Teaching AI to build more accessible software with skills

25 November 2026 – 1:30PM

AI can turn ideas into working software faster than ever. It can also create inaccessible software just as fast.

An accessibility skill can help with that. It can package accessibility principles, acceptance criteria, verification tooling and more into instructions the AI follows, so accessibility is built in by default.

This is a practical, hands-on session to help you build more accessible software with AI. You’ll build a prototype with no skill applied, check it for accessibility problems, discuss the problems as a group, then rebuild the same prompt with an accessibility skill applied and see what changes.

Same tool, same prompt, different (more accessible) output.

Along the way, we’ll cover what a skill is and what makes a good one. We’ll show where skills deliver measurable improvements, and where AI still falls short and human review is essential. We’ll also show how to use skills in your own workflow, and how they bring wider benefits, like more accessible prototypes you can test with users earlier.

You’ll leave able to improve AI-generated websites before inaccessible patterns become product decisions, and with skills you can put to use straight away.

You’ll leave knowing:

1. Where AI goes wrong. The common accessibility issues AI tools introduce or miss, and why.
2. What a good skill looks like. What a skill is, what makes a good one, where it improves output and where human review is still needed.
3. How to apply this yourself. Practical ways to apply accessibility skills earlier.