With agentic AI contributing to the AI hype-cycle, I've been giving it some thought from the point of view of a disabled person, and also as an accessibility specialist who's already witnessing the arrival of the agentic web first-hand:
tetralogical.com/blog/2025/08/…

#AI #agentics #accessibility #a11y

in reply to Sean Randall

@cachondo I've always had major concerns about equivalency/equity here. If *everyone* is using an agentic interface, I think it's okay: we all have a more or less equivalent experience. But while most people are using a curated, intentionally designed interface and people with disabilities are relegated to some AI because that's just easier than doing accessibility properly, we have a situation where the disability experience might be slower, less efficient, less accurate, less trustworthy, whatever. And if anything goes horribly wrong - you transfer money to the wrong account because the AI messed up - who is responsible?
in reply to Timothy Wynn

@cachondo @twynn I genuinely think things like shopping or searching for flights might be easier with agentic AI, but I think that's likely true for many people, with disabilities or without. It disproportionately helps people with disabilities because of the poor state of shopping accessibility. But while this isn't the approach used by everyone, I worry that promoting this as the *right* approach for accessibility creates inequity.
in reply to Jamie Teh

@jcsteh Right, and I could see it being useful for, say, buying a house or whatever, where you can only filter listings by price and location, but other more ambiguous datapoints like whether it is near a multi-lane area, close to amenities, etc, would be harder to filter for, but how can we trust its output and would it then make it problematic? Algorithms are great until you can't figure out how it determined that answer and whether or not it's an outlier. @cachondo @tink
in reply to Jamie Teh

@jcsteh @cachondo @twynn My hunch is that it'll be more like the arrival of touch screen interfaces. They're simpler and easier for a lot of people to use, irrespective of disability/accessibility, and that's more than likely why apps remain so popular - more popular than using the equivalent websites in many cases. That the simpler interface lends itself towards better accessibility is a bi-product not the stated intent.
in reply to Léonie Watson

Hey! I was reading your article, and one of your lines stood out to me.

we generally prefer to have a chat because (for most of us) it requires no effort at all - and what could be more convenient than that!


I would like to emphasise the difficulty that "having a chat can be" is for autistic people like myself. It requires a delicate understanding of tone, nuance and power dynamics which are often invisible to non-autistic people. Even writing this message requires a lot of drafting and re-drafting, and I might still get the tone wrong. By contrast, I find working with a structured interface - whether GUI or CLI - much easier. I also use contextual clues to help me navigate the web which disappear as soon as a chat interface appears.

I am writing to raise awareness of the differing needs that come with different disabilities - what might be necessary for one group of people may be overwhelming for another. Even for autistic people, some happily work with AI, while I find it very difficult to use

This entry was edited (3 weeks ago)
in reply to Léonie Watson

@sitcom_nemesis Even from the perspective of someone that finds chatting relatively easy, I'm curious: would you want to chat constantly? If everything was agentic and there were no structured interfaces any more, every single thing you do would have to be a chat. Aside from the impracticality of doing that in some situations where quiet or privacy is a concern, I think that would become extremely tiring and inefficient if you had to do it all day long for every technological interaction. I think there are some use cases where it might work well, but others where it doesn't, even in the same session.
@brib