that already fits her.

[we could add a positive illustration referencing the earlier slide]

Thank you!

Q&A

“Doesn't this just hand more power to whoever owns the AI?”

A generated UI can be empathetic to you — or to whoever owns it.

“If everyone's UI is different, don't we lose common ground?”

Real cost. Keep the periphery visible — personal shouldn't mean sealed.

“Will people actually use this, or just accept the default?”

The empathy framing sidesteps it — the default already fits.

“An interface that knows me that well — what about my privacy?”

A real risk. Local models may keep that context on your own device.

“But a UI that keeps changing breaks screen readers.”

Right: stability is a basic trait of accessibility and of good design in general. Our answer is memory, and least surprise is actually empathy applied. So hopefully this corrects itself with empathetic interfaces.

The sharpest objection: if interfaces are generated, don't you destroy what a screen-reader user relies on — the button being where it was yesterday? Stability is a basic trait of accessibility, and of good design in general. A constantly re-rolling UI would be a disaster.

Our answer is memory. The interface is generated once, for you, then remembered and reused. And an empathetic interface applies exactly this: the least surprise. It knows you navigate by memory, so it holds still — personalize the content, not the structure. So we hope the problem corrects itself: an interface that understands you wouldn't move your controls.

Today's one-size-fits-all UI is the volatile one: it changes under you whenever a vendor redesigns it — that's the Ravelry change that put Rose in bed for three days.

“Does this actually work today?”

Not quite. Models can generate working interfaces; the plumbing around them still fails.

Be honest about where we are. Eight months after our interviews, Rose tried to build a bespoke Steam library manager for herself. It didn't really work out: lack of support, lack of knowledge, still too time-consuming, too much tinkering. Rose is technical and motivated — if anyone could build her own tool, it's her. We replaced configuration toil with developer toil: the AI generated a front end, and what defeated her was the glue — storage, accounts, the plumbing you don't know exists unless you write software.

Is that failure telling us the idea is wrong, or that we're just early? We think it's the moment — and we have evidence, not just hope. Google Research found that generating a working interface is an emergent capability: their newest model produced valid UIs with zero errors, while a model one generation older failed on 60% of prompts. It crossed a threshold rather than improving gradually.

So our honest position: the plumbing that beat Rose is a property of today's tooling, not of the idea. We could be wrong — but we think it's a matter of time, and not much of it.

Your UI is not my UI

AI & accessibility today

Chatbots and agents are a revolution for accessibility

Interviews

Melody

Matthew

Sinead

No revolution

Allana

Rose

Myron

Today's model

One interface can't fit everyone

Personalization over adaptability

A UI generated for your needs.current task.preferences.accessibility needs.circumstances.state of mind.external environment.needs.

Your UI is not my UI

One task, many UIs

UI = f(Intent, Empathy)

The industry is already doing this

UIs with empathy

Apps can't predict you

The economics changed

What if launching an app means creating an app?

Imagine everything is bespoke

Key takeaways

Back to Sinead

Thank you!

Your UI is
not my UI

A UI generated for your needs.