Ability Bazaar

Inspiration

Digital accessibility remains a major challenge. Many websites and applications create barriers for people with disabilities, particularly people who are blind or have low vision. At the same time, the lived experience of people with disabilities is an important resource for building more inclusive technology.

We wanted to explore how AI could help connect accessibility needs with the people who understand those needs firsthand.

This inspired Ability Bazaar, a prototype for an accessibility-focused marketplace where organizations can request accessibility services and people with disabilities can participate as creators, testers, and reviewers.

Our goal is to explore a model in which accessibility is not just a technical requirement but a meaningful part of building inclusive digital products.

What it does

Ability Bazaar is a prototype designed to bring accessibility services and AI-assisted workflows together in one platform.

It focuses on four offerings:

Accessibility Audits: Help identify potential accessibility barriers through automated scanning and expert review.
Accessible Websites: Support the development of websites designed with accessibility in mind.
AI App Builder: Explore how natural-language ideas can be transformed into drafts of accessible mini-applications, subject to human review.
Training Sponsorships: Provide a concept for supporting training related to accessibility testing and accessible product development.

The platform is designed around a human-centered principle: AI can assist with repetitive tasks and initial drafts, but human judgment is essential when evaluating real-world accessibility.

These are the intended capabilities of the prototype, not a claim that every feature is production-ready or that employment outcomes have already been achieved.

How we built it

We designed Ability Bazaar around a combination of AI assistance, conventional application logic, accessibility tools, and human oversight.

Our technical approach includes:

Python and Flask for the application backend.
Google Cloud Run as the hosting environment.
Cloud Firestore for application data, including workflow-related records.
Google Gemini for AI-assisted classification, scoping, test-plan preparation, and application drafting.
axe-core and browser automation for collecting automated accessibility evidence.
PayPal Orders API for payment verification within the intended workflow.

The proposed workflow starts with a customer or creator describing a need. AI assists with organizing the request and preparing an initial scope. Accessibility tools help gather evidence, and a human reviewer evaluates the results before customer-facing work is approved.

We also designed a separation between AI-generated suggestions and authoritative operations. Pricing, payment verification, and workflow state should be handled by application code rather than delegated to an AI model.

As a prototype, Ability Bazaar is an exploration of how these components can work together reliably and responsibly.

Challenges we ran into

One of our central challenges was deciding what AI should and should not be allowed to do.

AI can generate drafts and help organize information, but its output may be incomplete or incorrect. We therefore designed the workflow around human review rather than treating AI-generated results as automatically trustworthy.

Another challenge was understanding the limitations of automated accessibility testing. Tools can detect many technical issues, but they cannot fully establish whether a digital product works well for someone using a screen reader or another assistive technology.

We also had to consider how to keep financial operations and workflow decisions reliable. These operations need predictable application logic, clear state management, and appropriate verification rather than relying on model-generated decisions.

These challenges shaped our approach: accessibility, accountability, and human oversight must be considered from the beginning.

Accomplishments that we're proud of

We are proud to have developed the concept and technical foundation for Ability Bazaar as a prototype addressing two connected problems: inaccessible digital experiences and barriers to participation in technology-related work.

We brought together AI-assisted coordination, automated accessibility evidence, a human-review workflow, and a marketplace model in one project.

We are particularly proud of the emphasis on responsible AI boundaries. The design keeps financial authority separate from the language model and places human approval at the center of customer-facing accessibility work.

Rather than claiming that the prototype has already solved these problems, we see it as a foundation for testing the approach, identifying limitations, and learning what would be required to make it dependable in real-world use.

What we learned

This project helped us recognize that using AI effectively involves more than generating useful output. It also requires clear responsibilities, reliable application logic, and ways to identify and correct mistakes.

We learned that automated accessibility tools and human testing serve different purposes. Automation can help identify potential problems, while people with disabilities can provide essential feedback about practical usability.

We also learned that a social-impact concept needs a way to measure its actual outcomes. Building a marketplace prototype does not, by itself, demonstrate that paid opportunities have been created or that accessibility has improved. Those outcomes must be evaluated with evidence as the project develops.

Most importantly, we learned to treat AI as an assistant to human expertise, not a replacement for it.

What's next for Ability Bazaar

Our next steps are to continue developing and evaluating the prototype, validate its core workflows, and determine which features are reliable enough for real-world testing.

We want to improve the clarity of accessibility audit results, test the AI-assisted workflow, and establish a practical review process involving people with disabilities.

We also plan to investigate how the platform could measure outcomes responsibly, including accessibility issues identified and resolved, the usability of delivered products, and any paid work actually completed through the platform.

Our longer-term vision is to explore whether a trustworthy accessibility marketplace can help organizations build more inclusive digital products while giving people with disabilities a meaningful role in creating and evaluating them.

Ability Bazaar is a step toward exploring that possibility—not a claim that the journey is complete.