Spreadsheets sound sensible until someone has to maintain them. This is especially true for accessibility where people already see it as extra work.
At first, everyone agrees with the idea that one score is too blunt. Lets add severity and probably we need user impact. What about risk? Add it! And we definitely need to know whether the issue blocks checkout, account recovery, search, publishing or some other flow that people depend on.
The spreadsheet is born.
Then someone has to copy findings from automated tools. Someone else has to add notes from manual testing. Yet another someone has to connect support tickets, QA bugs, design review comments and old component debt. And who gets to decide whether 20 similar failures are 20 separate problems or one broken pattern?
That's how it all starts to fall apart.
And it's not because teams hate accessibility, but because the work of keeping the spreadsheet useful becomes another job on top of the job.
This is where AI can help.
Don't get me wrong! It's not by replacing accessibility judgment or by declaring the product is accessible or not. Please no!
AI can help by organising that mess.
It can take findings from different places and group them by flow. Signup. Checkout. Search. Account recovery. Settings. It can notice that five pages have the same unlabeled button because one component is broken. It can turn technical vague issues like "input missing accessible name" into something everyone can understand, like maybe "screen reader users may not know what this field asks for." It can help separate cleanup from blockers.
AI can help sort everything into better questions:
- Can someone finish the task?
- Who is affected if this fails?
- Is this one bug or a product-wide pattern?
- Does this need manual testing?
All this doesn't make AI the authority.
Good AI accessibility work should show uncertainty. It should say what looks like a risk, but it'd be better to test. This alt text may describe the image, but a human should confirm the purpose. This issue shows across several pages and may come from the same component.
This is the kind of output that helps us spend time where we as humans matter most.
I don't want AI to tell me a product is accessible.
I want AI to help mee see the risks I'm currently missing, group the work in a way I can act on and keep the spreadsheet from taking over my life.
I don't need a smarter score. I need to be able to have better, more meaningful conversations.