← All work

Our Ability Inc.2026

Self-Disclosure and Accommodation

People with disabilities often have to decide if, when and how to tell an employer. This is a small web tool where you ask a question and get a short answer, with the sources shown underneath.

Overview

Role
Technical Product Designer
Company
Our Ability Inc.
Timeline
Jan to Jul 2026
Platform
Web
Status
Prototype

Problem

People with disabilities often have to decide if, when and how to tell an employer.

What I did

I led it from the founder's idea to a working prototype. I designed and built it with Claude Code.

Result

A working prototype that answers questions on both tabs. I have no numbers on how many people used it.

The solution

It has two tabs, Self-Disclosure and Accommodations. It searches the web, then an AI writes a short answer from what it found.

No account and no setup. You open the page and start asking. Three steps: pick, wait, read.

Step 1, ask. Choose a topic, tap an example question, or type your own.
Step 2, wait. A short message says it is searching and thinking. The button fades while it works.
Step 3, read. You get a short answer, then the sources it came from, with links. On the Self-Disclosure tab there is also a button that turns the answer into a short video.

Accommodations works the same way. It has its own topic filters, like Physical Accommodations and Technology & Software, and its own example questions. The answer card is called Guidance, and it has no video button.

The problem

People with disabilities often have to decide if, when and how to tell an employer.

John, the founder of Our Ability, had an idea for a tool like this. I led it from the first screen to a working prototype, using Claude Code.

Key decisions

Show where answers come from

Every answer lists the sources it came from, with a link and a short snippet. I did this so people can see where an answer comes from and trust the product.

If the web search finds nothing, the answer says so and falls back to general knowledge of the ADA, the US disability law.

The sources list. Raw HTML tags show in the snippets, which I list under what I would do differently.

Keep answers short and plain

The AI is told to keep answers to about 80 words, in plain text with simple bullets and no bold or markdown. I kept this because short answers are easy to read.

The 80 word limit is a request to the AI, not a rule in the code, so some answers run longer.

An Accommodations answer. Compare it with the Self-Disclosure answer in Step 3 above.

Start people off gently

Pick a topic, or tap an example question. The example fills the box but does not send it, so people can change the words before they ask.

Nobody has to start from a blank box.

Self-Disclosure.
Accommodations. Both tabs use the same pattern: topic filters, a question box and example questions.

Say what it is not

Every screen has a note saying this is general information, not legal advice. People should use the tool as a reference and not take the answers at face value.

Past questions are saved only in the person's own browser, the last 20 on each tab. There is no sign in, so there is no account to save them to, and nothing is stored on a server.

The note at the bottom of every screen, and the empty history list above it.

Edge cases I designed for

  • What if the web search finds nothing?

    The answer says so and falls back to general knowledge of the ADA, the US disability law.

  • What if someone wants to change an example question?

    Tapping an example fills the box. It does not send it, so you can change the words first.

  • What if the answer takes a while?

    A short message says it is searching and thinking. The button fades while it works.

  • What if someone takes an answer as legal advice?

    Every screen has a note saying this is general information, not legal advice.

  • What if there is no account to save questions to?

    Past questions are saved only in the person's own browser, the last 20 on each tab. Nothing is stored on a server.

Accessibility

White text on the orange buttons and selected topic pills measures 2.85:1. Normal text needs at least 4.5:1.

The text fields also lose their outline when you click into them, and only the border color changes.

I worked this out from the colors in the code and have not tested it with a screen reader.

The selected pill and the Ask Question button. White on this orange is 2.85:1.

Results

The tool works and answers questions on both tabs. I have no numbers on how many people used it. John is still shaping what it should become.

How I got there

Where it started

The name and colors are Our Ability's. The idea is still being worked on, so this case study covers the prototype as it is today.

How I worked

I designed and built it with Claude Code. It is a prototype, and the team is still deciding what to do with it.

Also in 2026: Abli.ai, making PDFs readable for screen readers. Earlier work for Our Ability: the job search dashboard redesign, 2022 to 2023.

What I'd do differently

These come from my own review of the screens and code, not from user feedback.

No safety path, and no note that answers come from AI

The tool answers questions about law and work, but it has no crisis or legal-help path. I also could not find any text on screen that says the answers are written by AI. The footer note helps, but I would add both.

Raw tags in the sources

HTML tags like <strong> and codes like &quot; show up in the source snippets. And because the 80 word limit is only a request to the AI, some answers run long.

The video needs work

The video button makes a short AI video from an answer. The text inside it comes out garbled, and there are no captions. Until that is fixed, I would hide the button.

While the video is made.
The result. The text in the video comes out garbled.

Details that add up

A missing setting can show a developer message to users. There is no way to copy or save an answer. Clear All deletes history without asking.