How it works
Operable Labs reads a vendor’s VPAT and tells you where the product falls short of the accessibility standard you choose.
- You give us the VPAT. Upload the vendor’s PDF, or search by product name and pick the right document from the results. It must be one the vendor has made public.
- A model reads it. Google’s Gemini copies each row of the VPAT’s tables into structured data: the requirement, the vendor’s claim and their remarks. We then check that the remarks really appear in the document.
- Fixed rules find the gaps. Ordinary code, not the model, compares the claims with the standard’s list of requirements from the W3C and sets each severity and due date. The same VPAT always gives the same findings.
- The model drafts the plan. For each gap it writes who is affected, what to ask the vendor and what you can do meanwhile, plus a summary and an email you can send. It cannot add, remove or re-rank a finding.
What you get
- A one-sentence answer you can repeat to your manager.
- A plan of action and milestones (POA&M), with one item per gap.
- An Excel file to track it, a printable report, and a draft email to the vendor.
What it cannot tell you
The review is based on the vendor’s own claims. The product is not tested. A clean result means the vendor reports no gaps, not that there are none. This is not a certification of ADA compliance and not legal advice. Results are AI-assisted, so check them against the source document.
Using your own AI
By default reviews run on our Google Gemini service. If your team has an API key for Google Gemini, OpenAI or Anthropic Claude, you can use it instead: open “Advanced” on the start page. Your reviews then run on your own account with that provider. The findings are decided by the same fixed rules whichever model reads the document.
Limits
During early access, each person can run a set number of reviews and searches a day. It is not available in the European Economic Area, the United Kingdom or Switzerland.