AI-Assisted Accessible Development: Building Workflows That Require Your Judgment
taught by: Kelsey Ruger
Session Summary
The thing experienced accessibility practitioners are quietly worried about isn't that AI will replace their judgment. It's that AI will make their judgment invisible. When the tools generate confident-looking, WCAG-adjacent markup that still fails real users, the gap between knowing and not knowing becomes much harder to see. This workshop builds both the domain knowledge and the AI workflow together, so the expertise is encoded in how you work, not just what you know.
Description
You've probably already seen this play out. Someone on the team runs a prompt, gets back what looks like accessible markup, and ships it — because the ARIA attributes are there, the labels are present, and the automated checker passed. What's missing is the kind of judgment that comes from actually understanding how people use assistive technology: whether the focus order makes sense in context, whether this interaction pattern is going to confuse a screen reader user, whether "technically correct" is the same thing as actually usable. It usually isn't, and the tools will not tell you that.
This is the core challenge with AI-assisted accessibility work. These tools amplify what you bring to them. If your accessibility knowledge is deep, they make you faster and more consistent. If it isn't, they help you produce inaccessible code with more confidence and less friction — which is arguably worse than the problem we started with.
This workshop works with current AI coding tools hands-on, but what you're building is bigger than any specific tool: workflows that encode your judgment rather than just hoping it shows up in the output. You'll structure projects so AI has real accessibility context to work with, build reusable commands that carry your expertise into every component you generate, and develop review processes that catch what automated testing misses. Then you'll stress-test all of it against deliberately flawed output to see where the workflow holds.
The tools will keep changing. The framework you leave with won't.
Practical Skills
- Structure AI-assisted projects so that accessibility standards, patterns, and constraints are available to the AI as working context.
- Build reusable commands and workflows that encode accessibility requirements into repeatable processes
- Evaluate AI-generated output for accessibility gaps that automated checks miss.